baseline_channelAcquisition.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 函数"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### import"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import os\n",
    "os.environ[\"CUDA_DEVICE_ORDER\"] = \"PCI_BUS_ID\"\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"  # 使用第二块GPU（从0开始）\n",
    "os.environ['KMP_DUPLICATE_LIB_OK']='True'\n",
    "import random\n",
    "import torch.nn.functional as F\n",
    "import torchvision\n",
    "import numpy as np\n",
    "from math import *\n",
    "import matplotlib.pyplot as plt\n",
    "from torch.autograd import Variable\n",
    "from IPython import display\n",
    "import torch.utils.data as Data\n",
    "import torch.nn as nn\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "from matplotlib.ticker import LinearLocator, FormatStrFormatter\n",
    "from matplotlib import cm\n",
    "from scipy.linalg import block_diag\n",
    "import datetime\n",
    "from torch.nn.utils import *\n",
    "import cvxpy as cp\n",
    "from Transformer_model import *\n",
    "from RIS import *\n",
    "\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 参数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "fc = 1.5e11; #150GHz  \n",
    "\n",
    "# BW = 1e9;    # 1GHz\n",
    "\n",
    "BW = 1e10;    # 10GHz  \n",
    "\n",
    "Nc = 64;     # 64个子载波  \n",
    "\n",
    "\n",
    "M = 2; #M×M子阵列行列数量\n",
    "N = 8; #每个子阵N×N天线\n",
    "c = 3e8;  \n",
    "lambda_c = c/fc;  \n",
    "D_sub = 100*lambda_c;  #子阵中心间隔 0.2m \n",
    "D_ant = lambda_c/2;    #天线间隔     2mm  \n",
    "R = M*D_sub**2/lambda_c;#距离         40m   满足LOS-MIMO距离  \n",
    "\n",
    "M_ant = M*M*N*N;\n",
    "\n",
    "h1 = 10;  #BS与RIS高度\n",
    "h2 = 1.5; #用户高度  此项作废，因为改为了每个散射体0-4米之间分布\n",
    "\n",
    "L  = 8  #观测时隙数\n",
    "Lp  = 1  #多径数\n",
    "sigma = 1e5\n",
    "B = 128\n",
    "\n",
    "K = M*M\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "H_BR = Channel_BS_RIS(param_list) * 1e6"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 定义csiNet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def mish(x):\n",
    "    x = x * (torch.tanh(F.softplus(x)))\n",
    "    return x\n",
    "class Mish(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "#         print(\"Mish activation loaded...\")\n",
    "    def forward(self,x):\n",
    "        x = x * (torch.tanh(F.softplus(x)))\n",
    "        return x\n",
    "class RES_BLOCK(nn.Module): #输入信道输出 量化后的B比特反馈信息\n",
    "    def __init__(self,channel_list):\n",
    "        super(RES_BLOCK,self).__init__()\n",
    "        self.conv1 = nn.Sequential(\n",
    "            nn.Conv2d(\n",
    "                in_channels = channel_list[0], #图片通道数\n",
    "                out_channels = channel_list[1],#filter数量\n",
    "                kernel_size = (5,5), #filter大小\n",
    "                stride=1,  #扫描步进\n",
    "                padding=(2,2), #周围围上2圈0 使得输出的宽和高和之前一样不变小\n",
    "            ),\n",
    "            nn.BatchNorm2d(channel_list[1], eps=1e-05, momentum=0.1, affine=True),\n",
    "            Mish(),\n",
    "        )\n",
    "        self.conv2 = nn.Sequential(\n",
    "            nn.Conv2d(\n",
    "                in_channels = channel_list[1], #图片通道数\n",
    "                out_channels = channel_list[2],#filter数量\n",
    "                kernel_size = (5,5), #filter大小\n",
    "                stride=1,  #扫描步进\n",
    "                padding=(2,2), #周围围上2圈0 使得输出的宽和高和之前一样不变小\n",
    "            ),\n",
    "            nn.BatchNorm2d(channel_list[2], eps=1e-05, momentum=0.1, affine=True),\n",
    "            Mish(),\n",
    "        )\n",
    "        self.conv3 = nn.Sequential(\n",
    "            nn.Conv2d(\n",
    "                in_channels = channel_list[2], #图片通道数\n",
    "                out_channels = channel_list[0],#filter数量\n",
    "                kernel_size = (5,5), #filter大小\n",
    "                stride=1,  #扫描步进\n",
    "                padding=(2,2), #周围围上2圈0 使得输出的宽和高和之前一样不变小\n",
    "            ),\n",
    "            nn.BatchNorm2d(channel_list[0], eps=1e-05, momentum=0.1, affine=True),\n",
    "        )\n",
    "        \n",
    "        \n",
    "    def forward(self,x_ini):\n",
    "        \n",
    "        \n",
    "        x = self.conv1(x_ini)\n",
    "        x = self.conv2(x)\n",
    "        x = self.conv3(x)\n",
    "        x = mish(x+x_ini)\n",
    "        return x\n",
    "\n",
    "def conv7x7(in_planes, out_planes, stride=1):\n",
    "    return nn.Conv2d(in_planes, out_planes, kernel_size=7, stride=stride,\n",
    "                     padding=3, bias=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#csiNet\n",
    "class CSI_EN(nn.Module): #单层\n",
    "    def __init__(self, param_list): #H [batch,G,K]为角度域信道, A [M,G]为观测矩阵 G为原始稀疏数据长度，T为迭代次数，K为子载波数也就是MMV的数量，M为观测维度\n",
    "        super(CSI_EN,self).__init__()\n",
    "        \n",
    "#         param_list = [fc,B,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L]\n",
    "\n",
    "        fc = param_list[0]\n",
    "        BW  = param_list[1]\n",
    "        Nc = param_list[2]\n",
    "        M  = param_list[3]\n",
    "        N  = param_list[4]\n",
    "        D_sub = param_list[5]\n",
    "        D_ant = param_list[6]\n",
    "        R = param_list[7]\n",
    "        M_ant = param_list[8]\n",
    "        h1 = param_list[9]\n",
    "        h2 = param_list[10]\n",
    "        L =  param_list[11]\n",
    "        sigma = param_list[12]\n",
    "        B = param_list[13]\n",
    "        \n",
    "\n",
    "        \n",
    "\n",
    "\n",
    "        #############\n",
    "        self.B_quan = 2\n",
    "        self.res1  = RES_BLOCK([2,16,32])\n",
    "        self.res2  = RES_BLOCK([2,16,32])\n",
    "        self.FC = nn.Linear(2*Nc*M_ant,B//self.B_quan)#全连接\n",
    "        self.bn = nn.BatchNorm1d(B//self.B_quan)\n",
    "        self.QL = QuantizationLayer(self.B_quan)\n",
    "        \n",
    "        \n",
    "        \n",
    "        \n",
    "        \n",
    "    def forward(self, param_list, H_BR, H_RU):\n",
    "        #H_BR [ M_ant, M_ant]  H_RU [batch,Nc,M_ant]\n",
    "        # param_list = [fc,B,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,SNR]\n",
    "        fc = param_list[0]\n",
    "        B  = param_list[1]\n",
    "        Nc = param_list[2]\n",
    "        M  = param_list[3]\n",
    "        N  = param_list[4]\n",
    "        D_sub = param_list[5]\n",
    "        D_ant = param_list[6]\n",
    "        R = param_list[7]\n",
    "        M_ant = param_list[8]\n",
    "        h1 = param_list[9]\n",
    "        h2 = param_list[10]\n",
    "        L =  param_list[11]\n",
    "        sigma = param_list[12]\n",
    "        B = param_list[13]\n",
    "        \n",
    "        H_real = torch.zeros(H_RU.shape[0],2,Nc,M_ant).cuda()#总观测【batch,Nc,L】\n",
    "        H_real[:,0,:,:] = torch.real(H_RU)\n",
    "        H_real[:,1,:,:] = torch.imag(H_RU)\n",
    "        \n",
    "        x = (self.res1(H_real))\n",
    "        x = (self.res2(x))\n",
    "\n",
    "\n",
    "        \n",
    "        x = x.reshape(-1,2*Nc*M_ant)\n",
    "        x = self.FC(x)\n",
    "        x = self.bn(x)\n",
    "        x = torch.sigmoid(x)\n",
    "        x = self.QL(x)\n",
    "        return x\n",
    "\n",
    "class CSI_DE(nn.Module): #单层\n",
    "    def __init__(self, param_list): #H [batch,G,K]为角度域信道, A [M,G]为观测矩阵 G为原始稀疏数据长度，T为迭代次数，K为子载波数也就是MMV的数量，M为观测维度\n",
    "        super(CSI_DE,self).__init__()\n",
    "        \n",
    "#         param_list = [fc,B,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L]\n",
    "\n",
    "        fc = param_list[0]\n",
    "        BW  = param_list[1]\n",
    "        Nc = param_list[2]\n",
    "        M  = param_list[3]\n",
    "        N  = param_list[4]\n",
    "        D_sub = param_list[5]\n",
    "        D_ant = param_list[6]\n",
    "        R = param_list[7]\n",
    "        M_ant = param_list[8]\n",
    "        h1 = param_list[9]\n",
    "        h2 = param_list[10]\n",
    "        L =  param_list[11]\n",
    "        sigma = param_list[12]\n",
    "        B = param_list[13]\n",
    "        \n",
    "        self.B_quan = 2\n",
    "        self.DQL = DequantizationLayer(self.B_quan)\n",
    "        self.FC1 = nn.Linear(B//self.B_quan,2*Nc*M_ant)#全连接\n",
    "        self.bn1 = nn.BatchNorm1d(2*Nc*M_ant)\n",
    "        \n",
    "        \n",
    "        self.res1  = RES_BLOCK([2,16,32])\n",
    "        self.res2  = RES_BLOCK([2,16,32])\n",
    "        self.res3  = RES_BLOCK([2,16,32])\n",
    "        self.res4  = RES_BLOCK([2,16,32])\n",
    "        self.out_cov = conv7x7(2, 2)\n",
    "        \n",
    "        \n",
    "        \n",
    "        \n",
    "    def forward(self, param_list,  x):\n",
    "        #H_BR [ M_ant, M_ant]  H_RU [batch,Nc,M_ant]\n",
    "        # param_list = [fc,B,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,SNR]\n",
    "        fc = param_list[0]\n",
    "        B  = param_list[1]\n",
    "        Nc = param_list[2]\n",
    "        M  = param_list[3]\n",
    "        N  = param_list[4]\n",
    "        D_sub = param_list[5]\n",
    "        D_ant = param_list[6]\n",
    "        R = param_list[7]\n",
    "        M_ant = param_list[8]\n",
    "        h1 = param_list[9]\n",
    "        h2 = param_list[10]\n",
    "        L =  param_list[11]\n",
    "        sigma = param_list[12]\n",
    "        B = param_list[13]\n",
    "        \n",
    "        x = self.DQL(x)-0.5\n",
    "        x = self.FC1(x)\n",
    "        x = self.bn1(x)\n",
    "        x = mish(x)\n",
    "        \n",
    "        x = x.reshape(-1,2,Nc,M_ant)\n",
    "        x = (self.res1(x))\n",
    "        x = (self.res2(x))\n",
    "        x = (self.res3(x))\n",
    "        x = (self.res4(x))\n",
    "        x = self.out_cov(x)\n",
    "        \n",
    "        H_RU = x[:,0,:,:] + 1j*x[:,1,:,:]\n",
    "        return H_RU\n",
    "\n",
    "class csiNet(nn.Module): #单层\n",
    "    def __init__(self, param_list): #H [batch,G,K]为角度域信道, A [M,G]为观测矩阵 G为原始稀疏数据长度，T为迭代次数，K为子载波数也就是MMV的数量，M为观测维度\n",
    "        super(csiNet,self).__init__()\n",
    "        self.encoder = CSI_EN(param_list)\n",
    "        self.decoder = CSI_DE(param_list)\n",
    "        \n",
    "        \n",
    "    def forward(self, param_list,H_BR, H_RU1):\n",
    "        \n",
    "        out = self.encoder(param_list,H_BR,H_RU1)\n",
    "        H_RU2 = self.decoder(param_list,out)\n",
    "        #H_BR [ M_ant, M_ant]  H_RU [batch,Nc,M_ant]\n",
    "        # param_list = [fc,B,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,SNR]\n",
    "\n",
    "        return H_RU2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_csiNet(param_list,num_batch, batch,STEPS,model_name_AE):\n",
    "    \n",
    "    fc = param_list[0]\n",
    "#     B  = param_list1[1]\n",
    "#     Nc = param_list1[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    Lp = param_list[15]\n",
    "\n",
    "    net = csiNet(param_list).cuda()\n",
    "        \n",
    "    optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, net.parameters()),lr=0.001)\n",
    "    \n",
    "    Ay = (M-1)*D_sub + (N-1)*(D_ant);#阵列长度\n",
    "    Az = Ay;\n",
    "\n",
    "    r_BS  = (torch.tensor([0,0,h1]) - torch.tensor([0,Ay/2,Az/2])).cuda()\n",
    "    r_RIS = (torch.tensor([R,0,h1]) - torch.tensor([0,Ay/2,Az/2])).cuda()\n",
    "\n",
    "    \n",
    "    ##随机生成信道\n",
    "    # H_BR = Channel_BS_RIS(param_list) * 1e6\n",
    "    \n",
    "    \n",
    "    \n",
    "    num_train = 0\n",
    "    train_nmse = 0\n",
    "    best_nmse = 1\n",
    "    start = datetime.datetime.now()\n",
    "    numStep = 1\n",
    "    warmup_steps = 4000\n",
    "    d_model = 256\n",
    "    for step in range(STEPS):\n",
    "        # optimizer.param_groups[0]['lr'] = (d_model**(-0.5))*min(numStep**(-0.5),numStep*warmup_steps**(-1.5))\n",
    "        numStep = numStep+1\n",
    "        net.train()\n",
    "        optimizer.zero_grad()\n",
    "        for num_b in range(num_batch):\n",
    "\n",
    "            alpha = 1e5\n",
    "            H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "\n",
    "            H_RU0 = net(param_list, H_BR, H_RU)        \n",
    "            loss = NMSE(H_RU0,H_RU)/num_batch\n",
    "            loss.backward()\n",
    "            train_nmse = train_nmse + loss\n",
    "\n",
    "        optimizer.step()\n",
    "        optimizer.zero_grad()\n",
    "        num_train = num_train + 1\n",
    "        \n",
    "        if (step+1)%50==0:\n",
    "            with torch.no_grad():\n",
    "                net.eval()\n",
    "                alpha = 1e5\n",
    "                H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "        \n",
    "                H_RU0 = net(param_list, H_BR, H_RU)\n",
    "        \n",
    "                \n",
    "                loss = NMSE(H_RU0,H_RU)\n",
    "#                 print(H_BR)\n",
    "#                 print(H_RU)\n",
    "\n",
    "                \n",
    "#                 print(R_UE0)\n",
    "#                 print(R_UE)\n",
    "#                 print(Theta_UE0)\n",
    "#                 print(Theta_UE)\n",
    "                \n",
    "#                 r_UE0 = net(param_list, H_BR, H_RU)\n",
    "#                 loss = loss_func(r_UE0, r_UE)\n",
    "#             print(r_UE0)\n",
    "#             print(r_UE)\n",
    "            \n",
    "            train_nmse = train_nmse/num_train\n",
    "            time0 =  datetime.datetime.now()-start\n",
    "            print('step:',step,'time',time0,'train NMSE %.5f' % train_nmse.cpu(),'test NMSE %.5f' % loss.cpu()) \n",
    "            if loss < best_nmse:\n",
    "                best_nmse = loss\n",
    "                torch.save(net, model_name_AE)\n",
    "                print('Model saved!')\n",
    "            num_train = 0\n",
    "            train_nmse = 0\n",
    "            start = datetime.datetime.now()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 压缩感知算法"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def SOMP_GMMV(A,Y,K1,K2): #A为感知矩阵[M,L,N]，Y为观测信号矩阵[batch,M,L,1]，K1为用户数，K2为稀疏度,恢复稀疏信号X[batch,M,N,1]\n",
    "    #Y为观测信号矩阵[batch,L,M]，K为稀疏度,恢复稀疏信号X[batch,N,M]\n",
    "    batch = Y.shape[0]\n",
    "    N = A.shape[2] #稀疏信号总长度\n",
    "    M = Y.shape[1] #MMV数\n",
    "    N_sub = N//K1  #子阵稀疏信号\n",
    "    L = Y.shape[2] #压缩观测数\n",
    "    R = Y #残差 [batch,M,L,1]\n",
    "    ind_set = torch.floor(torch.zeros(batch,K2,1)).cuda().long()\n",
    "\n",
    "    AA = A.repeat(batch, 1, 1 ,1) #取索引的时候gather需要输入每一个样本的 #[batch,M,L,N_sub]\n",
    "\n",
    "    for a in range(K2):\n",
    "        C = Hermitian(A) @ R #[batch,M,N,1]\n",
    "        C = C.reshape(batch,M,N)\n",
    "        C2 = Hermitian(C) @ C\n",
    "        D = torch.abs(torch.diagonal(C2,dim1=-2, dim2=-1))\n",
    "        ind = torch.max(D, 1)[1].reshape(-1,1,1,1)   #[batch,1,1,1]\n",
    "        ind_set[:,a,:] = ind.reshape(-1,1)\n",
    "        if a == 0:\n",
    "            Supp = torch.gather(AA, 3, ind.repeat(1, M, L, 1))  #按照batch取索引,获得支撑集[batch,M,L,1],最终为[batch,M,L,K]\n",
    "        else:\n",
    "            Supp = torch.cat([Supp,torch.gather(AA, 3, ind.repeat(1, M, L, 1))],3)\n",
    "        X_LS = torch.pinverse(Supp) @ Y #[batch,M,K,1]\n",
    "        R = Y - Supp @ X_LS\n",
    "    X_hat = torch.zeros(batch,M,N,1).cuda() + 0j\n",
    "    for b in range(batch):  #最终的输出X_hat [batch, N, M]\n",
    "        for k in range(K2):\n",
    "            X_hat[b,:,ind_set[b,k,0],:] = X_LS[b,:,k,:]\n",
    "    return X_hat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#考虑不同子载波不同观测\n",
    "def SS_OMP_GMMV(A,Y,K1,K2): #A为感知矩阵[M,L,N]，Y为观测信号矩阵[batch,M,L,1]，K为稀疏度,恢复稀疏信号X[batch,M,N,1],SA为子阵数\n",
    "    batch = Y.shape[0]\n",
    "    N = A.shape[2] #稀疏信号总长度\n",
    "    M = Y.shape[1] #MMV数\n",
    "    N_sub = N//K1  #子阵稀疏信号\n",
    "    L = Y.shape[2] #压缩观测数\n",
    "    R = Y #残差 [batch,M,L,1]\n",
    "    ind_set = torch.floor(torch.zeros(batch,K2,1)).cuda().long()\n",
    "    for b in range(K2//K1):\n",
    "        for a in range(K1):\n",
    "            AA_k = A[:,:,(a*N_sub):((a+1)*N_sub)] #[M,L,N_sub]\n",
    "            C = Hermitian(AA_k) @ R #[batch,M,N_sub,1]\n",
    "            C = C.reshape(batch,M,N_sub)\n",
    "            if a == 0:\n",
    "                C2 = Hermitian(C) @ C\n",
    "            else:\n",
    "                C2 = C2 + Hermitian(C) @ C\n",
    "\n",
    "        \n",
    "        D = torch.abs(torch.diagonal(C2,dim1=-2, dim2=-1))\n",
    "        for a in range(K):\n",
    "            AA_k = A[:,:,(a*N_sub):((a+1)*N_sub)] #[M,L,N_sub]\n",
    "            ind = torch.max(D, 1)[1].reshape(-1,1,1,1)   #[batch,1,1,1]\n",
    "            ind_set[:,b*K1+a,:] = ind.reshape(-1,1) + a*N_sub\n",
    "            AA_k = AA_k.repeat(batch, 1, 1 ,1) #取索引的时候gather需要输入每一个样本的 #[batch,M,L,N_sub]\n",
    "            if b*K1+a == 0:\n",
    "                Supp = torch.gather(AA_k, 3, ind.repeat(1, M, L, 1))  #按照batch取索引,获得支撑集[batch,M,L,1],最终为[batch,M,L,K]\n",
    "            else:\n",
    "                Supp = torch.cat([Supp,torch.gather(AA_k, 3, ind.repeat(1, M, L, 1)) ],3) #[batch,M,L,K]\n",
    "        X_LS = torch.pinverse(Supp) @ Y #[batch,M,K,1]\n",
    "        R = Y - Supp @ X_LS\n",
    "\n",
    "    X_hat = torch.zeros(batch,M,N,1).cuda() + 0j\n",
    "    for b in range(batch):  #最终的输出X_hat [batch, M, N, 1]\n",
    "        for k in range(K):\n",
    "            X_hat[b,:,ind_set[b,k,0],:] = X_LS[b,:,k,:]\n",
    "    return X_hat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigmoid(x):\n",
    "    return torch.sigmoid(torch.from_numpy(x)).numpy()\n",
    "\n",
    "# MMV-AMP algorithm Channel Estimation\n",
    "def AMP_GMMV_Meng(y, H, IteN):\n",
    "\n",
    "# y [N_RAs,N_dim] 观测值\n",
    "# H [N_dim,N_RAs,N_UEs] 观测矩阵\n",
    "\n",
    "    N_RAs = H.shape[1]   #观测维度\n",
    "    N_UEs = H.shape[2]   #稀疏维度\n",
    "    N_dim = y.shape[1]   #MMV维度\n",
    "\n",
    "    y = y.astype(np.complex128)\n",
    "    H = H.astype(np.complex128)\n",
    " \n",
    "    import scipy.stats as st\n",
    "    import math\n",
    "    lam = N_RAs/N_UEs\n",
    "    c = np.arange(0.01, 10, 10/1024)\n",
    "    rho = (1 - 2 * N_UEs * ((1 + c**2) * st.norm.cdf(-c)-c * st.norm.pdf(c))/N_RAs)/(1 + c**2 - 2 * ((1 + c**2) * st.norm.cdf(-c) - c * st.norm.pdf(c)))\n",
    "    alpha = lam * np.amax(rho) * np.ones((N_UEs, N_dim)) ##[N_UEs, N_dim]\n",
    "\n",
    "    x_hat = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "    var_hat = np.ones((N_UEs, N_dim), dtype='float64')   ##[N_UEs, N_dim]\n",
    "\n",
    "    V = np.ones((N_RAs, N_dim))  ##[N_RAs, N_dim]\n",
    "    Z = y + 0 ##[N_RAs, N_dim]\n",
    "    SNR0 = 100\n",
    "    sigma2 = np.linalg.norm(y)**2/((SNR0 + 1) * N_RAs * N_dim)  ##1\n",
    "    hvar = (np.linalg.norm(y)**2 - N_RAs * sigma2)/(N_dim * lam * np.amax(rho) * ((np.linalg.norm(H)**2)/N_dim) )  ##1\n",
    "    hmean = 0  ##1\n",
    "\n",
    "    pai = np.zeros((N_UEs, N_dim), dtype='float64')   ##\n",
    "\n",
    "    t = 1\n",
    "    tol = 1e-5\n",
    "    damp = 0.3\n",
    "\n",
    "    while t < IteN:\n",
    "        x_hat_new = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "        x_hat_pre = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "        Vi_tol = np.zeros((N_UEs, N_dim), dtype='float64')  ##[N_UEs, N_dim]\n",
    "        B_tol = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "        for a in range(N_dim):\n",
    "            x_hat_pre[:,a:a+1] = x_hat[:,a:a+1]\n",
    "            V_new = np.dot(np.abs(H[a])*np.abs(H[a]), var_hat[:,a:a+1])\n",
    "            Z_new = np.dot(H[a], x_hat[:,a:a+1]) - ((y[:,a:a+1] - Z[:,a:a+1]) / (sigma2 + V[:,a:a+1])) * V_new   # + 1e-8\n",
    "\n",
    "            Z_new = damp * Z[:,a:a+1] + (1 - damp) * Z_new\n",
    "            V_new = damp * V[:,a:a+1] + (1 - damp) * V_new\n",
    "\n",
    "            var1 = np.dot((np.abs(H[a])*np.abs(H[a])).T, 1/(sigma2 + V_new))\n",
    "            var2 = np.dot(H[a].conj().T, (y[:,a:a+1] - Z_new)/(sigma2 + V_new))\n",
    "\n",
    "            Ri = var2/(var1) + x_hat[:,a:a+1]\n",
    "            Vi = 1/(var1)\n",
    "            Vi_tol[:,a:a+1] = Vi\n",
    "\n",
    "            # A\n",
    "            A = (hvar * Vi)/(Vi + hvar)\n",
    "            # B\n",
    "            B = (hvar * Ri + Vi * hmean)/(Vi + hvar)\n",
    "            B_tol[:,a:a+1] = B\n",
    "            # lll\n",
    "            lll = np.log(Vi/(Vi + hvar))/2 + np.abs(Ri)**2/2/Vi - np.abs(Ri - hmean)**2/2/(Vi + hvar)\n",
    "            # pai\n",
    "            # pai[:,a:a+1] = alpha[:,a:a+1]/(alpha[:,a:a+1] + (1 - alpha[:,a:a+1]) * np.exp(-lll))\n",
    "            pai[:,a:a+1] = sigmoid(lll-np.log((1-alpha[:,a:a+1])/alpha[:,a:a+1]))\n",
    "\n",
    "            x_hat_new[:,a:a+1] = pai[:,a:a+1] * B\n",
    "\n",
    "            var_hat_new = pai[:,a:a+1]*(np.abs(B)**2 + Vi) - np.abs(x_hat_new[:,a:a+1])**2\n",
    "\n",
    "            x_hat[:,a:a+1] = x_hat_new[:,a:a+1]\n",
    "            var_hat[:,a:a+1] = var_hat_new\n",
    "            \n",
    "            V[:,a:a+1] = V_new\n",
    "            # Z[:,a:a+1] = Z_new + 0\n",
    "            # Z = 0\n",
    "            Z[:,a:a+1] = Z_new.copy() + 0\n",
    "\n",
    "        # 噪声方差 自学习\n",
    "        sigma2 = ((np.abs(y-Z)**2)/(np.abs(1+V/sigma2)**2)+sigma2*V/(V+sigma2)).mean()\n",
    "        # 先验均值 自学习\n",
    "        hmean = (np.sum(pai * B_tol, axis=0)/np.sum(pai, axis=0)).mean()\n",
    "        # 先验方差 自学习\n",
    "        hvar = (np.sum(pai * (np.abs(hmean - B_tol)**2 + Vi_tol), axis=0)/np.sum(pai, axis=0)).mean()\n",
    "        # 活跃因子 EM学习\n",
    "        pai[pai < 1e-5] = 1e-5\n",
    "        pai[pai > 1 - 1e-5] = 1 - 1e-5\n",
    "        alpha_new00 = (np.sum(pai, axis=1)/N_dim).reshape(-1, 1)\n",
    "\n",
    "        ###取平均\n",
    "\n",
    "        alpha_new = np.tile(alpha_new00, (1, N_dim))\n",
    "        alpha = alpha_new\n",
    "\n",
    "        NMSE_iter = np.sum(np.abs(x_hat_pre - x_hat_new)**2)/np.sum(np.abs(x_hat_new)**2)\n",
    "        if NMSE_iter < tol:\n",
    "            break\n",
    "        xx = x_hat.transpose(1,0).reshape(N_dim,N_UEs,1)\n",
    "        y_hat = H @ xx #[N_dim,N_RAs,1]\n",
    "        y_hat = y_hat.transpose(1,0,2).reshape(N_RAs,N_dim)\n",
    "        NMSE_iter = np.sum(np.abs(y_hat - y)**2)/np.sum(np.abs(y)**2)\n",
    "        if NMSE_iter > 10:\n",
    "            x_hat = x_hat - x_hat\n",
    "            break\n",
    "\n",
    "\n",
    "        t = t + 1\n",
    "    return x_hat, var_hat, alpha, t\n",
    "# MMV-AMP algorithm Channel Estimation\n",
    "def BAMP_GMMV_Meng(y, H, IteN):\n",
    "\n",
    "# y [N_RAs,N_dim] 观测值\n",
    "# H [N_dim,N_RAs,N_UEs] 观测矩阵\n",
    "\n",
    "    N_RAs = H.shape[1]   #观测维度\n",
    "    N_UEs = H.shape[2]   #稀疏维度\n",
    "    N_dim = y.shape[1]   #MMV维度\n",
    "\n",
    "    y = y.astype(np.complex128)\n",
    "    H = H.astype(np.complex128)\n",
    " \n",
    "    import scipy.stats as st\n",
    "    import math\n",
    "    lam = N_RAs/N_UEs\n",
    "    c = np.arange(0.01, 10, 10/1024)\n",
    "    rho = (1 - 2 * N_UEs * ((1 + c**2) * st.norm.cdf(-c)-c * st.norm.pdf(c))/N_RAs)/(1 + c**2 - 2 * ((1 + c**2) * st.norm.cdf(-c) - c * st.norm.pdf(c)))\n",
    "    alpha = lam * np.amax(rho) * np.ones((N_UEs, N_dim)) ##[N_UEs, N_dim]\n",
    "\n",
    "    x_hat = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "    var_hat = np.ones((N_UEs, N_dim), dtype='float64')   ##[N_UEs, N_dim]\n",
    "\n",
    "    V = np.ones((N_RAs, N_dim))  ##[N_RAs, N_dim]\n",
    "    Z = y + 0 ##[N_RAs, N_dim]\n",
    "    SNR0 = 100\n",
    "    sigma2 = np.linalg.norm(y)**2/((SNR0 + 1) * N_RAs * N_dim)  ##1\n",
    "    hvar = (np.linalg.norm(y)**2 - N_RAs * sigma2)/(N_dim * lam * np.amax(rho) * ((np.linalg.norm(H)**2)/N_dim) )  ##1\n",
    "    hmean = 0  ##1\n",
    "\n",
    "    pai = np.zeros((N_UEs, N_dim), dtype='float64')   ##\n",
    "\n",
    "    t = 1\n",
    "    tol = 1e-5\n",
    "    damp = 0.3\n",
    "\n",
    "    while t < IteN:\n",
    "        x_hat_new = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "        x_hat_pre = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "        Vi_tol = np.zeros((N_UEs, N_dim), dtype='float64')  ##[N_UEs, N_dim]\n",
    "        B_tol = np.zeros((N_UEs, N_dim), dtype='complex128')  ##[N_UEs, N_dim]\n",
    "        for a in range(N_dim):\n",
    "            x_hat_pre[:,a:a+1] = x_hat[:,a:a+1]\n",
    "            V_new = np.dot(np.abs(H[a])*np.abs(H[a]), var_hat[:,a:a+1])\n",
    "            Z_new = np.dot(H[a], x_hat[:,a:a+1]) - ((y[:,a:a+1] - Z[:,a:a+1]) / (sigma2 + V[:,a:a+1])) * V_new   # + 1e-8\n",
    "\n",
    "            Z_new = damp * Z[:,a:a+1] + (1 - damp) * Z_new\n",
    "            V_new = damp * V[:,a:a+1] + (1 - damp) * V_new\n",
    "\n",
    "            var1 = np.dot((np.abs(H[a])*np.abs(H[a])).T, 1/(sigma2 + V_new))\n",
    "            var2 = np.dot(H[a].conj().T, (y[:,a:a+1] - Z_new)/(sigma2 + V_new))\n",
    "\n",
    "            Ri = var2/(var1) + x_hat[:,a:a+1]\n",
    "            Vi = 1/(var1)\n",
    "            Vi_tol[:,a:a+1] = Vi\n",
    "\n",
    "            # A\n",
    "            A = (hvar * Vi)/(Vi + hvar)\n",
    "            # B\n",
    "            B = (hvar * Ri + Vi * hmean)/(Vi + hvar)\n",
    "            B_tol[:,a:a+1] = B\n",
    "            # lll\n",
    "            lll = np.log(Vi/(Vi + hvar))/2 + np.abs(Ri)**2/2/Vi - np.abs(Ri - hmean)**2/2/(Vi + hvar)\n",
    "            # pai\n",
    "            # pai[:,a:a+1] = alpha[:,a:a+1]/(alpha[:,a:a+1] + (1 - alpha[:,a:a+1]) * np.exp(-lll))\n",
    "            pai[:,a:a+1] = sigmoid(lll-np.log((1-alpha[:,a:a+1])/alpha[:,a:a+1]))\n",
    "\n",
    "            x_hat_new[:,a:a+1] = pai[:,a:a+1] * B\n",
    "\n",
    "            var_hat_new = pai[:,a:a+1]*(np.abs(B)**2 + Vi) - np.abs(x_hat_new[:,a:a+1])**2\n",
    "\n",
    "            x_hat[:,a:a+1] = x_hat_new[:,a:a+1]\n",
    "            var_hat[:,a:a+1] = var_hat_new\n",
    "            \n",
    "            V[:,a:a+1] = V_new\n",
    "            # Z[:,a:a+1] = Z_new + 0\n",
    "            # Z = 0\n",
    "            Z[:,a:a+1] = Z_new.copy() + 0\n",
    "\n",
    "        # 噪声方差 自学习\n",
    "        sigma2 = ((np.abs(y-Z)**2)/(np.abs(1+V/sigma2)**2)+sigma2*V/(V+sigma2)).mean()\n",
    "        # 先验均值 自学习\n",
    "        hmean = (np.sum(pai * B_tol, axis=0)/np.sum(pai, axis=0)).mean()\n",
    "        # 先验方差 自学习\n",
    "        hvar = (np.sum(pai * (np.abs(hmean - B_tol)**2 + Vi_tol), axis=0)/np.sum(pai, axis=0)).mean()\n",
    "        # 活跃因子 EM学习\n",
    "        pai[pai < 1e-5] = 1e-5\n",
    "        pai[pai > 1 - 1e-5] = 1 - 1e-5\n",
    "        alpha_new00 = (np.sum(pai, axis=1)/N_dim).reshape(-1, 1)\n",
    "\n",
    "        ###取平均\n",
    "\n",
    "        alpha_new = np.tile(alpha_new00, (1, N_dim))\n",
    "        alpha = alpha_new\n",
    "\n",
    "        alpha = alpha.reshape(K, N_UEs//K, N_dim)\n",
    "        alpha = alpha.mean(axis=0)\n",
    "        alpha = np.tile(alpha, (K, 1))\n",
    "\n",
    "\n",
    "\n",
    "        NMSE_iter = np.sum(np.abs(x_hat_pre - x_hat_new)**2)/np.sum(np.abs(x_hat_new)**2)\n",
    "        if NMSE_iter < tol:\n",
    "            break\n",
    "        xx = x_hat.transpose(1,0).reshape(N_dim,N_UEs,1)\n",
    "        y_hat = H @ xx #[N_dim,N_RAs,1]\n",
    "        y_hat = y_hat.transpose(1,0,2).reshape(N_RAs,N_dim)\n",
    "        NMSE_iter = np.sum(np.abs(y_hat - y)**2)/np.sum(np.abs(y)**2)\n",
    "        if NMSE_iter > 10:\n",
    "            x_hat = x_hat - x_hat\n",
    "            break\n",
    "\n",
    "\n",
    "        t = t + 1\n",
    "    return x_hat, var_hat, alpha, t\n",
    "\n",
    "# MMV-AMP algorithm Channel Estimation\n",
    "def AMP_GMMV(H, y, IteN):#A为感知矩阵[M,L,N]，Y为观测信号矩阵[batch,M,L,1]，K为稀疏度,恢复稀疏信号X[batch,M,N,1],SA为子阵数\n",
    "    L = H.shape[1]   #观测维度\n",
    "    N = H.shape[2]   #稀疏维度\n",
    "    M = y.shape[1]   #MMV维度\n",
    "    batch = y.shape[0]\n",
    "    X_hat = torch.zeros(batch,N,M).cuda() + 0j#[batch,N,M]\n",
    "    for i in range(batch):\n",
    "        yy = y[i].reshape(M,L)\n",
    "        yy = yy.permute(1,0) #[L,M]\n",
    "        a,b,c,d = AMP_GMMV_Meng(yy.cpu().numpy(),H.cpu().numpy(),20) # Y为观测信号矩阵[batch,L,M]\n",
    "        # a,b,c,d = AMP_MMV_Meng(yy.cpu().numpy(),H[0].cpu().numpy(),50) # Y为观测信号矩阵[batch,L,M]\n",
    "        X_hat[i] = torch.from_numpy(a).cuda()\n",
    "    X_hat = X_hat.permute(0,2,1)\n",
    "    X_hat = X_hat.reshape(batch,M,N,1)\n",
    "    return X_hat\n",
    "def BAMP_GMMV(H, y, IteN):#A为感知矩阵[M,L,N]，Y为观测信号矩阵[batch,M,L,1]，K为稀疏度,恢复稀疏信号X[batch,M,N,1],SA为子阵数\n",
    "    L = H.shape[1]   #观测维度\n",
    "    N = H.shape[2]   #稀疏维度\n",
    "    M = y.shape[1]   #MMV维度\n",
    "    batch = y.shape[0]\n",
    "    X_hat = torch.zeros(batch,N,M).cuda() + 0j#[batch,N,M]\n",
    "    for i in range(batch):\n",
    "        yy = y[i].reshape(M,L)\n",
    "        yy = yy.permute(1,0) #[L,M]\n",
    "        a,b,c,d = BAMP_GMMV_Meng(yy.cpu().numpy(),H.cpu().numpy(),20) # Y为观测信号矩阵[batch,L,M]\n",
    "        # a,b,c,d = AMP_MMV_Meng(yy.cpu().numpy(),H[0].cpu().numpy(),50) # Y为观测信号矩阵[batch,L,M]\n",
    "        X_hat[i] = torch.from_numpy(a).cuda()\n",
    "    X_hat = X_hat.permute(0,2,1)\n",
    "    X_hat = X_hat.reshape(batch,M,N,1)\n",
    "    return X_hat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def CE_SOMP_GMMV(param_list,batch,pho):\n",
    "    #Phi 为观测矩阵 Theta为感知矩阵\n",
    "\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "        S_digital = torch.randn(L,Nc,4,1).cuda()+1j*torch.randn(L,Nc,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = SOMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K1,K2)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "\n",
    "def CE_SOMP_MMV(param_list,batch,pho):\n",
    "    #Phi 为观测矩阵 Theta为感知矩阵\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "        S_digital = torch.randn(L,1,4,1).cuda()+1j*torch.randn(L,1,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(1,M_ant,1)  #导频s [1,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(1))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = SOMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K1,K2)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "\n",
    "def CE_BSOMP_GMMV(param_list,batch,pho): #K1用户数 K2稀疏度\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "        S_digital = torch.randn(L,Nc,4,1).cuda()+1j*torch.randn(L,Nc,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = SS_OMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K1,K2)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "\n",
    "\n",
    "def CE_BSOMP_MMV(param_list,batch,pho): #K1用户数 K2稀疏度\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "        S_digital = torch.randn(L,1,4,1).cuda()+1j*torch.randn(L,1,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(1,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(1))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = SS_OMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K1,K2)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "def CE_AMP_MMV(param_list,batch,pho):\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "        S_digital = torch.randn(L,1,4,1).cuda()+1j*torch.randn(L,1,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(1,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(1))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = AMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "def CE_BAMP_MMV(param_list,batch,pho):\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "        S_digital = torch.randn(L,1,4,1).cuda()+1j*torch.randn(L,1,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(1,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(1))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = BAMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "\n",
    "def CE_BAMP_GMMV(param_list,batch,pho):\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "        S_digital = torch.randn(L,Nc,4,1).cuda()+1j*torch.randn(L,Nc,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = BAMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)\n",
    "\n",
    "def CE_AMP_GMMV(param_list,batch,pho):\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "    fc = param_list[0]\n",
    "    BW  = param_list[1]\n",
    "    Nc = param_list[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    K = param_list[14]\n",
    "    K1 = K\n",
    "    K2 = K\n",
    "    N_2 = pho*N \n",
    "\n",
    "    alpha = 1e5/sqrt(Lp)\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 1\n",
    "\n",
    "\n",
    "    NMSE = 0\n",
    "\n",
    "    for step in range(Step):\n",
    "\n",
    "        A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "        A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "        for i in range(N_2):\n",
    "            for j in range(N_2):\n",
    "                a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "                a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "                a = kron_add(a_h,a_v).reshape(N*N)\n",
    "                A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "        for i in range(M*M):\n",
    "            A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A\n",
    "\n",
    "        H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "        S_digital = torch.randn(L,Nc,4,1).cuda()+1j*torch.randn(L,Nc,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "        S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "        Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "\n",
    "        Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "        Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "        for l in range(L):\n",
    "            s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "            \n",
    "            s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "            s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "        #             print(s)\n",
    "            n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "            \n",
    "        #             print(n)\n",
    "            \n",
    "            Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "            a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "            Phi[:,l,:] = a[:,0,:]\n",
    "        #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "            y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "            Y[:,:,l] = y\n",
    "\n",
    "        Theta = Phi @ A_t\n",
    "\n",
    "        H_t_hat = AMP_GMMV(Theta,Y.reshape(batch,Nc,L,1),K)\n",
    "        H_hat = A_t @ H_t_hat\n",
    "        NMSE = NMSE + ((torch.mean(torch.abs((H_RU-H_hat.reshape(batch,Nc,M_ant))*(H_RU-H_hat.reshape(batch,Nc,M_ant)))))/(torch.mean(torch.abs(H_RU*H_RU))))\n",
    "    NMSE = NMSE/Step\n",
    "    return H_RU,H_hat\n",
    "    # print(NMSE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test_channelAcquisition(param_list,batch,model_name_AE):\n",
    "    \n",
    "    fc = param_list[0]\n",
    "#     B  = param_list1[1]\n",
    "#     Nc = param_list1[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    Lp = param_list[15]\n",
    "\n",
    "    net = csiNet(param_list).cuda()\n",
    "    net = torch.load(model_name_AE)\n",
    "        \n",
    "    \n",
    "    Ay = (M-1)*D_sub + (N-1)*(D_ant);#阵列长度\n",
    "    Az = Ay;\n",
    "\n",
    "    r_BS  = (torch.tensor([0,0,h1]) - torch.tensor([0,Ay/2,Az/2])).cuda()\n",
    "    r_RIS = (torch.tensor([R,0,h1]) - torch.tensor([0,Ay/2,Az/2])).cuda()\n",
    "\n",
    "    \n",
    "    ##随机生成信道\n",
    "    # H_BR = Channel_BS_RIS(param_list) * 1e6\n",
    "    \n",
    "    \n",
    "    \n",
    "    num_train = 0\n",
    "    train_nmse = 0\n",
    "    best_nmse = 1\n",
    "    start = datetime.datetime.now()\n",
    "    numStep = 1\n",
    "    warmup_steps = 4000\n",
    "    d_model = 256\n",
    "\n",
    "    with torch.no_grad():\n",
    "        net.eval()\n",
    "        alpha = 1e5/sqrt(Lp)\n",
    "\n",
    "        H_RU_SOMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_hat_SOMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_RU_BSOMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_hat_BSOMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_RU_AMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_hat_AMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "\n",
    "        H_RU_LAMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_hat_LAMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "\n",
    "\n",
    "\n",
    "        for ii in range(batch):\n",
    "\n",
    "            H_RU_SOMP11,H_hat_SOMP11 = CE_SOMP_GMMV(param_list,1,4)\n",
    "            H_RU_BSOMP11,H_hat_BSOMP11 = CE_BSOMP_GMMV(param_list,1,4)\n",
    "            H_RU_AMP11,H_hat_AMP11 = CE_AMP_GMMV(param_list,1,2)\n",
    "            \n",
    "\n",
    "            H_RU_LAMP11,H_hat_LAMP11 = Test_LAMP_GMMV(param_list, T, 1, 1, L,sigma)\n",
    "\n",
    "            H_hat_SOMP11 = H_hat_SOMP11.reshape(1,Nc,M_ant)\n",
    "            H_hat_BSOMP11 = H_hat_BSOMP11.reshape(1,Nc,M_ant)\n",
    "            H_hat_AMP11 = H_hat_AMP11.reshape(1,Nc,M_ant)\n",
    "\n",
    "            if NMSE(H_hat_SOMP11,H_RU_SOMP11)>1:\n",
    "                H_hat_SOMP11 = H_hat_SOMP11-H_hat_SOMP11\n",
    "            if NMSE(H_hat_BSOMP11,H_RU_BSOMP11)>1:\n",
    "                H_hat_BSOMP11 = H_hat_BSOMP11-H_hat_BSOMP11\n",
    "            if NMSE(H_hat_AMP11,H_RU_AMP11)>1:\n",
    "                H_hat_AMP11 = H_hat_AMP11-H_hat_AMP11\n",
    "            if NMSE(H_hat_LAMP11,H_RU_LAMP11)>1:\n",
    "                H_hat_LAMP11 = H_hat_LAMP11-H_hat_LAMP11\n",
    "\n",
    "            \n",
    "\n",
    "            H_RU_SOMP[ii,:,:] = H_RU_SOMP11[0,:,:]\n",
    "            H_hat_SOMP[ii,:,:] = H_hat_SOMP11[0,:,:]\n",
    "            H_RU_BSOMP[ii,:,:] = H_RU_BSOMP11[0,:,:]\n",
    "            H_hat_BSOMP[ii,:,:] = H_hat_BSOMP11[0,:,:]\n",
    "            H_RU_AMP[ii,:,:] = H_RU_AMP11[0,:,:]\n",
    "            H_hat_AMP[ii,:,:] = H_hat_AMP11[0,:,:]\n",
    "            H_RU_LAMP[ii,:,:] = H_RU_LAMP11[0,:,:]\n",
    "            H_hat_LAMP[ii,:,:] = H_hat_LAMP11[0,:,:]\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "        H_RU_SOMP0 = net(param_list, H_BR, H_RU_SOMP)\n",
    "        H_hat_SOMP0 = net(param_list, H_BR, H_hat_SOMP)\n",
    "        H_hat_BSOMP0 = net(param_list, H_BR, H_hat_BSOMP)\n",
    "        H_hat_AMP0 = net(param_list, H_BR, H_hat_AMP)\n",
    "        H_hat_LAMP0 = net(param_list, H_BR, H_hat_LAMP)\n",
    "\n",
    "\n",
    "        \n",
    "        NMSE1 = NMSE(H_RU_SOMP0,H_RU_SOMP)\n",
    "        NMSE2 = NMSE(H_hat_SOMP0,H_RU_SOMP)\n",
    "        NMSE3 = NMSE(H_hat_BSOMP0,H_RU_BSOMP)\n",
    "        NMSE4 = NMSE(H_hat_AMP0,H_RU_AMP)\n",
    "\n",
    "        NMSE5 = NMSE(H_hat_SOMP,H_RU_SOMP)\n",
    "        NMSE6 = NMSE(H_hat_BSOMP,H_RU_BSOMP)\n",
    "        NMSE7 = NMSE(H_hat_AMP,H_RU_AMP)\n",
    "\n",
    "        NMSE8 = NMSE(H_hat_LAMP,H_RU_LAMP)\n",
    "        NMSE9 = NMSE(H_hat_LAMP0,H_RU_LAMP)\n",
    "        print('完美信道估计+csiNet反馈 NMSE %.5f' % NMSE1.cpu()) \n",
    "        print('SOMP道估计+完美反馈 NMSE %.5f' % NMSE5.cpu())\n",
    "        print('BSOMP道估计+完美反馈 NMSE %.5f' % NMSE6.cpu())\n",
    "        print('AMP道估计+完美反馈 NMSE %.5f' % NMSE7.cpu())\n",
    "        print('GMMV-LAMP道估计+完美反馈 NMSE %.5f' % NMSE8.cpu())\n",
    "\n",
    "        print('SOMP道估计+csiNet反馈 NMSE %.5f' % NMSE2.cpu())\n",
    "        print('BSOMP道估计+csiNet反馈 NMSE %.5f' % NMSE3.cpu())\n",
    "        print('AMP道估计+csiNet反馈 NMSE %.5f' % NMSE4.cpu())\n",
    "        print('GMMV-LAMP道估计+csiNet反馈 NMSE %.5f' % NMSE9.cpu())\n",
    "        "
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### GMMV-LAMP网络定义"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def eta(R, sigma_t2, K, G, M, theta):  #R [batch,G,K] G为原始稀疏数据长度，K为子载波数也就是MMV的数量，M为观测维度\n",
    "    theta1 = torch.abs(theta[0])\n",
    "    theta2 = theta[1]\n",
    "    pi_t = 1 + sigma_t2/theta1\n",
    "    psi_t = 0.5*K*torch.log(torch.tensor(1+theta1/sigma_t2).cuda()) + theta2\n",
    "    \n",
    "#     aa = (psi_t.reshape(R.shape[0],1,1)-torch.sum(torch.abs(R*R),[2]).reshape(R.shape[0],R.shape[1],1)/((2*sigma_t2*pi_t).reshape(R.shape[0],1,1)))\n",
    "#     frac = 1/torch.sigmoid(-aa)*\n",
    "    \n",
    "    aa = (psi_t.reshape(R.shape[0],1,1)-torch.sum(torch.abs(R*R),[2]).reshape(R.shape[0],R.shape[1],1)/((2*sigma_t2*pi_t).reshape(R.shape[0],1,1)))\n",
    "    frac_div = torch.sigmoid(-aa)/pi_t.reshape(R.shape[0],1,1)\n",
    "\n",
    "\n",
    "    H_res = R*frac_div\n",
    "    b = torch.sum(frac_div,[1,2])/M\n",
    "    return H_res,b\n",
    "class LAMP_Net(nn.Module): #单层\n",
    "    def __init__(self, Theta , Nc, N, L): #H [batch,G,K]为角度域信道, Theta [Nc,L,N]为感知矩阵 N为原始稀疏数据长度，T为迭代次数，Nc为子载波数也就是MMV的数量，L为观测维度\n",
    "        super(LAMP_Net,self).__init__()\n",
    "        self.B = torch.nn.Parameter(Hermitian(Theta))    #[Nc,N,L]\n",
    "        self.theta = torch.nn.Parameter(torch.tensor([1.0,2.0]))\n",
    "        \n",
    "        \n",
    "    def forward(self, Y, Nc, N, L, Theta): #输入Y[batch,Nc,L] 要改成[batch,L,Nc]\n",
    "        \n",
    "        # N = ((torch.randn(H.shape[0],M,K).cuda() + 1j*torch.randn(H.shape[0],M,K).cuda())*(sqrt(sigma2/2))).cuda()\n",
    "        # Y = torch.matmul(self.A,H) + N\n",
    "\n",
    "        Y = Y.permute(0,2,1).reshape(Y.shape[0],L,Nc)\n",
    "\n",
    "        H_res = torch.zeros(Y.shape[0],N,Nc).cuda()\n",
    "        V = Y #[batch,L,Nc]\n",
    "        \n",
    "        B = self.B\n",
    "        theta = self.theta\n",
    "\n",
    "        BV = B @ V.permute(0,2,1).reshape(Y.shape[0],Nc,L,1) #[batch,Nc,N,1]\n",
    "        BV = BV.permute(0,2,1,3).reshape(Y.shape[0],N,Nc) #[batch,N,Nc]\n",
    "\n",
    "        R = H_res + BV\n",
    "        sigma_t2 = torch.sum(torch.abs(V*V),[1,2])/(L*Nc)\n",
    "        H_res,b = eta(R, sigma_t2, Nc, N, L, theta)\n",
    "\n",
    "        HH = Theta @ H_res.permute(0,2,1).reshape(Y.shape[0],Nc,N,1)   #[batch,Nc,L,1]\n",
    "        HH = HH.permute(0,2,1,3).reshape(Y.shape[0],L,Nc) #[batch,L,Nc]\n",
    "\n",
    "        V = Y-HH+b.reshape(-1,1,1)*V\n",
    "        \n",
    "        return H_res,V\n",
    "\n",
    "class LAMP_MuNet(nn.Module): #多层\n",
    "    def __init__(self, Theta , Nc, N, L, net): #H [batch,G,K]为角度域信道, Theta [Nc,L,N]为感知矩阵 N为原始稀疏数据长度，T为迭代次数，Nc为子载波数也就是MMV的数量，L为观测维度\n",
    "        super(LAMP_MuNet,self).__init__()\n",
    "        self.net = net\n",
    "        for p in self.net.parameters():\n",
    "            (p.requires_grad) = False\n",
    "        \n",
    "        self.B = torch.nn.Parameter(Hermitian(Theta))    #[Nc,N,L]\n",
    "        self.theta = torch.nn.Parameter(torch.tensor([1.0,2.0]))\n",
    "        \n",
    "        \n",
    "    def forward(self, Y, Nc, N, L, Theta): #输入Y[batch,Nc,L] 要改成[batch,L,Nc]\n",
    "\n",
    "        H_res,V = self.net(Y, Nc, N, L, Theta) \n",
    "\n",
    "        Y = Y.permute(0,2,1).reshape(Y.shape[0],L,Nc)\n",
    "\n",
    "        B = self.B\n",
    "        theta = self.theta\n",
    "\n",
    "        BV = B @ V.permute(0,2,1).reshape(Y.shape[0],Nc,L,1) #[batch,Nc,N,1]\n",
    "        BV = BV.permute(0,2,1,3).reshape(Y.shape[0],N,Nc) #[batch,N,Nc]\n",
    "\n",
    "        R = H_res + BV\n",
    "        sigma_t2 = torch.sum(torch.abs(V*V),[1,2])/(L*Nc)\n",
    "        H_res,b = eta(R, sigma_t2, Nc, N, L, theta)\n",
    "\n",
    "        HH = Theta @ H_res.permute(0,2,1).reshape(Y.shape[0],Nc,N,1)   #[batch,Nc,L,1]\n",
    "        HH = HH.permute(0,2,1,3).reshape(Y.shape[0],L,Nc) #[batch,L,Nc]\n",
    "\n",
    "        V = Y-HH+b.reshape(-1,1,1)*V\n",
    "        \n",
    "        return H_res,V"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def NMSE(H_RU_hat,H_RU):    \n",
    "    NMSE = torch.sum(torch.abs((H_RU_hat-H_RU))**2)/torch.sum(torch.abs(H_RU)**2)\n",
    "    return NMSE  \n",
    "def CE_LAMP_GMMV(param_list, L,sigma): #K1用户数 K2稀疏度\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "\n",
    "    factor = 100000*L/32\n",
    "    B = 4096\n",
    "    # param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "\n",
    "    alpha = 1e5\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "    Step = 3000\n",
    "\n",
    "    batch = 64\n",
    "\n",
    "    S_digital = torch.randn(L,Nc,4,1).cuda()+1j*torch.randn(L,Nc,4,1).cuda() #L个时隙BS导频,这里假设频率选择\n",
    "    S_analog = torch.rand(L,1,M_ant,4).cuda() * 2*pi\n",
    "    Phase = torch.rand(L,M_ant).cuda() * 2*pi   #L个时隙RIS相位\n",
    "    Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "    for l in range(L):\n",
    "        s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "        s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "        s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "    #             print(s)\n",
    "        \n",
    "    #             print(n)\n",
    "        \n",
    "        Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "        a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "        Phi[:,l,:] = a[:,0,:]\n",
    "    #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "    Theta = Phi @ A_t\n",
    "    Theta = Theta/sqrt(factor)\n",
    "\n",
    "    for T in range(5):\n",
    "        if T==0:\n",
    "            net = LAMP_Net(Theta, Nc, M_ant*pho*pho, L).cuda()\n",
    "        else:  \n",
    "            net_1 =  torch.load('./models/GMMV_LAMP'+str(T)+'_'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "            net = LAMP_MuNet(Theta, Nc, M_ant*pho*pho, L, net_1).cuda()\n",
    "            S_digital = torch.load('./models/S_digital'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "            S_analog  = torch.load('./models/S_analog'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "            Phase = torch.load('./models/Phase'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "            # net = LAMP_MuNet(Theta , Nc, M_ant*pho*pho, L, net).cuda()\n",
    "        optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, net.parameters()),lr=0.001)\n",
    "\n",
    "        num_train = 0\n",
    "        train_nmse = 0\n",
    "        best_nmse = 2\n",
    "        start = datetime.datetime.now()\n",
    "        \n",
    "\n",
    "        for step in range(Step):\n",
    "            net.train()\n",
    "            H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "\n",
    "            Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "            Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "            for l in range(L):\n",
    "                s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "                \n",
    "                s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "                s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "            #             print(s)\n",
    "                n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "                \n",
    "            #             print(n)\n",
    "                \n",
    "                Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "                a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "                Phi[:,l,:] = a[:,0,:]\n",
    "            #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "                y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "                Y[:,:,l] = y\n",
    "\n",
    "            Theta = Phi @ A_t\n",
    "            Theta = Theta/sqrt(factor)\n",
    "            Y = Y/sqrt(factor)\n",
    "\n",
    "            H_t_hat,V = net(Y, Nc, M_ant*pho*pho, L, Theta) #[batch,过采样,Nc]\n",
    "            H_t_hat = H_t_hat.permute(0,2,1).reshape(Y.shape[0],Nc,M_ant*pho*pho,1)\n",
    "            H_hat = A_t @ H_t_hat\n",
    "            H_hat = H_hat.reshape(-1,Nc,M_ant)\n",
    "            loss = NMSE(H_hat,H_RU)\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            num_train = num_train + 1\n",
    "            train_nmse = train_nmse + loss\n",
    "\n",
    "            if (step+1)%100==0:\n",
    "                net.eval() \n",
    "                with torch.no_grad():\n",
    "                    H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "\n",
    "                    Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "                    Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "                    for l in range(L):\n",
    "                        s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "                        \n",
    "                        s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "                        s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "                    #             print(s)\n",
    "                        n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "                        \n",
    "                    #             print(n)\n",
    "                        \n",
    "                        Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "                        a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "                        Phi[:,l,:] = a[:,0,:]\n",
    "                    #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "                        y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "                        Y[:,:,l] = y\n",
    "\n",
    "                    Theta = Phi @ A_t\n",
    "                    Theta = Theta/sqrt(factor)\n",
    "                    Y = Y/sqrt(factor)\n",
    "\n",
    "                    H_t_hat,V = net(Y, Nc, M_ant*pho*pho, L, Theta) #[batch,过采样,Nc]\n",
    "                    H_t_hat = H_t_hat.permute(0,2,1).reshape(Y.shape[0],Nc,M_ant*pho*pho,1)\n",
    "                    H_hat = A_t @ H_t_hat\n",
    "                    H_hat = H_hat.reshape(-1,Nc,M_ant)\n",
    "                    loss = NMSE(H_hat, H_RU)\n",
    "\n",
    "                    train_nmse = train_nmse/num_train\n",
    "                    time0 =  datetime.datetime.now()-start\n",
    "                    print('step:',step,'time',time0,'train NMSE %.5f' % train_nmse.cpu(),'test NMSE %.5f' % loss.cpu()) \n",
    "\n",
    "                    if loss < best_nmse:\n",
    "                        best_nmse = loss\n",
    "                        torch.save(net, './models/GMMV_LAMP'+str(T+1)+'_'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "                        torch.save(S_digital, './models/S_digital'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "                        torch.save(S_analog, './models/S_analog'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "                        torch.save(Phase, './models/Phase'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "                        print('Model saved!')\n",
    "                    num_train = 0\n",
    "                    train_nmse = 0\n",
    "                    start = datetime.datetime.now()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def Test_LAMP_GMMV(param_list, T, batch, Step, L,sigma): #K1用户数 K2稀疏度\n",
    "#Phi 为观测矩阵 Theta为感知矩阵\n",
    "\n",
    "    factor = 100000*L/32\n",
    "    B = 4096\n",
    "    # param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "\n",
    "    alpha = 1e5\n",
    "    # H_RU = H_RU.reshape(batch,Nc,M_ant,1)\n",
    "\n",
    "    batch = 64\n",
    "\n",
    "    S_digital = torch.load('./models/S_digital'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "    S_analog  = torch.load('./models/S_analog'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "    Phase = torch.load('./models/Phase'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "    Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "    for l in range(L):\n",
    "        s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "        s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "        s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "    #             print(s)\n",
    "        \n",
    "    #             print(n)\n",
    "        \n",
    "        Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "        a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "        Phi[:,l,:] = a[:,0,:]\n",
    "    #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "    Theta = Phi @ A_t\n",
    "    Theta = Theta/sqrt(factor)\n",
    "\n",
    "    for TT in range(T):\n",
    "        if TT==0:\n",
    "            net = LAMP_Net(Theta, Nc, M_ant*pho*pho, L).cuda()\n",
    "        else:  \n",
    "            net = LAMP_MuNet(Theta, Nc, M_ant*pho*pho, L, net).cuda()\n",
    "    net =  torch.load('./models/GMMV_LAMP'+str(T)+'_'+str(L)+'pilots_'+str(B)+'bits_'+str(Lp)+'paths_'+str(floor(log10(sigma)))+'sigma'+'.pth')\n",
    "\n",
    "            # net = LAMP_MuNet(Theta , Nc, M_ant*pho*pho, L, net).cuda()\n",
    "\n",
    "    \n",
    "\n",
    "    for step in range(Step):\n",
    "\n",
    "        net.eval() \n",
    "        with torch.no_grad():\n",
    "            H_RU = Channel_RIS_UE(param_list,batch,alpha)\n",
    "\n",
    "\n",
    "            Y = torch.zeros(H_RU.shape[0],Nc,L).cuda() + 0j#总观测【batch,Nc,L】\n",
    "\n",
    "            Phi = torch.zeros(Nc,L,M_ant).cuda() + 0j\n",
    "\n",
    "            for l in range(L):\n",
    "                s = (torch.exp(1j*S_analog[l,:,:,:]) @ S_digital[l,:,:,:] ).reshape(Nc,M_ant,1)  #导频s [Nc,M_ant]    \n",
    "                \n",
    "                s_sigma = torch.sqrt(torch.sum(torch.abs(s*s)))\n",
    "                s = s/s_sigma.reshape(1,1,1)*torch.min(s_sigma,torch.tensor(sqrt(Nc))).reshape(1,1,1)\n",
    "            #             print(s)\n",
    "                n = (torch.randn(H_RU.shape[0],Nc) + 1j*torch.randn(H_RU.shape[0],Nc)).cuda()*sqrt(sigma/2)\n",
    "                \n",
    "            #             print(n)\n",
    "                \n",
    "                Theta = torch.diag(torch.exp(1j*Phase[l,:])) #RIS反射矩阵【M_ant,M_ant】 \n",
    "                a =  (Theta @ H_BR @ s).permute(0,2,1) #a[Nc,1,M_ant] = Theta * G * s_l\n",
    "                Phi[:,l,:] = a[:,0,:]\n",
    "            #             print((a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc))\n",
    "                y = (a @ H_RU.reshape(-1,Nc,M_ant,1)).reshape(-1,Nc) + n #[batch, Nc]\n",
    "                Y[:,:,l] = y\n",
    "\n",
    "            Theta = Phi @ A_t\n",
    "            Theta = Theta/sqrt(factor)\n",
    "            Y = Y/sqrt(factor)\n",
    "\n",
    "            H_t_hat,V = net(Y, Nc, M_ant*pho*pho, L, Theta) #[batch,过采样,Nc]\n",
    "            H_t_hat = H_t_hat.permute(0,2,1).reshape(Y.shape[0],Nc,M_ant*pho*pho,1)\n",
    "            H_hat = A_t @ H_t_hat\n",
    "            H_hat = H_hat.reshape(-1,Nc,M_ant)\n",
    "\n",
    "            if step == 0:\n",
    "                H_RU_set = H_RU\n",
    "                H_hat_set = H_hat\n",
    "            else:\n",
    "                H_RU_set = torch.cat([H_RU_set,H_RU],0)\n",
    "                H_hat_set = torch.cat([H_hat_set,H_hat],0)\n",
    "    return H_RU_set,H_hat_set\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test_CF_MDDL(param_list,batch,model_name_AE):\n",
    "    \n",
    "    fc = param_list[0]\n",
    "#     B  = param_list1[1]\n",
    "#     Nc = param_list1[2]\n",
    "    M  = param_list[3]\n",
    "    N  = param_list[4]\n",
    "    D_sub = param_list[5]\n",
    "    D_ant = param_list[6]\n",
    "    R = param_list[7]\n",
    "    M_ant = param_list[8]\n",
    "    h1 = param_list[9]\n",
    "    h2 = param_list[10]\n",
    "    L =  param_list[11]\n",
    "    sigma = param_list[12]\n",
    "    B = param_list[13]\n",
    "    Lp = param_list[15]\n",
    "\n",
    "    net = csiNet(param_list).cuda()\n",
    "    net = torch.load(model_name_AE)\n",
    "        \n",
    "    \n",
    "    Ay = (M-1)*D_sub + (N-1)*(D_ant);#阵列长度\n",
    "    Az = Ay;\n",
    "\n",
    "    r_BS  = (torch.tensor([0,0,h1]) - torch.tensor([0,Ay/2,Az/2])).cuda()\n",
    "    r_RIS = (torch.tensor([R,0,h1]) - torch.tensor([0,Ay/2,Az/2])).cuda()\n",
    "\n",
    "    \n",
    "    ##随机生成信道\n",
    "    # H_BR = Channel_BS_RIS(param_list) * 1e6\n",
    "    \n",
    "    \n",
    "    \n",
    "    num_train = 0\n",
    "    train_nmse = 0\n",
    "    best_nmse = 1\n",
    "    start = datetime.datetime.now()\n",
    "    numStep = 1\n",
    "    warmup_steps = 4000\n",
    "    d_model = 256\n",
    "\n",
    "    with torch.no_grad():\n",
    "        net.eval()\n",
    "        alpha = 1e5\n",
    "\n",
    "\n",
    "\n",
    "        H_RU_LAMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "        H_hat_LAMP = (torch.zeros(batch,Nc,M_ant) + 0j).cuda()\n",
    "\n",
    "\n",
    "\n",
    "        for ii in range(batch):\n",
    "\n",
    "\n",
    "            \n",
    "            T = 5\n",
    "            H_RU_LAMP11,H_hat_LAMP11 = Test_LAMP_GMMV(param_list, T, 1, 1, L,sigma)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "            if NMSE(H_hat_LAMP11,H_RU_LAMP11)>1:\n",
    "                H_hat_LAMP11 = H_hat_LAMP11-H_hat_LAMP11\n",
    "\n",
    "            \n",
    "\n",
    "            H_RU_LAMP[ii,:,:] = H_RU_LAMP11[0,:,:]\n",
    "            H_hat_LAMP[ii,:,:] = H_hat_LAMP11[0,:,:]\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "        H_RU_LAMP0 = net(param_list, H_BR, H_RU_LAMP)\n",
    "        H_hat_LAMP0 = net(param_list, H_BR, H_hat_LAMP)\n",
    "\n",
    "\n",
    "        \n",
    "        NMSE1 = NMSE(H_RU_LAMP0,H_RU_LAMP)\n",
    "\n",
    "\n",
    "        NMSE8 = NMSE(H_hat_LAMP,H_RU_LAMP)\n",
    "        NMSE9 = NMSE(H_hat_LAMP0,H_RU_LAMP)\n",
    "        print('完美信道估计+csiNet反馈 NMSE %.5f' % NMSE1.cpu()) \n",
    "        print('GMMV-LAMP道估计+完美反馈 NMSE %.5f' % NMSE8.cpu())\n",
    "        print('GMMV-LAMP道估计+csiNet反馈 NMSE %.5f' % NMSE9.cpu())\n",
    "        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 训练CSINET"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 49 time 0:00:13.818283 train NMSE 1.05149 test NMSE 1.00230\n",
      "step: 99 time 0:00:13.591619 train NMSE 1.00139 test NMSE 1.00089\n",
      "step: 149 time 0:00:13.683874 train NMSE 1.00060 test NMSE 1.00020\n",
      "step: 199 time 0:00:13.792758 train NMSE 1.00003 test NMSE 0.99912\n",
      "Model saved!\n",
      "step: 249 time 0:00:13.936893 train NMSE 0.99300 test NMSE 0.98218\n",
      "Model saved!\n",
      "step: 299 time 0:00:14.084904 train NMSE 0.96622 test NMSE 0.97257\n",
      "Model saved!\n",
      "step: 349 time 0:00:14.185765 train NMSE 0.93593 test NMSE 0.93049\n",
      "Model saved!\n",
      "step: 399 time 0:00:14.242995 train NMSE 0.90795 test NMSE 0.88426\n",
      "Model saved!\n",
      "step: 449 time 0:00:14.221688 train NMSE 0.88884 test NMSE 0.88310\n",
      "Model saved!\n",
      "step: 499 time 0:00:14.219037 train NMSE 0.88253 test NMSE 0.86145\n",
      "Model saved!\n",
      "step: 549 time 0:00:14.294446 train NMSE 0.86688 test NMSE 0.84659\n",
      "Model saved!\n",
      "step: 599 time 0:00:14.317091 train NMSE 0.85586 test NMSE 0.86890\n",
      "step: 649 time 0:00:14.311648 train NMSE 0.84578 test NMSE 0.82185\n",
      "Model saved!\n",
      "step: 699 time 0:00:14.334680 train NMSE 0.83403 test NMSE 0.80800\n",
      "Model saved!\n",
      "step: 749 time 0:00:14.344417 train NMSE 0.83507 test NMSE 0.84398\n",
      "step: 799 time 0:00:14.340082 train NMSE 0.83506 test NMSE 0.77439\n",
      "Model saved!\n",
      "step: 849 time 0:00:14.337357 train NMSE 0.82688 test NMSE 0.82985\n",
      "step: 899 time 0:00:14.298534 train NMSE 0.81561 test NMSE 0.78978\n",
      "step: 949 time 0:00:14.297137 train NMSE 0.81280 test NMSE 0.80858\n",
      "step: 999 time 0:00:14.296187 train NMSE 0.81149 test NMSE 0.79819\n",
      "step: 1049 time 0:00:14.333878 train NMSE 0.81046 test NMSE 0.79471\n",
      "step: 1099 time 0:00:14.366644 train NMSE 0.80651 test NMSE 0.83341\n",
      "step: 1149 time 0:00:14.344172 train NMSE 0.80750 test NMSE 0.79735\n",
      "step: 1199 time 0:00:14.353481 train NMSE 0.80715 test NMSE 0.80906\n",
      "step: 1249 time 0:00:14.349700 train NMSE 0.80748 test NMSE 0.81048\n",
      "step: 1299 time 0:00:14.307284 train NMSE 0.80104 test NMSE 0.80101\n",
      "step: 1349 time 0:00:14.315856 train NMSE 0.80126 test NMSE 0.77233\n",
      "Model saved!\n",
      "step: 1399 time 0:00:14.313084 train NMSE 0.79780 test NMSE 0.80055\n",
      "step: 1449 time 0:00:14.316775 train NMSE 0.79336 test NMSE 0.79380\n",
      "step: 1499 time 0:00:14.327230 train NMSE 0.79529 test NMSE 0.76355\n",
      "Model saved!\n",
      "step: 1549 time 0:00:14.327277 train NMSE 0.79411 test NMSE 0.77979\n",
      "step: 1599 time 0:00:14.328762 train NMSE 0.79221 test NMSE 0.75387\n",
      "Model saved!\n",
      "step: 1649 time 0:00:14.333777 train NMSE 0.78838 test NMSE 0.77747\n",
      "step: 1699 time 0:00:14.322778 train NMSE 0.78632 test NMSE 0.76992\n",
      "step: 1749 time 0:00:14.327641 train NMSE 0.78344 test NMSE 0.79393\n",
      "step: 1799 time 0:00:14.333516 train NMSE 0.78758 test NMSE 0.81277\n",
      "step: 1849 time 0:00:14.318757 train NMSE 0.78831 test NMSE 0.79276\n",
      "step: 1899 time 0:00:14.337743 train NMSE 0.78035 test NMSE 0.76827\n",
      "step: 1949 time 0:00:14.331737 train NMSE 0.78193 test NMSE 0.77613\n",
      "step: 1999 time 0:00:14.328763 train NMSE 0.77615 test NMSE 0.74199\n",
      "Model saved!\n",
      "step: 2049 time 0:00:14.328793 train NMSE 0.77832 test NMSE 0.76203\n",
      "step: 2099 time 0:00:14.337259 train NMSE 0.77941 test NMSE 0.76707\n",
      "step: 2149 time 0:00:14.328259 train NMSE 0.77557 test NMSE 0.78364\n",
      "step: 2199 time 0:00:14.348310 train NMSE 0.77698 test NMSE 0.70697\n",
      "Model saved!\n",
      "step: 2249 time 0:00:14.334728 train NMSE 0.77832 test NMSE 0.75746\n",
      "step: 2299 time 0:00:14.331284 train NMSE 0.76989 test NMSE 0.75385\n",
      "step: 2349 time 0:00:14.343728 train NMSE 0.76301 test NMSE 0.75446\n",
      "step: 2399 time 0:00:14.337793 train NMSE 0.77574 test NMSE 0.73152\n",
      "step: 2449 time 0:00:14.348691 train NMSE 0.77360 test NMSE 0.72493\n",
      "step: 2499 time 0:00:14.346726 train NMSE 0.76893 test NMSE 0.76678\n",
      "step: 2549 time 0:00:14.349708 train NMSE 0.76664 test NMSE 0.75234\n",
      "step: 2599 time 0:00:14.340316 train NMSE 0.76852 test NMSE 0.74972\n",
      "step: 2649 time 0:00:14.352756 train NMSE 0.76712 test NMSE 0.73952\n",
      "step: 2699 time 0:00:14.332779 train NMSE 0.76457 test NMSE 0.71213\n",
      "step: 2749 time 0:00:14.344700 train NMSE 0.76796 test NMSE 0.75469\n",
      "step: 2799 time 0:00:14.330754 train NMSE 0.77009 test NMSE 0.74188\n",
      "step: 2849 time 0:00:14.340731 train NMSE 0.76038 test NMSE 0.75030\n",
      "step: 2899 time 0:00:14.336959 train NMSE 0.76480 test NMSE 0.72350\n",
      "step: 2949 time 0:00:14.342715 train NMSE 0.76415 test NMSE 0.73063\n",
      "step: 2999 time 0:00:14.335751 train NMSE 0.76406 test NMSE 0.75086\n",
      "step: 3049 time 0:00:14.340728 train NMSE 0.75844 test NMSE 0.76957\n",
      "step: 3099 time 0:00:14.330749 train NMSE 0.76975 test NMSE 0.74418\n",
      "step: 3149 time 0:00:14.343695 train NMSE 0.75907 test NMSE 0.77633\n",
      "step: 3199 time 0:00:14.331726 train NMSE 0.76209 test NMSE 0.74636\n",
      "step: 3249 time 0:00:14.350697 train NMSE 0.75907 test NMSE 0.73872\n",
      "step: 3299 time 0:00:14.344705 train NMSE 0.76166 test NMSE 0.71086\n",
      "step: 3349 time 0:00:14.342801 train NMSE 0.75954 test NMSE 0.74722\n",
      "step: 3399 time 0:00:14.338251 train NMSE 0.76109 test NMSE 0.73326\n",
      "step: 3449 time 0:00:14.338750 train NMSE 0.75950 test NMSE 0.75918\n",
      "step: 3499 time 0:00:14.350732 train NMSE 0.75732 test NMSE 0.72697\n",
      "step: 3549 time 0:00:14.344704 train NMSE 0.76433 test NMSE 0.73564\n",
      "step: 3599 time 0:00:14.346688 train NMSE 0.75643 test NMSE 0.75291\n",
      "step: 3649 time 0:00:14.347726 train NMSE 0.76036 test NMSE 0.75714\n",
      "step: 3699 time 0:00:14.354699 train NMSE 0.75840 test NMSE 0.72060\n",
      "step: 3749 time 0:00:14.347682 train NMSE 0.74812 test NMSE 0.72992\n",
      "step: 3799 time 0:00:14.338227 train NMSE 0.75777 test NMSE 0.72575\n",
      "step: 3849 time 0:00:14.369679 train NMSE 0.75454 test NMSE 0.72406\n",
      "step: 3899 time 0:00:14.357684 train NMSE 0.75539 test NMSE 0.68174\n",
      "Model saved!\n",
      "step: 3949 time 0:00:14.341718 train NMSE 0.75472 test NMSE 0.70713\n",
      "step: 3999 time 0:00:14.357659 train NMSE 0.74729 test NMSE 0.71328\n",
      "step: 4049 time 0:00:14.354665 train NMSE 0.75148 test NMSE 0.71327\n",
      "step: 4099 time 0:00:14.348723 train NMSE 0.75780 test NMSE 0.69919\n",
      "step: 4149 time 0:00:14.359232 train NMSE 0.75431 test NMSE 0.73495\n",
      "step: 4199 time 0:00:14.356183 train NMSE 0.75175 test NMSE 0.73979\n",
      "step: 4249 time 0:00:14.355175 train NMSE 0.75040 test NMSE 0.77064\n",
      "step: 4299 time 0:00:14.361667 train NMSE 0.75329 test NMSE 0.72017\n",
      "step: 4349 time 0:00:14.346226 train NMSE 0.74890 test NMSE 0.71565\n",
      "step: 4399 time 0:00:14.353173 train NMSE 0.76042 test NMSE 0.69953\n",
      "step: 4449 time 0:00:14.354193 train NMSE 0.75300 test NMSE 0.69700\n",
      "step: 4499 time 0:00:14.352704 train NMSE 0.74892 test NMSE 0.74529\n",
      "step: 4549 time 0:00:14.345430 train NMSE 0.74913 test NMSE 0.73644\n",
      "step: 4599 time 0:00:14.356169 train NMSE 0.75072 test NMSE 0.74674\n",
      "step: 4649 time 0:00:14.343680 train NMSE 0.74864 test NMSE 0.74663\n",
      "step: 4699 time 0:00:14.360147 train NMSE 0.74345 test NMSE 0.75515\n",
      "step: 4749 time 0:00:14.347211 train NMSE 0.75411 test NMSE 0.71592\n",
      "step: 4799 time 0:00:14.344703 train NMSE 0.75380 test NMSE 0.73022\n",
      "step: 4849 time 0:00:14.351689 train NMSE 0.74893 test NMSE 0.73383\n",
      "step: 4899 time 0:00:14.357677 train NMSE 0.74958 test NMSE 0.69806\n",
      "step: 4949 time 0:00:14.344215 train NMSE 0.74249 test NMSE 0.70436\n",
      "step: 4999 time 0:00:14.366677 train NMSE 0.74560 test NMSE 0.66732\n",
      "Model saved!\n",
      "step: 5049 time 0:00:14.356211 train NMSE 0.75551 test NMSE 0.71777\n",
      "step: 5099 time 0:00:14.358714 train NMSE 0.74932 test NMSE 0.75825\n",
      "step: 5149 time 0:00:14.353804 train NMSE 0.74071 test NMSE 0.70159\n",
      "step: 5199 time 0:00:14.354710 train NMSE 0.74718 test NMSE 0.71478\n",
      "step: 5249 time 0:00:14.348713 train NMSE 0.75143 test NMSE 0.71167\n",
      "step: 5299 time 0:00:14.355350 train NMSE 0.73945 test NMSE 0.73484\n",
      "step: 5349 time 0:00:14.344209 train NMSE 0.74829 test NMSE 0.70506\n",
      "step: 5399 time 0:00:14.352670 train NMSE 0.75033 test NMSE 0.69683\n",
      "step: 5449 time 0:00:14.346690 train NMSE 0.73952 test NMSE 0.70472\n",
      "step: 5499 time 0:00:14.346177 train NMSE 0.74564 test NMSE 0.70681\n",
      "step: 5549 time 0:00:14.342342 train NMSE 0.74050 test NMSE 0.71699\n",
      "step: 5599 time 0:00:14.340722 train NMSE 0.74554 test NMSE 0.72570\n",
      "step: 5649 time 0:00:14.358631 train NMSE 0.74645 test NMSE 0.72673\n",
      "step: 5699 time 0:00:14.341216 train NMSE 0.74451 test NMSE 0.72795\n",
      "step: 5749 time 0:00:14.354459 train NMSE 0.74714 test NMSE 0.73419\n",
      "step: 5799 time 0:00:14.344034 train NMSE 0.73671 test NMSE 0.70534\n",
      "step: 5849 time 0:00:14.338182 train NMSE 0.74120 test NMSE 0.68698\n",
      "step: 5899 time 0:00:14.346277 train NMSE 0.74444 test NMSE 0.73230\n",
      "step: 5949 time 0:00:14.344063 train NMSE 0.73638 test NMSE 0.76498\n",
      "step: 5999 time 0:00:14.359081 train NMSE 0.74232 test NMSE 0.74479\n",
      "step: 6049 time 0:00:14.348024 train NMSE 0.74290 test NMSE 0.71157\n",
      "step: 6099 time 0:00:14.352061 train NMSE 0.73934 test NMSE 0.73741\n",
      "step: 6149 time 0:00:14.345250 train NMSE 0.73898 test NMSE 0.72913\n",
      "step: 6199 time 0:00:14.350927 train NMSE 0.73398 test NMSE 0.75287\n",
      "step: 6249 time 0:00:14.348660 train NMSE 0.73918 test NMSE 0.70289\n",
      "step: 6299 time 0:00:14.349340 train NMSE 0.74489 test NMSE 0.71495\n",
      "step: 6349 time 0:00:14.347904 train NMSE 0.74037 test NMSE 0.71176\n",
      "step: 6399 time 0:00:14.383285 train NMSE 0.74677 test NMSE 0.71423\n",
      "step: 6449 time 0:00:14.357947 train NMSE 0.74411 test NMSE 0.71850\n",
      "step: 6499 time 0:00:14.397158 train NMSE 0.74483 test NMSE 0.71480\n",
      "step: 6549 time 0:00:14.399228 train NMSE 0.73671 test NMSE 0.70917\n",
      "step: 6599 time 0:00:14.370947 train NMSE 0.73903 test NMSE 0.72400\n",
      "step: 6649 time 0:00:14.365724 train NMSE 0.73554 test NMSE 0.72694\n",
      "step: 6699 time 0:00:14.392904 train NMSE 0.73730 test NMSE 0.70490\n",
      "step: 6749 time 0:00:14.418755 train NMSE 0.73788 test NMSE 0.66695\n",
      "Model saved!\n",
      "step: 6799 time 0:00:14.406220 train NMSE 0.74730 test NMSE 0.70633\n",
      "step: 6849 time 0:00:18.309757 train NMSE 0.74246 test NMSE 0.72253\n",
      "step: 6899 time 0:00:32.571038 train NMSE 0.74737 test NMSE 0.69684\n",
      "step: 6949 time 0:00:34.205968 train NMSE 0.73512 test NMSE 0.72435\n",
      "step: 6999 time 0:00:32.811029 train NMSE 0.73827 test NMSE 0.74644\n",
      "step: 7049 time 0:00:32.909813 train NMSE 0.73558 test NMSE 0.70852\n",
      "step: 7099 time 0:00:32.956762 train NMSE 0.73494 test NMSE 0.68597\n",
      "step: 7149 time 0:00:34.265581 train NMSE 0.74131 test NMSE 0.72282\n",
      "step: 7199 time 0:00:32.840801 train NMSE 0.74152 test NMSE 0.71879\n",
      "step: 7249 time 0:00:32.793434 train NMSE 0.74027 test NMSE 0.74392\n",
      "step: 7299 time 0:00:32.756651 train NMSE 0.73800 test NMSE 0.72574\n",
      "step: 7349 time 0:00:32.866760 train NMSE 0.74048 test NMSE 0.69285\n",
      "step: 7399 time 0:00:32.779510 train NMSE 0.73942 test NMSE 0.72501\n",
      "step: 7449 time 0:00:34.107423 train NMSE 0.74193 test NMSE 0.69591\n",
      "step: 7499 time 0:00:32.808890 train NMSE 0.74029 test NMSE 0.68717\n",
      "step: 7549 time 0:00:32.823132 train NMSE 0.74381 test NMSE 0.71504\n",
      "step: 7599 time 0:00:32.788444 train NMSE 0.73208 test NMSE 0.71383\n",
      "step: 7649 time 0:00:32.847467 train NMSE 0.74393 test NMSE 0.73987\n",
      "step: 7699 time 0:00:32.852583 train NMSE 0.74117 test NMSE 0.66647\n",
      "Model saved!\n",
      "step: 7749 time 0:00:32.851192 train NMSE 0.73953 test NMSE 0.72094\n",
      "step: 7799 time 0:00:34.214144 train NMSE 0.73970 test NMSE 0.72865\n",
      "step: 7849 time 0:00:32.815678 train NMSE 0.73887 test NMSE 0.74040\n",
      "step: 7899 time 0:00:32.820092 train NMSE 0.74277 test NMSE 0.70112\n",
      "step: 7949 time 0:00:32.855303 train NMSE 0.73213 test NMSE 0.70010\n",
      "step: 7999 time 0:00:32.801426 train NMSE 0.73218 test NMSE 0.67829\n",
      "step: 8049 time 0:00:32.774657 train NMSE 0.73831 test NMSE 0.74688\n",
      "step: 8099 time 0:00:32.817337 train NMSE 0.73463 test NMSE 0.71299\n",
      "step: 8149 time 0:00:32.858873 train NMSE 0.73520 test NMSE 0.69116\n",
      "step: 8199 time 0:00:32.881675 train NMSE 0.73140 test NMSE 0.70521\n",
      "step: 8249 time 0:00:32.845277 train NMSE 0.73921 test NMSE 0.71062\n",
      "step: 8299 time 0:00:32.814347 train NMSE 0.73627 test NMSE 0.71695\n",
      "step: 8349 time 0:00:32.833791 train NMSE 0.73743 test NMSE 0.69277\n",
      "step: 8399 time 0:00:32.864526 train NMSE 0.72553 test NMSE 0.72854\n",
      "step: 8449 time 0:00:32.884308 train NMSE 0.73691 test NMSE 0.70930\n",
      "step: 8499 time 0:00:32.819582 train NMSE 0.73333 test NMSE 0.74634\n",
      "step: 8549 time 0:00:34.156493 train NMSE 0.73532 test NMSE 0.73445\n",
      "step: 8599 time 0:00:32.819530 train NMSE 0.72590 test NMSE 0.70444\n",
      "step: 8649 time 0:00:32.866261 train NMSE 0.73179 test NMSE 0.67466\n",
      "step: 8699 time 0:00:32.776490 train NMSE 0.73660 test NMSE 0.71135\n",
      "step: 8749 time 0:00:32.793446 train NMSE 0.73746 test NMSE 0.72523\n",
      "step: 8799 time 0:00:32.789995 train NMSE 0.73496 test NMSE 0.70490\n",
      "step: 8849 time 0:00:32.844369 train NMSE 0.73438 test NMSE 0.72877\n",
      "step: 8899 time 0:00:32.817042 train NMSE 0.73295 test NMSE 0.72902\n",
      "step: 8949 time 0:00:32.807548 train NMSE 0.73599 test NMSE 0.74785\n",
      "step: 8999 time 0:00:32.806378 train NMSE 0.73968 test NMSE 0.71508\n",
      "step: 9049 time 0:00:32.817067 train NMSE 0.73523 test NMSE 0.72817\n",
      "step: 9099 time 0:00:32.877236 train NMSE 0.72966 test NMSE 0.73652\n",
      "step: 9149 time 0:00:32.812853 train NMSE 0.74508 test NMSE 0.70719\n",
      "step: 9199 time 0:00:32.894201 train NMSE 0.73560 test NMSE 0.69473\n",
      "step: 9249 time 0:00:32.828804 train NMSE 0.74045 test NMSE 0.73996\n",
      "step: 9299 time 0:00:34.198371 train NMSE 0.73037 test NMSE 0.71297\n",
      "step: 9349 time 0:00:32.862291 train NMSE 0.73364 test NMSE 0.69947\n",
      "step: 9399 time 0:00:32.764530 train NMSE 0.73483 test NMSE 0.71460\n",
      "step: 9449 time 0:00:32.864153 train NMSE 0.73861 test NMSE 0.71757\n",
      "step: 9499 time 0:00:32.848982 train NMSE 0.73516 test NMSE 0.68813\n",
      "step: 9549 time 0:00:32.772562 train NMSE 0.73304 test NMSE 0.71556\n",
      "step: 9599 time 0:00:32.846818 train NMSE 0.73940 test NMSE 0.72434\n",
      "step: 9649 time 0:00:34.168830 train NMSE 0.73121 test NMSE 0.70018\n",
      "step: 9699 time 0:00:32.848158 train NMSE 0.73295 test NMSE 0.71874\n",
      "step: 9749 time 0:00:32.808425 train NMSE 0.74148 test NMSE 0.73316\n",
      "step: 9799 time 0:00:32.845933 train NMSE 0.73150 test NMSE 0.69820\n",
      "step: 9849 time 0:00:32.848338 train NMSE 0.73321 test NMSE 0.74566\n",
      "step: 9899 time 0:00:32.816367 train NMSE 0.73332 test NMSE 0.75222\n",
      "step: 9949 time 0:00:32.847669 train NMSE 0.73732 test NMSE 0.73250\n",
      "step: 9999 time 0:00:32.789444 train NMSE 0.73641 test NMSE 0.71650\n"
     ]
    }
   ],
   "source": [
    "B = 8\n",
    "\n",
    "batch = 64;   #用户数\n",
    "num_batch = 1\n",
    "STEPS = 10000\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "train_csiNet(param_list,num_batch,batch,STEPS,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 49 time 0:00:32.817091 train NMSE 1.03937 test NMSE 1.00066\n",
      "step: 99 time 0:00:32.864644 train NMSE 0.99924 test NMSE 0.99973\n",
      "Model saved!\n",
      "step: 149 time 0:00:32.863741 train NMSE 0.99372 test NMSE 0.99121\n",
      "Model saved!\n",
      "step: 199 time 0:00:32.790917 train NMSE 0.96281 test NMSE 0.93677\n",
      "Model saved!\n",
      "step: 249 time 0:00:32.835745 train NMSE 0.89593 test NMSE 0.86316\n",
      "Model saved!\n",
      "step: 299 time 0:00:32.828158 train NMSE 0.83325 test NMSE 0.81329\n",
      "Model saved!\n",
      "step: 349 time 0:00:34.166342 train NMSE 0.80315 test NMSE 0.79293\n",
      "Model saved!\n",
      "step: 399 time 0:00:32.831539 train NMSE 0.78369 test NMSE 0.76983\n",
      "Model saved!\n",
      "step: 449 time 0:00:32.828366 train NMSE 0.77479 test NMSE 0.74959\n",
      "Model saved!\n",
      "step: 499 time 0:00:32.842850 train NMSE 0.75628 test NMSE 0.74660\n",
      "Model saved!\n",
      "step: 549 time 0:00:32.799507 train NMSE 0.74996 test NMSE 0.73093\n",
      "Model saved!\n",
      "step: 599 time 0:00:32.822356 train NMSE 0.73641 test NMSE 0.72133\n",
      "Model saved!\n",
      "step: 649 time 0:00:32.832841 train NMSE 0.72793 test NMSE 0.71116\n",
      "Model saved!\n",
      "step: 699 time 0:00:34.146328 train NMSE 0.71583 test NMSE 0.71965\n",
      "step: 749 time 0:00:32.845843 train NMSE 0.70171 test NMSE 0.69837\n",
      "Model saved!\n",
      "step: 799 time 0:00:32.827694 train NMSE 0.68836 test NMSE 0.66836\n",
      "Model saved!\n",
      "step: 849 time 0:00:32.788472 train NMSE 0.67415 test NMSE 0.63070\n",
      "Model saved!\n",
      "step: 899 time 0:00:32.846491 train NMSE 0.65587 test NMSE 0.62531\n",
      "Model saved!\n",
      "step: 949 time 0:00:32.821407 train NMSE 0.64206 test NMSE 0.60773\n",
      "Model saved!\n",
      "step: 999 time 0:00:32.827504 train NMSE 0.63068 test NMSE 0.59689\n",
      "Model saved!\n",
      "step: 1049 time 0:00:34.129412 train NMSE 0.62543 test NMSE 0.60393\n",
      "step: 1099 time 0:00:32.836319 train NMSE 0.61733 test NMSE 0.58975\n",
      "Model saved!\n",
      "step: 1149 time 0:00:32.772485 train NMSE 0.60957 test NMSE 0.60175\n",
      "step: 1199 time 0:00:32.839315 train NMSE 0.60477 test NMSE 0.59873\n",
      "step: 1249 time 0:00:32.880418 train NMSE 0.60035 test NMSE 0.57947\n",
      "Model saved!\n",
      "step: 1299 time 0:00:32.808838 train NMSE 0.60192 test NMSE 0.59336\n",
      "step: 1349 time 0:00:32.786197 train NMSE 0.60219 test NMSE 0.58274\n",
      "step: 1399 time 0:00:34.141862 train NMSE 0.59355 test NMSE 0.57394\n",
      "Model saved!\n",
      "step: 1449 time 0:00:32.855824 train NMSE 0.59820 test NMSE 0.57980\n",
      "step: 1499 time 0:00:32.795434 train NMSE 0.59379 test NMSE 0.56907\n",
      "Model saved!\n",
      "step: 1549 time 0:00:32.817110 train NMSE 0.59052 test NMSE 0.57727\n",
      "step: 1599 time 0:00:32.820313 train NMSE 0.58783 test NMSE 0.57882\n",
      "step: 1649 time 0:00:32.814821 train NMSE 0.58714 test NMSE 0.57326\n",
      "step: 1699 time 0:00:32.801424 train NMSE 0.58509 test NMSE 0.57239\n",
      "step: 1749 time 0:00:34.214623 train NMSE 0.58779 test NMSE 0.56873\n",
      "Model saved!\n",
      "step: 1799 time 0:00:32.841798 train NMSE 0.58463 test NMSE 0.57779\n",
      "step: 1849 time 0:00:32.838597 train NMSE 0.58262 test NMSE 0.55991\n",
      "Model saved!\n",
      "step: 1899 time 0:00:32.770336 train NMSE 0.58583 test NMSE 0.57641\n",
      "step: 1949 time 0:00:32.796386 train NMSE 0.58235 test NMSE 0.56765\n",
      "step: 1999 time 0:00:32.872371 train NMSE 0.58390 test NMSE 0.57226\n",
      "step: 2049 time 0:00:32.846795 train NMSE 0.58127 test NMSE 0.57523\n",
      "step: 2099 time 0:00:32.830554 train NMSE 0.58169 test NMSE 0.56489\n",
      "step: 2149 time 0:00:34.171653 train NMSE 0.58097 test NMSE 0.56375\n",
      "step: 2199 time 0:00:32.843333 train NMSE 0.57542 test NMSE 0.56520\n",
      "step: 2249 time 0:00:32.869240 train NMSE 0.58256 test NMSE 0.55721\n",
      "Model saved!\n",
      "step: 2299 time 0:00:32.822410 train NMSE 0.57845 test NMSE 0.56406\n",
      "step: 2349 time 0:00:32.843822 train NMSE 0.57289 test NMSE 0.56380\n",
      "step: 2399 time 0:00:32.877241 train NMSE 0.57223 test NMSE 0.56375\n",
      "step: 2449 time 0:00:32.871234 train NMSE 0.57118 test NMSE 0.54745\n",
      "Model saved!\n",
      "step: 2499 time 0:00:32.854955 train NMSE 0.56800 test NMSE 0.55041\n",
      "step: 2549 time 0:00:32.801416 train NMSE 0.56360 test NMSE 0.54533\n",
      "Model saved!\n",
      "step: 2599 time 0:00:32.812273 train NMSE 0.56247 test NMSE 0.52000\n",
      "Model saved!\n",
      "step: 2649 time 0:00:32.813034 train NMSE 0.56015 test NMSE 0.53003\n",
      "step: 2699 time 0:00:32.842829 train NMSE 0.55898 test NMSE 0.52423\n",
      "step: 2749 time 0:00:32.815108 train NMSE 0.55010 test NMSE 0.53937\n",
      "step: 2799 time 0:00:32.809936 train NMSE 0.55126 test NMSE 0.54903\n",
      "step: 2849 time 0:00:32.805956 train NMSE 0.55242 test NMSE 0.54476\n",
      "step: 2899 time 0:00:34.146849 train NMSE 0.54480 test NMSE 0.53151\n",
      "step: 2949 time 0:00:32.862776 train NMSE 0.54680 test NMSE 0.52712\n",
      "step: 2999 time 0:00:32.830393 train NMSE 0.54103 test NMSE 0.52753\n",
      "step: 3049 time 0:00:32.838263 train NMSE 0.54012 test NMSE 0.51151\n",
      "Model saved!\n",
      "step: 3099 time 0:00:32.886698 train NMSE 0.53649 test NMSE 0.52087\n",
      "step: 3149 time 0:00:32.855261 train NMSE 0.53404 test NMSE 0.51348\n",
      "step: 3199 time 0:00:32.830228 train NMSE 0.52833 test NMSE 0.51189\n",
      "step: 3249 time 0:00:34.189741 train NMSE 0.53338 test NMSE 0.50494\n",
      "Model saved!\n",
      "step: 3299 time 0:00:32.821242 train NMSE 0.52712 test NMSE 0.49801\n",
      "Model saved!\n",
      "step: 3349 time 0:00:32.868286 train NMSE 0.52414 test NMSE 0.48854\n",
      "Model saved!\n",
      "step: 3399 time 0:00:32.900175 train NMSE 0.52602 test NMSE 0.51574\n",
      "step: 3449 time 0:00:32.845352 train NMSE 0.51972 test NMSE 0.52361\n",
      "step: 3499 time 0:00:32.880529 train NMSE 0.51794 test NMSE 0.49885\n",
      "step: 3549 time 0:00:34.379933 train NMSE 0.51287 test NMSE 0.52657\n",
      "step: 3599 time 0:00:33.100168 train NMSE 0.51625 test NMSE 0.49540\n",
      "step: 3649 time 0:00:33.076736 train NMSE 0.51306 test NMSE 0.51990\n",
      "step: 3699 time 0:00:33.063251 train NMSE 0.50886 test NMSE 0.49226\n",
      "step: 3749 time 0:00:33.042051 train NMSE 0.50582 test NMSE 0.48987\n",
      "step: 3799 time 0:00:34.408607 train NMSE 0.50530 test NMSE 0.47364\n",
      "Model saved!\n",
      "step: 3849 time 0:00:33.065901 train NMSE 0.50093 test NMSE 0.44633\n",
      "Model saved!\n",
      "step: 3899 time 0:00:33.053059 train NMSE 0.48888 test NMSE 0.48858\n",
      "step: 3949 time 0:00:33.066020 train NMSE 0.49120 test NMSE 0.45579\n",
      "step: 3999 time 0:00:33.077741 train NMSE 0.48151 test NMSE 0.47078\n",
      "step: 4049 time 0:00:34.386126 train NMSE 0.48422 test NMSE 0.47069\n",
      "step: 4099 time 0:00:33.072737 train NMSE 0.48269 test NMSE 0.46556\n",
      "step: 4149 time 0:00:33.071337 train NMSE 0.48004 test NMSE 0.45292\n",
      "step: 4199 time 0:00:33.076167 train NMSE 0.47775 test NMSE 0.49770\n",
      "step: 4249 time 0:00:33.062263 train NMSE 0.47462 test NMSE 0.41802\n",
      "Model saved!\n",
      "step: 4299 time 0:00:33.100279 train NMSE 0.46847 test NMSE 0.48143\n",
      "step: 4349 time 0:00:33.075139 train NMSE 0.46900 test NMSE 0.46671\n",
      "step: 4399 time 0:00:33.054950 train NMSE 0.47153 test NMSE 0.43140\n",
      "step: 4449 time 0:00:33.065783 train NMSE 0.46645 test NMSE 0.44937\n",
      "step: 4499 time 0:00:33.060534 train NMSE 0.46944 test NMSE 0.45666\n",
      "step: 4549 time 0:00:33.035219 train NMSE 0.46128 test NMSE 0.43906\n",
      "step: 4599 time 0:00:33.034496 train NMSE 0.46279 test NMSE 0.43942\n",
      "step: 4649 time 0:00:33.007618 train NMSE 0.46526 test NMSE 0.47122\n",
      "step: 4699 time 0:00:33.080176 train NMSE 0.46485 test NMSE 0.41075\n",
      "Model saved!\n",
      "step: 4749 time 0:00:33.042574 train NMSE 0.45196 test NMSE 0.41560\n",
      "step: 4799 time 0:00:34.370627 train NMSE 0.45261 test NMSE 0.44791\n",
      "step: 4849 time 0:00:33.100101 train NMSE 0.45504 test NMSE 0.39917\n",
      "Model saved!\n",
      "step: 4899 time 0:00:33.072617 train NMSE 0.45381 test NMSE 0.45284\n",
      "step: 4949 time 0:00:33.141952 train NMSE 0.45694 test NMSE 0.44783\n",
      "step: 4999 time 0:00:33.018536 train NMSE 0.44233 test NMSE 0.42333\n",
      "step: 5049 time 0:00:34.399163 train NMSE 0.45071 test NMSE 0.38239\n",
      "Model saved!\n",
      "step: 5099 time 0:00:33.040304 train NMSE 0.45245 test NMSE 0.42301\n",
      "step: 5149 time 0:00:33.010064 train NMSE 0.44051 test NMSE 0.40363\n",
      "step: 5199 time 0:00:33.049791 train NMSE 0.44788 test NMSE 0.41390\n",
      "step: 5249 time 0:00:33.048218 train NMSE 0.44096 test NMSE 0.41976\n",
      "step: 5299 time 0:00:33.066250 train NMSE 0.44854 test NMSE 0.38708\n",
      "step: 5349 time 0:00:34.412633 train NMSE 0.44600 test NMSE 0.40318\n",
      "step: 5399 time 0:00:33.045828 train NMSE 0.44587 test NMSE 0.38865\n",
      "step: 5449 time 0:00:33.060263 train NMSE 0.43589 test NMSE 0.39131\n",
      "step: 5499 time 0:00:33.040267 train NMSE 0.43260 test NMSE 0.40705\n",
      "step: 5549 time 0:00:32.998904 train NMSE 0.44034 test NMSE 0.41303\n",
      "step: 5599 time 0:00:33.069711 train NMSE 0.43202 test NMSE 0.39521\n",
      "step: 5649 time 0:00:33.022872 train NMSE 0.43683 test NMSE 0.37738\n",
      "Model saved!\n",
      "step: 5699 time 0:00:33.023711 train NMSE 0.44299 test NMSE 0.37335\n",
      "Model saved!\n",
      "step: 5749 time 0:00:33.066220 train NMSE 0.42783 test NMSE 0.38030\n",
      "step: 5799 time 0:00:33.048241 train NMSE 0.43301 test NMSE 0.41024\n",
      "step: 5849 time 0:00:33.036149 train NMSE 0.43521 test NMSE 0.38349\n",
      "step: 5899 time 0:00:33.072815 train NMSE 0.42973 test NMSE 0.41118\n",
      "step: 5949 time 0:00:34.361275 train NMSE 0.42742 test NMSE 0.38105\n",
      "step: 5999 time 0:00:33.010899 train NMSE 0.43208 test NMSE 0.44566\n",
      "step: 6049 time 0:00:33.043743 train NMSE 0.42417 test NMSE 0.35284\n",
      "Model saved!\n",
      "step: 6099 time 0:00:33.054280 train NMSE 0.42704 test NMSE 0.41028\n",
      "step: 6149 time 0:00:33.039694 train NMSE 0.43120 test NMSE 0.35397\n",
      "step: 6199 time 0:00:32.968881 train NMSE 0.43706 test NMSE 0.37932\n",
      "step: 6249 time 0:00:32.989463 train NMSE 0.42400 test NMSE 0.39496\n",
      "step: 6299 time 0:00:33.061025 train NMSE 0.41761 test NMSE 0.38347\n",
      "step: 6349 time 0:00:33.045081 train NMSE 0.43589 test NMSE 0.38076\n",
      "step: 6399 time 0:00:33.009871 train NMSE 0.43011 test NMSE 0.40517\n",
      "step: 6449 time 0:00:33.042262 train NMSE 0.43152 test NMSE 0.38250\n",
      "step: 6499 time 0:00:33.055755 train NMSE 0.42254 test NMSE 0.41619\n",
      "step: 6549 time 0:00:34.416593 train NMSE 0.42869 test NMSE 0.34591\n",
      "Model saved!\n",
      "step: 6599 time 0:00:33.057836 train NMSE 0.42298 test NMSE 0.41495\n",
      "step: 6649 time 0:00:33.050319 train NMSE 0.43938 test NMSE 0.38484\n",
      "step: 6699 time 0:00:33.051102 train NMSE 0.42557 test NMSE 0.40416\n",
      "step: 6749 time 0:00:33.043255 train NMSE 0.41943 test NMSE 0.39205\n",
      "step: 6799 time 0:00:33.033217 train NMSE 0.42177 test NMSE 0.35106\n",
      "step: 6849 time 0:00:33.069180 train NMSE 0.41924 test NMSE 0.42082\n",
      "step: 6899 time 0:00:33.085615 train NMSE 0.41648 test NMSE 0.36034\n",
      "step: 6949 time 0:00:33.011064 train NMSE 0.42250 test NMSE 0.39886\n",
      "step: 6999 time 0:00:33.039555 train NMSE 0.42999 test NMSE 0.39946\n",
      "step: 7049 time 0:00:33.032852 train NMSE 0.41423 test NMSE 0.39458\n",
      "step: 7099 time 0:00:33.056262 train NMSE 0.41854 test NMSE 0.32885\n",
      "Model saved!\n",
      "step: 7149 time 0:00:33.030011 train NMSE 0.40917 test NMSE 0.38222\n",
      "step: 7199 time 0:00:33.050754 train NMSE 0.41639 test NMSE 0.37127\n",
      "step: 7249 time 0:00:32.985478 train NMSE 0.41341 test NMSE 0.36733\n",
      "step: 7299 time 0:00:33.016011 train NMSE 0.40984 test NMSE 0.40635\n",
      "step: 7349 time 0:00:33.064286 train NMSE 0.41575 test NMSE 0.40414\n",
      "step: 7399 time 0:00:33.059216 train NMSE 0.41494 test NMSE 0.41950\n",
      "step: 7449 time 0:00:33.028361 train NMSE 0.41320 test NMSE 0.37683\n",
      "step: 7499 time 0:00:34.388778 train NMSE 0.41274 test NMSE 0.41330\n",
      "step: 7549 time 0:00:33.006899 train NMSE 0.40991 test NMSE 0.40756\n",
      "step: 7599 time 0:00:33.037739 train NMSE 0.41504 test NMSE 0.38571\n",
      "step: 7649 time 0:00:33.024880 train NMSE 0.41513 test NMSE 0.41950\n",
      "step: 7699 time 0:00:33.041951 train NMSE 0.41408 test NMSE 0.38592\n",
      "step: 7749 time 0:00:33.028294 train NMSE 0.40371 test NMSE 0.38675\n",
      "step: 7799 time 0:00:33.017198 train NMSE 0.40743 test NMSE 0.37200\n",
      "step: 7849 time 0:00:34.342333 train NMSE 0.41066 test NMSE 0.40658\n",
      "step: 7899 time 0:00:33.007086 train NMSE 0.40722 test NMSE 0.37531\n",
      "step: 7949 time 0:00:32.964475 train NMSE 0.40395 test NMSE 0.40913\n",
      "step: 7999 time 0:00:33.020882 train NMSE 0.40155 test NMSE 0.39445\n",
      "step: 8049 time 0:00:33.031342 train NMSE 0.39631 test NMSE 0.38182\n",
      "step: 8099 time 0:00:33.032998 train NMSE 0.40177 test NMSE 0.36122\n",
      "step: 8149 time 0:00:33.021875 train NMSE 0.40219 test NMSE 0.34211\n",
      "step: 8199 time 0:00:34.384172 train NMSE 0.40257 test NMSE 0.36072\n",
      "step: 8249 time 0:00:33.031815 train NMSE 0.40620 test NMSE 0.40191\n",
      "step: 8299 time 0:00:33.032977 train NMSE 0.40606 test NMSE 0.38444\n",
      "step: 8349 time 0:00:33.032829 train NMSE 0.39904 test NMSE 0.34660\n",
      "step: 8399 time 0:00:33.004376 train NMSE 0.41076 test NMSE 0.37180\n",
      "step: 8449 time 0:00:33.031801 train NMSE 0.40398 test NMSE 0.38377\n",
      "step: 8499 time 0:00:33.028295 train NMSE 0.40002 test NMSE 0.37148\n",
      "step: 8549 time 0:00:34.423567 train NMSE 0.40077 test NMSE 0.35459\n",
      "step: 8599 time 0:00:33.058329 train NMSE 0.39924 test NMSE 0.33891\n",
      "step: 8649 time 0:00:33.022340 train NMSE 0.39794 test NMSE 0.34983\n",
      "step: 8699 time 0:00:33.035321 train NMSE 0.39965 test NMSE 0.35117\n",
      "step: 8749 time 0:00:32.992454 train NMSE 0.39538 test NMSE 0.35178\n",
      "step: 8799 time 0:00:33.030313 train NMSE 0.39212 test NMSE 0.32534\n",
      "Model saved!\n",
      "step: 8849 time 0:00:33.054228 train NMSE 0.39291 test NMSE 0.33021\n",
      "step: 8899 time 0:00:34.392698 train NMSE 0.39552 test NMSE 0.34928\n",
      "step: 8949 time 0:00:32.969702 train NMSE 0.39420 test NMSE 0.33623\n",
      "step: 8999 time 0:00:33.041130 train NMSE 0.39570 test NMSE 0.34134\n",
      "step: 9049 time 0:00:33.025378 train NMSE 0.40139 test NMSE 0.35761\n",
      "step: 9099 time 0:00:33.044223 train NMSE 0.39565 test NMSE 0.34815\n",
      "step: 9149 time 0:00:33.035825 train NMSE 0.38734 test NMSE 0.35167\n",
      "step: 9199 time 0:00:33.029328 train NMSE 0.39361 test NMSE 0.35834\n",
      "step: 9249 time 0:00:33.040294 train NMSE 0.38954 test NMSE 0.36333\n",
      "step: 9299 time 0:00:33.036836 train NMSE 0.38955 test NMSE 0.36159\n",
      "step: 9349 time 0:00:33.050765 train NMSE 0.39179 test NMSE 0.34267\n",
      "step: 9399 time 0:00:33.007367 train NMSE 0.38288 test NMSE 0.39348\n",
      "step: 9449 time 0:00:32.983787 train NMSE 0.38842 test NMSE 0.30172\n",
      "Model saved!\n",
      "step: 9499 time 0:00:33.010896 train NMSE 0.38278 test NMSE 0.33774\n",
      "step: 9549 time 0:00:34.424248 train NMSE 0.38863 test NMSE 0.36531\n",
      "step: 9599 time 0:00:33.027306 train NMSE 0.39086 test NMSE 0.35488\n",
      "step: 9649 time 0:00:33.042411 train NMSE 0.38129 test NMSE 0.35173\n",
      "step: 9699 time 0:00:32.979433 train NMSE 0.37775 test NMSE 0.37075\n",
      "step: 9749 time 0:00:37.699839 train NMSE 0.37939 test NMSE 0.36014\n",
      "step: 9799 time 0:00:36.017724 train NMSE 0.38213 test NMSE 0.34039\n",
      "step: 9849 time 0:00:36.120663 train NMSE 0.37991 test NMSE 0.36220\n",
      "step: 9899 time 0:00:37.760629 train NMSE 0.38307 test NMSE 0.36370\n",
      "step: 9949 time 0:00:37.818553 train NMSE 0.38350 test NMSE 0.38073\n",
      "step: 9999 time 0:00:37.782588 train NMSE 0.38165 test NMSE 0.31987\n"
     ]
    }
   ],
   "source": [
    "B = 16\n",
    "\n",
    "batch = 64;   #用户数\n",
    "num_batch = 1\n",
    "STEPS = 10000\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "train_csiNet(param_list,num_batch,batch,STEPS,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 49 time 0:00:57.854232 train NMSE 1.02997 test NMSE 1.00084\n",
      "step: 99 time 0:00:55.953507 train NMSE 0.99885 test NMSE 0.99408\n",
      "Model saved!\n",
      "step: 149 time 0:00:56.005501 train NMSE 0.96687 test NMSE 0.93665\n",
      "Model saved!\n",
      "step: 199 time 0:00:55.928976 train NMSE 0.86472 test NMSE 0.82075\n",
      "Model saved!\n",
      "step: 249 time 0:00:52.908381 train NMSE 0.79274 test NMSE 0.78855\n",
      "Model saved!\n",
      "step: 299 time 0:00:53.001062 train NMSE 0.76019 test NMSE 0.74179\n",
      "Model saved!\n",
      "step: 349 time 0:00:56.020829 train NMSE 0.73327 test NMSE 0.72676\n",
      "Model saved!\n",
      "step: 399 time 0:00:56.254396 train NMSE 0.70989 test NMSE 0.67509\n",
      "Model saved!\n",
      "step: 449 time 0:00:53.310002 train NMSE 0.67045 test NMSE 0.64165\n",
      "Model saved!\n",
      "step: 499 time 0:00:56.349970 train NMSE 0.62024 test NMSE 0.60026\n",
      "Model saved!\n",
      "step: 549 time 0:00:56.086186 train NMSE 0.58869 test NMSE 0.56823\n",
      "Model saved!\n",
      "step: 599 time 0:00:52.949418 train NMSE 0.55407 test NMSE 0.54256\n",
      "Model saved!\n",
      "step: 649 time 0:00:55.912656 train NMSE 0.53958 test NMSE 0.50858\n",
      "Model saved!\n",
      "step: 699 time 0:00:55.916894 train NMSE 0.52138 test NMSE 0.53443\n",
      "step: 749 time 0:00:52.939292 train NMSE 0.50467 test NMSE 0.46992\n",
      "Model saved!\n",
      "step: 799 time 0:00:55.933542 train NMSE 0.49027 test NMSE 0.47481\n",
      "step: 849 time 0:00:52.868759 train NMSE 0.46929 test NMSE 0.46304\n",
      "Model saved!\n",
      "step: 899 time 0:00:52.904726 train NMSE 0.45777 test NMSE 0.41641\n",
      "Model saved!\n",
      "step: 949 time 0:00:55.883176 train NMSE 0.44146 test NMSE 0.42522\n",
      "step: 999 time 0:00:52.900201 train NMSE 0.43152 test NMSE 0.42144\n",
      "step: 1049 time 0:00:55.898734 train NMSE 0.41752 test NMSE 0.36078\n",
      "Model saved!\n",
      "step: 1099 time 0:00:55.893758 train NMSE 0.40041 test NMSE 0.40202\n",
      "step: 1149 time 0:00:52.850884 train NMSE 0.39578 test NMSE 0.40588\n",
      "step: 1199 time 0:00:55.952041 train NMSE 0.37872 test NMSE 0.36777\n",
      "step: 1249 time 0:00:52.911663 train NMSE 0.36311 test NMSE 0.34727\n",
      "Model saved!\n",
      "step: 1299 time 0:00:55.908861 train NMSE 0.34418 test NMSE 0.31393\n",
      "Model saved!\n",
      "step: 1349 time 0:00:56.144107 train NMSE 0.33109 test NMSE 0.30682\n",
      "Model saved!\n",
      "step: 1399 time 0:00:53.139140 train NMSE 0.31659 test NMSE 0.29945\n",
      "Model saved!\n",
      "step: 1449 time 0:00:55.970593 train NMSE 0.31020 test NMSE 0.27885\n",
      "Model saved!\n",
      "step: 1499 time 0:00:52.930384 train NMSE 0.29966 test NMSE 0.26038\n",
      "Model saved!\n",
      "step: 1549 time 0:00:55.970957 train NMSE 0.29033 test NMSE 0.28843\n",
      "step: 1599 time 0:00:55.968529 train NMSE 0.27810 test NMSE 0.25394\n",
      "Model saved!\n",
      "step: 1649 time 0:00:52.943669 train NMSE 0.26983 test NMSE 0.26223\n",
      "step: 1699 time 0:00:55.970009 train NMSE 0.26192 test NMSE 0.26352\n",
      "step: 1749 time 0:00:53.164029 train NMSE 0.26379 test NMSE 0.23264\n",
      "Model saved!\n",
      "step: 1799 time 0:00:55.976367 train NMSE 0.24552 test NMSE 0.24587\n",
      "step: 1849 time 0:00:52.953663 train NMSE 0.23719 test NMSE 0.22479\n",
      "Model saved!\n",
      "step: 1899 time 0:00:52.933632 train NMSE 0.23079 test NMSE 0.20955\n",
      "Model saved!\n",
      "step: 1949 time 0:00:55.993446 train NMSE 0.22578 test NMSE 0.19391\n",
      "Model saved!\n",
      "step: 1999 time 0:00:52.946147 train NMSE 0.21507 test NMSE 0.17824\n",
      "Model saved!\n",
      "step: 2049 time 0:00:55.947384 train NMSE 0.22055 test NMSE 0.20156\n",
      "step: 2099 time 0:00:52.930088 train NMSE 0.20929 test NMSE 0.17671\n",
      "Model saved!\n",
      "step: 2149 time 0:00:56.028642 train NMSE 0.20626 test NMSE 0.17199\n",
      "Model saved!\n",
      "step: 2199 time 0:00:55.950587 train NMSE 0.20210 test NMSE 0.17323\n",
      "step: 2249 time 0:00:52.930681 train NMSE 0.20051 test NMSE 0.15800\n",
      "Model saved!\n",
      "step: 2299 time 0:00:55.982631 train NMSE 0.19860 test NMSE 0.18777\n",
      "step: 2349 time 0:00:52.960793 train NMSE 0.18303 test NMSE 0.16829\n",
      "step: 2399 time 0:00:55.976245 train NMSE 0.18903 test NMSE 0.17322\n",
      "step: 2449 time 0:00:55.963233 train NMSE 0.18600 test NMSE 0.16036\n",
      "step: 2499 time 0:00:52.947842 train NMSE 0.19231 test NMSE 0.14918\n",
      "Model saved!\n",
      "step: 2549 time 0:00:55.986793 train NMSE 0.18060 test NMSE 0.15106\n",
      "step: 2599 time 0:00:52.940580 train NMSE 0.18319 test NMSE 0.15441\n",
      "step: 2649 time 0:00:56.175435 train NMSE 0.18260 test NMSE 0.14101\n",
      "Model saved!\n",
      "step: 2699 time 0:00:53.043559 train NMSE 0.18250 test NMSE 0.14488\n",
      "step: 2749 time 0:00:55.955422 train NMSE 0.17434 test NMSE 0.15394\n",
      "step: 2799 time 0:00:55.949060 train NMSE 0.17507 test NMSE 0.17671\n",
      "step: 2849 time 0:00:52.963518 train NMSE 0.17514 test NMSE 0.13890\n",
      "Model saved!\n",
      "step: 2899 time 0:00:55.976001 train NMSE 0.17851 test NMSE 0.13110\n",
      "Model saved!\n",
      "step: 2949 time 0:00:52.962271 train NMSE 0.16772 test NMSE 0.14099\n",
      "step: 2999 time 0:00:55.944588 train NMSE 0.16947 test NMSE 0.13747\n",
      "step: 3049 time 0:00:53.145234 train NMSE 0.16540 test NMSE 0.12772\n",
      "Model saved!\n",
      "step: 3099 time 0:00:55.970165 train NMSE 0.17025 test NMSE 0.13780\n",
      "step: 3149 time 0:00:55.974979 train NMSE 0.16476 test NMSE 0.13039\n",
      "step: 3199 time 0:00:52.937979 train NMSE 0.16583 test NMSE 0.13189\n",
      "step: 3249 time 0:00:55.971021 train NMSE 0.16923 test NMSE 0.13268\n",
      "step: 3299 time 0:00:53.017891 train NMSE 0.17414 test NMSE 0.13173\n",
      "step: 3349 time 0:00:56.178465 train NMSE 0.16001 test NMSE 0.12802\n",
      "step: 3399 time 0:00:52.942717 train NMSE 0.15676 test NMSE 0.14007\n",
      "step: 3449 time 0:00:55.970548 train NMSE 0.16280 test NMSE 0.14640\n",
      "step: 3499 time 0:00:55.975536 train NMSE 0.16111 test NMSE 0.13128\n",
      "step: 3549 time 0:00:53.146520 train NMSE 0.15597 test NMSE 0.12045\n",
      "Model saved!\n",
      "step: 3599 time 0:00:55.972464 train NMSE 0.16369 test NMSE 0.12654\n",
      "step: 3649 time 0:00:53.151326 train NMSE 0.15887 test NMSE 0.11867\n",
      "Model saved!\n",
      "step: 3699 time 0:00:56.189597 train NMSE 0.15661 test NMSE 0.13582\n",
      "step: 3749 time 0:00:55.944630 train NMSE 0.15646 test NMSE 0.14734\n",
      "step: 3799 time 0:00:53.009622 train NMSE 0.15408 test NMSE 0.13623\n",
      "step: 3849 time 0:00:55.998510 train NMSE 0.15601 test NMSE 0.12035\n",
      "step: 3899 time 0:00:52.936760 train NMSE 0.15467 test NMSE 0.11719\n",
      "Model saved!\n",
      "step: 3949 time 0:00:55.977358 train NMSE 0.15038 test NMSE 0.13096\n",
      "step: 3999 time 0:00:52.949090 train NMSE 0.15026 test NMSE 0.12008\n",
      "step: 4049 time 0:00:55.988391 train NMSE 0.15059 test NMSE 0.13030\n",
      "step: 4099 time 0:00:55.977393 train NMSE 0.15425 test NMSE 0.11458\n",
      "Model saved!\n",
      "step: 4149 time 0:00:52.929708 train NMSE 0.15161 test NMSE 0.13523\n",
      "step: 4199 time 0:00:55.970327 train NMSE 0.14641 test NMSE 0.14699\n",
      "step: 4249 time 0:00:53.126041 train NMSE 0.15249 test NMSE 0.11826\n",
      "step: 4299 time 0:00:56.143334 train NMSE 0.14824 test NMSE 0.10965\n",
      "Model saved!\n",
      "step: 4349 time 0:00:55.966441 train NMSE 0.15025 test NMSE 0.12472\n",
      "step: 4399 time 0:00:52.955082 train NMSE 0.14624 test NMSE 0.12401\n",
      "step: 4449 time 0:00:56.178940 train NMSE 0.15265 test NMSE 0.11825\n",
      "step: 4499 time 0:00:53.134608 train NMSE 0.14976 test NMSE 0.12258\n",
      "step: 4549 time 0:00:55.962118 train NMSE 0.14536 test NMSE 0.12245\n",
      "step: 4599 time 0:00:52.964335 train NMSE 0.14128 test NMSE 0.11444\n",
      "step: 4649 time 0:00:55.979434 train NMSE 0.14605 test NMSE 0.13410\n",
      "step: 4699 time 0:00:55.995004 train NMSE 0.14789 test NMSE 0.12388\n",
      "step: 4749 time 0:00:53.236707 train NMSE 0.14474 test NMSE 0.10961\n",
      "Model saved!\n",
      "step: 4799 time 0:00:55.934663 train NMSE 0.15277 test NMSE 0.13268\n",
      "step: 4849 time 0:00:52.976962 train NMSE 0.14477 test NMSE 0.10647\n",
      "Model saved!\n",
      "step: 4899 time 0:00:55.964090 train NMSE 0.14380 test NMSE 0.11985\n",
      "step: 4949 time 0:00:55.986154 train NMSE 0.14477 test NMSE 0.12767\n",
      "step: 4999 time 0:00:53.039766 train NMSE 0.15145 test NMSE 0.12783\n",
      "step: 5049 time 0:00:55.966103 train NMSE 0.14550 test NMSE 0.10333\n",
      "Model saved!\n",
      "step: 5099 time 0:00:52.961085 train NMSE 0.14083 test NMSE 0.11676\n",
      "step: 5149 time 0:00:55.977868 train NMSE 0.13703 test NMSE 0.11072\n",
      "step: 5199 time 0:00:53.022480 train NMSE 0.13931 test NMSE 0.10599\n",
      "step: 5249 time 0:00:53.134162 train NMSE 0.14147 test NMSE 0.11603\n",
      "step: 5299 time 0:00:55.969348 train NMSE 0.14128 test NMSE 0.11615\n",
      "step: 5349 time 0:00:53.387108 train NMSE 0.13979 test NMSE 0.10553\n",
      "step: 5399 time 0:00:55.962062 train NMSE 0.13453 test NMSE 0.10516\n",
      "step: 5449 time 0:00:52.924199 train NMSE 0.13879 test NMSE 0.10932\n",
      "step: 5499 time 0:00:55.982972 train NMSE 0.13546 test NMSE 0.11051\n",
      "step: 5549 time 0:00:52.947655 train NMSE 0.13650 test NMSE 0.10430\n",
      "step: 5599 time 0:00:55.969135 train NMSE 0.13874 test NMSE 0.11429\n",
      "step: 5649 time 0:00:55.972637 train NMSE 0.13933 test NMSE 0.10530\n",
      "step: 5699 time 0:00:53.167021 train NMSE 0.13682 test NMSE 0.09244\n",
      "Model saved!\n",
      "step: 5749 time 0:00:55.982819 train NMSE 0.13210 test NMSE 0.11228\n",
      "step: 5799 time 0:00:52.930191 train NMSE 0.13448 test NMSE 0.10019\n",
      "step: 5849 time 0:00:55.953661 train NMSE 0.13766 test NMSE 0.09880\n",
      "step: 5899 time 0:00:55.922928 train NMSE 0.13634 test NMSE 0.11243\n",
      "step: 5949 time 0:00:52.905768 train NMSE 0.13450 test NMSE 0.10569\n",
      "step: 5999 time 0:00:55.969412 train NMSE 0.13758 test NMSE 0.11027\n",
      "step: 6049 time 0:00:52.951678 train NMSE 0.13781 test NMSE 0.11419\n",
      "step: 6099 time 0:00:55.931424 train NMSE 0.13636 test NMSE 0.11373\n",
      "step: 6149 time 0:00:52.956305 train NMSE 0.13480 test NMSE 0.12754\n",
      "step: 6199 time 0:00:55.919200 train NMSE 0.12975 test NMSE 0.10907\n",
      "step: 6249 time 0:00:55.948043 train NMSE 0.13430 test NMSE 0.10919\n",
      "step: 6299 time 0:00:53.131757 train NMSE 0.13397 test NMSE 0.11487\n",
      "step: 6349 time 0:00:55.913069 train NMSE 0.13381 test NMSE 0.13239\n",
      "step: 6399 time 0:00:52.890321 train NMSE 0.14028 test NMSE 0.10933\n",
      "step: 6449 time 0:00:55.951536 train NMSE 0.13573 test NMSE 0.10924\n",
      "step: 6499 time 0:00:52.929856 train NMSE 0.13483 test NMSE 0.09852\n",
      "step: 6549 time 0:00:55.925314 train NMSE 0.13159 test NMSE 0.11977\n",
      "step: 6599 time 0:00:52.921753 train NMSE 0.13083 test NMSE 0.10826\n",
      "step: 6649 time 0:00:52.991869 train NMSE 0.13004 test NMSE 0.10249\n",
      "step: 6699 time 0:00:55.952838 train NMSE 0.13230 test NMSE 0.09847\n",
      "step: 6749 time 0:00:52.943781 train NMSE 0.12789 test NMSE 0.09980\n",
      "step: 6799 time 0:00:55.951983 train NMSE 0.13124 test NMSE 0.10059\n",
      "step: 6849 time 0:00:52.935693 train NMSE 0.13006 test NMSE 0.09028\n",
      "Model saved!\n",
      "step: 6899 time 0:00:55.949968 train NMSE 0.12550 test NMSE 0.10482\n",
      "step: 6949 time 0:00:52.900300 train NMSE 0.12520 test NMSE 0.09901\n",
      "step: 6999 time 0:00:52.921592 train NMSE 0.13516 test NMSE 0.10496\n",
      "step: 7049 time 0:00:55.921211 train NMSE 0.13249 test NMSE 0.10508\n",
      "step: 7099 time 0:00:52.914128 train NMSE 0.13245 test NMSE 0.10697\n",
      "step: 7149 time 0:00:56.146143 train NMSE 0.12539 test NMSE 0.11002\n",
      "step: 7199 time 0:00:52.897422 train NMSE 0.12722 test NMSE 0.10195\n",
      "step: 7249 time 0:00:55.952108 train NMSE 0.12652 test NMSE 0.11384\n",
      "step: 7299 time 0:00:52.934224 train NMSE 0.13032 test NMSE 0.09726\n",
      "step: 7349 time 0:00:52.928962 train NMSE 0.13035 test NMSE 0.09644\n",
      "step: 7399 time 0:00:55.942438 train NMSE 0.12570 test NMSE 0.09528\n",
      "step: 7449 time 0:00:52.928101 train NMSE 0.12574 test NMSE 0.10705\n",
      "step: 7499 time 0:00:55.972861 train NMSE 0.12542 test NMSE 0.09310\n",
      "step: 7549 time 0:00:52.934327 train NMSE 0.13029 test NMSE 0.09801\n",
      "step: 7599 time 0:00:55.925759 train NMSE 0.12840 test NMSE 0.09705\n",
      "step: 7649 time 0:00:55.953433 train NMSE 0.12459 test NMSE 0.10638\n",
      "step: 7699 time 0:00:53.306875 train NMSE 0.12716 test NMSE 0.09093\n",
      "step: 7749 time 0:00:55.941379 train NMSE 0.12656 test NMSE 0.09764\n",
      "step: 7799 time 0:00:53.350477 train NMSE 0.12686 test NMSE 0.10593\n",
      "step: 7849 time 0:00:55.943725 train NMSE 0.12504 test NMSE 0.10833\n",
      "step: 7899 time 0:00:52.933891 train NMSE 0.12451 test NMSE 0.09369\n",
      "step: 7949 time 0:00:55.918649 train NMSE 0.12312 test NMSE 0.10365\n",
      "step: 7999 time 0:00:53.178696 train NMSE 0.12093 test NMSE 0.10415\n",
      "step: 8049 time 0:00:52.917340 train NMSE 0.12290 test NMSE 0.09888\n",
      "step: 8099 time 0:00:55.959109 train NMSE 0.12349 test NMSE 0.08842\n",
      "Model saved!\n",
      "step: 8149 time 0:00:52.884769 train NMSE 0.12488 test NMSE 0.09680\n",
      "step: 8199 time 0:00:55.955158 train NMSE 0.12142 test NMSE 0.09587\n",
      "step: 8249 time 0:00:52.922444 train NMSE 0.12083 test NMSE 0.09892\n",
      "step: 8299 time 0:00:55.911861 train NMSE 0.12493 test NMSE 0.09257\n",
      "step: 8349 time 0:00:52.926141 train NMSE 0.12424 test NMSE 0.10098\n",
      "step: 8399 time 0:00:52.958981 train NMSE 0.12481 test NMSE 0.09791\n",
      "step: 8449 time 0:00:55.947988 train NMSE 0.12159 test NMSE 0.08476\n",
      "Model saved!\n",
      "step: 8499 time 0:00:52.916583 train NMSE 0.12203 test NMSE 0.10355\n",
      "step: 8549 time 0:00:56.180833 train NMSE 0.12744 test NMSE 0.09880\n",
      "step: 8599 time 0:00:53.121216 train NMSE 0.12324 test NMSE 0.09433\n",
      "step: 8649 time 0:00:56.035291 train NMSE 0.12170 test NMSE 0.10684\n",
      "step: 8699 time 0:00:56.051780 train NMSE 0.11888 test NMSE 0.09126\n",
      "step: 8749 time 0:00:52.954402 train NMSE 0.12230 test NMSE 0.09710\n",
      "step: 8799 time 0:00:55.980500 train NMSE 0.12429 test NMSE 0.10112\n",
      "step: 8849 time 0:00:52.931553 train NMSE 0.12064 test NMSE 0.09711\n",
      "step: 8899 time 0:00:55.977051 train NMSE 0.12233 test NMSE 0.08387\n",
      "Model saved!\n",
      "step: 8949 time 0:00:52.947692 train NMSE 0.12103 test NMSE 0.09121\n",
      "step: 8999 time 0:00:55.972945 train NMSE 0.12583 test NMSE 0.08803\n",
      "step: 9049 time 0:00:55.943181 train NMSE 0.12374 test NMSE 0.09517\n",
      "step: 9099 time 0:00:52.946601 train NMSE 0.11769 test NMSE 0.09133\n",
      "step: 9149 time 0:00:55.959041 train NMSE 0.12217 test NMSE 0.08658\n",
      "step: 9199 time 0:00:52.954081 train NMSE 0.11956 test NMSE 0.09747\n",
      "step: 9249 time 0:00:55.967563 train NMSE 0.11948 test NMSE 0.09737\n",
      "step: 9299 time 0:00:52.950321 train NMSE 0.11916 test NMSE 0.09183\n",
      "step: 9349 time 0:00:55.964013 train NMSE 0.11638 test NMSE 0.10123\n",
      "step: 9399 time 0:00:55.961509 train NMSE 0.11866 test NMSE 0.08192\n",
      "Model saved!\n",
      "step: 9449 time 0:00:52.961563 train NMSE 0.11948 test NMSE 0.08050\n",
      "Model saved!\n",
      "step: 9499 time 0:00:55.963084 train NMSE 0.11820 test NMSE 0.09793\n",
      "step: 9549 time 0:00:52.960501 train NMSE 0.11781 test NMSE 0.09206\n",
      "step: 9599 time 0:00:55.979451 train NMSE 0.11904 test NMSE 0.09646\n",
      "step: 9649 time 0:00:52.943625 train NMSE 0.11927 test NMSE 0.09463\n",
      "step: 9699 time 0:00:52.935098 train NMSE 0.11516 test NMSE 0.09479\n",
      "step: 9749 time 0:00:55.963344 train NMSE 0.11614 test NMSE 0.09703\n",
      "step: 9799 time 0:00:52.969608 train NMSE 0.12271 test NMSE 0.10307\n",
      "step: 9849 time 0:00:55.951575 train NMSE 0.12047 test NMSE 0.09713\n",
      "step: 9899 time 0:00:52.939087 train NMSE 0.12076 test NMSE 0.09138\n",
      "step: 9949 time 0:00:55.960245 train NMSE 0.11692 test NMSE 0.09468\n",
      "step: 9999 time 0:00:56.004454 train NMSE 0.11595 test NMSE 0.10974\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "\n",
    "batch = 64;   #用户数\n",
    "num_batch = 1\n",
    "STEPS = 10000\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "train_csiNet(param_list,num_batch,batch,STEPS,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 49 time 0:00:53.428184 train NMSE 1.02426 test NMSE 0.99577\n",
      "Model saved!\n",
      "step: 99 time 0:00:56.512791 train NMSE 0.89524 test NMSE 0.81480\n",
      "Model saved!\n",
      "step: 149 time 0:00:53.444332 train NMSE 0.77403 test NMSE 0.72931\n",
      "Model saved!\n",
      "step: 199 time 0:00:56.614122 train NMSE 0.73024 test NMSE 0.69389\n",
      "Model saved!\n",
      "step: 249 time 0:00:53.451454 train NMSE 0.68771 test NMSE 0.66576\n",
      "Model saved!\n",
      "step: 299 time 0:00:56.522473 train NMSE 0.64074 test NMSE 0.60118\n",
      "Model saved!\n",
      "step: 349 time 0:00:53.452244 train NMSE 0.57866 test NMSE 0.55672\n",
      "Model saved!\n",
      "step: 399 time 0:00:56.537518 train NMSE 0.53622 test NMSE 0.49849\n",
      "Model saved!\n",
      "step: 449 time 0:00:56.736795 train NMSE 0.48895 test NMSE 0.47014\n",
      "Model saved!\n",
      "step: 499 time 0:00:53.341505 train NMSE 0.43360 test NMSE 0.39906\n",
      "Model saved!\n",
      "step: 549 time 0:00:56.501018 train NMSE 0.38708 test NMSE 0.35589\n",
      "Model saved!\n",
      "step: 599 time 0:00:56.586394 train NMSE 0.36202 test NMSE 0.34834\n",
      "Model saved!\n",
      "step: 649 time 0:00:53.462579 train NMSE 0.34106 test NMSE 0.29655\n",
      "Model saved!\n",
      "step: 699 time 0:00:56.501333 train NMSE 0.31118 test NMSE 0.30125\n",
      "step: 749 time 0:00:56.436260 train NMSE 0.30545 test NMSE 0.28112\n",
      "Model saved!\n",
      "step: 799 time 0:00:53.690059 train NMSE 0.28805 test NMSE 0.27264\n",
      "Model saved!\n",
      "step: 849 time 0:00:56.559450 train NMSE 0.27164 test NMSE 0.25502\n",
      "Model saved!\n",
      "step: 899 time 0:00:53.564435 train NMSE 0.25230 test NMSE 0.23604\n",
      "Model saved!\n",
      "step: 949 time 0:00:53.492643 train NMSE 0.23615 test NMSE 0.22609\n",
      "Model saved!\n",
      "step: 999 time 0:00:56.535294 train NMSE 0.22036 test NMSE 0.16933\n",
      "Model saved!\n",
      "step: 1049 time 0:00:53.438221 train NMSE 0.19790 test NMSE 0.18488\n",
      "step: 1099 time 0:00:56.485527 train NMSE 0.18796 test NMSE 0.16253\n",
      "Model saved!\n",
      "step: 1149 time 0:00:53.429795 train NMSE 0.17450 test NMSE 0.14690\n",
      "Model saved!\n",
      "step: 1199 time 0:00:53.432959 train NMSE 0.16400 test NMSE 0.13807\n",
      "Model saved!\n",
      "step: 1249 time 0:00:56.505566 train NMSE 0.15656 test NMSE 0.15221\n",
      "step: 1299 time 0:00:53.477763 train NMSE 0.15383 test NMSE 0.12370\n",
      "Model saved!\n",
      "step: 1349 time 0:00:56.730700 train NMSE 0.15507 test NMSE 0.12072\n",
      "Model saved!\n",
      "step: 1399 time 0:00:53.469626 train NMSE 0.14252 test NMSE 0.12554\n",
      "step: 1449 time 0:00:56.522683 train NMSE 0.13944 test NMSE 0.12155\n",
      "step: 1499 time 0:00:53.487014 train NMSE 0.13873 test NMSE 0.11913\n",
      "Model saved!\n",
      "step: 1549 time 0:00:56.540861 train NMSE 0.13011 test NMSE 0.12100\n",
      "step: 1599 time 0:00:53.476187 train NMSE 0.12860 test NMSE 0.10905\n",
      "Model saved!\n",
      "step: 1649 time 0:00:56.567401 train NMSE 0.13217 test NMSE 0.10531\n",
      "Model saved!\n",
      "step: 1699 time 0:00:53.477213 train NMSE 0.12714 test NMSE 0.11313\n",
      "step: 1749 time 0:00:56.578339 train NMSE 0.11950 test NMSE 0.10828\n",
      "step: 1799 time 0:00:53.471156 train NMSE 0.12180 test NMSE 0.09645\n",
      "Model saved!\n",
      "step: 1849 time 0:00:53.514144 train NMSE 0.12123 test NMSE 0.10489\n",
      "step: 1899 time 0:00:56.584965 train NMSE 0.11812 test NMSE 0.10007\n",
      "step: 1949 time 0:00:53.506241 train NMSE 0.11740 test NMSE 0.08459\n",
      "Model saved!\n",
      "step: 1999 time 0:00:56.579921 train NMSE 0.11570 test NMSE 0.10213\n",
      "step: 2049 time 0:00:53.513078 train NMSE 0.11111 test NMSE 0.08782\n",
      "step: 2099 time 0:00:56.769428 train NMSE 0.11008 test NMSE 0.08328\n",
      "Model saved!\n",
      "step: 2149 time 0:00:53.476215 train NMSE 0.10559 test NMSE 0.08124\n",
      "Model saved!\n",
      "step: 2199 time 0:00:56.560023 train NMSE 0.10759 test NMSE 0.09234\n",
      "step: 2249 time 0:00:53.484179 train NMSE 0.10503 test NMSE 0.07781\n",
      "Model saved!\n",
      "step: 2299 time 0:00:56.542111 train NMSE 0.10496 test NMSE 0.09296\n",
      "step: 2349 time 0:00:53.478237 train NMSE 0.10382 test NMSE 0.08260\n",
      "step: 2399 time 0:00:56.735432 train NMSE 0.10546 test NMSE 0.08934\n",
      "step: 2449 time 0:00:53.493114 train NMSE 0.10395 test NMSE 0.08042\n",
      "step: 2499 time 0:00:56.648690 train NMSE 0.10075 test NMSE 0.08111\n",
      "step: 2549 time 0:00:53.709106 train NMSE 0.10362 test NMSE 0.09445\n",
      "step: 2599 time 0:00:56.563447 train NMSE 0.09763 test NMSE 0.07831\n",
      "step: 2649 time 0:00:53.488116 train NMSE 0.09519 test NMSE 0.07866\n",
      "step: 2699 time 0:00:56.532564 train NMSE 0.09574 test NMSE 0.07404\n",
      "Model saved!\n",
      "step: 2749 time 0:00:53.510226 train NMSE 0.09984 test NMSE 0.08177\n",
      "step: 2799 time 0:00:56.547999 train NMSE 0.09374 test NMSE 0.07810\n",
      "step: 2849 time 0:00:53.495169 train NMSE 0.09337 test NMSE 0.07839\n",
      "step: 2899 time 0:00:56.759629 train NMSE 0.09166 test NMSE 0.07762\n",
      "step: 2949 time 0:00:53.485364 train NMSE 0.09216 test NMSE 0.07504\n",
      "step: 2999 time 0:00:56.546497 train NMSE 0.09115 test NMSE 0.06926\n",
      "Model saved!\n",
      "step: 3049 time 0:00:53.468767 train NMSE 0.09208 test NMSE 0.07097\n",
      "step: 3099 time 0:00:53.534267 train NMSE 0.09439 test NMSE 0.07520\n",
      "step: 3149 time 0:00:53.475386 train NMSE 0.08379 test NMSE 0.07408\n",
      "step: 3199 time 0:00:53.474946 train NMSE 0.08724 test NMSE 0.06387\n",
      "Model saved!\n",
      "step: 3249 time 0:00:53.446727 train NMSE 0.08483 test NMSE 0.06394\n",
      "step: 3299 time 0:00:53.478260 train NMSE 0.08599 test NMSE 0.07580\n",
      "step: 3349 time 0:00:56.544302 train NMSE 0.08108 test NMSE 0.06615\n",
      "step: 3399 time 0:00:53.468396 train NMSE 0.08577 test NMSE 0.07257\n",
      "step: 3449 time 0:00:56.561569 train NMSE 0.08733 test NMSE 0.07158\n",
      "step: 3499 time 0:00:53.478414 train NMSE 0.08745 test NMSE 0.07394\n",
      "step: 3549 time 0:00:56.552179 train NMSE 0.08350 test NMSE 0.06335\n",
      "Model saved!\n",
      "step: 3599 time 0:00:53.476956 train NMSE 0.08710 test NMSE 0.06841\n",
      "step: 3649 time 0:00:56.537472 train NMSE 0.08382 test NMSE 0.06518\n",
      "step: 3699 time 0:00:53.477641 train NMSE 0.08373 test NMSE 0.06731\n",
      "step: 3749 time 0:00:56.532958 train NMSE 0.08263 test NMSE 0.06186\n",
      "Model saved!\n",
      "step: 3799 time 0:00:53.478440 train NMSE 0.08612 test NMSE 0.08078\n",
      "step: 3849 time 0:00:56.523026 train NMSE 0.07995 test NMSE 0.06766\n",
      "step: 3899 time 0:00:53.565641 train NMSE 0.08271 test NMSE 0.07155\n",
      "step: 3949 time 0:00:56.542535 train NMSE 0.07922 test NMSE 0.06214\n",
      "step: 3999 time 0:00:53.485246 train NMSE 0.08022 test NMSE 0.06216\n",
      "step: 4049 time 0:00:56.521487 train NMSE 0.07930 test NMSE 0.06594\n",
      "step: 4099 time 0:00:53.485045 train NMSE 0.07758 test NMSE 0.06688\n",
      "step: 4149 time 0:00:56.550951 train NMSE 0.08040 test NMSE 0.05861\n",
      "Model saved!\n",
      "step: 4199 time 0:00:53.476545 train NMSE 0.07744 test NMSE 0.06372\n",
      "step: 4249 time 0:00:56.530589 train NMSE 0.07817 test NMSE 0.06451\n",
      "step: 4299 time 0:00:53.472515 train NMSE 0.07393 test NMSE 0.06156\n",
      "step: 4349 time 0:00:56.572706 train NMSE 0.07454 test NMSE 0.05739\n",
      "Model saved!\n",
      "step: 4399 time 0:00:53.492383 train NMSE 0.07693 test NMSE 0.05844\n",
      "step: 4449 time 0:00:56.559802 train NMSE 0.07486 test NMSE 0.05651\n",
      "Model saved!\n",
      "step: 4499 time 0:00:56.535885 train NMSE 0.07551 test NMSE 0.06296\n",
      "step: 4549 time 0:00:53.486691 train NMSE 0.07449 test NMSE 0.05486\n",
      "Model saved!\n",
      "step: 4599 time 0:00:56.519841 train NMSE 0.07563 test NMSE 0.06051\n",
      "step: 4649 time 0:00:53.470166 train NMSE 0.07544 test NMSE 0.06354\n",
      "step: 4699 time 0:00:56.533982 train NMSE 0.07604 test NMSE 0.06625\n",
      "step: 4749 time 0:00:53.491188 train NMSE 0.07524 test NMSE 0.05963\n",
      "step: 4799 time 0:00:56.579312 train NMSE 0.07460 test NMSE 0.05377\n",
      "Model saved!\n",
      "step: 4849 time 0:00:53.467583 train NMSE 0.07286 test NMSE 0.05205\n",
      "Model saved!\n",
      "step: 4899 time 0:00:56.560423 train NMSE 0.07564 test NMSE 0.05919\n",
      "step: 4949 time 0:00:53.683628 train NMSE 0.07463 test NMSE 0.05539\n",
      "step: 4999 time 0:00:56.544313 train NMSE 0.07105 test NMSE 0.05602\n",
      "step: 5049 time 0:00:53.482699 train NMSE 0.07387 test NMSE 0.06784\n",
      "step: 5099 time 0:00:56.543155 train NMSE 0.07387 test NMSE 0.06326\n",
      "step: 5149 time 0:00:53.482665 train NMSE 0.07286 test NMSE 0.05240\n",
      "step: 5199 time 0:00:56.561767 train NMSE 0.06979 test NMSE 0.05472\n",
      "step: 5249 time 0:00:53.475723 train NMSE 0.06792 test NMSE 0.05267\n",
      "step: 5299 time 0:00:56.531364 train NMSE 0.07069 test NMSE 0.05007\n",
      "Model saved!\n",
      "step: 5349 time 0:00:53.693121 train NMSE 0.07016 test NMSE 0.05773\n",
      "step: 5399 time 0:00:56.561992 train NMSE 0.06989 test NMSE 0.05825\n",
      "step: 5449 time 0:00:53.706930 train NMSE 0.07111 test NMSE 0.05239\n",
      "step: 5499 time 0:00:56.541416 train NMSE 0.06666 test NMSE 0.05525\n",
      "step: 5549 time 0:00:53.483226 train NMSE 0.07049 test NMSE 0.05448\n",
      "step: 5599 time 0:00:56.555490 train NMSE 0.07177 test NMSE 0.04963\n",
      "Model saved!\n",
      "step: 5649 time 0:00:53.475734 train NMSE 0.07166 test NMSE 0.05243\n",
      "step: 5699 time 0:00:56.507104 train NMSE 0.06836 test NMSE 0.05256\n",
      "step: 5749 time 0:00:53.473220 train NMSE 0.06777 test NMSE 0.05564\n",
      "step: 5799 time 0:00:56.523549 train NMSE 0.06911 test NMSE 0.05251\n",
      "step: 5849 time 0:00:53.564926 train NMSE 0.06664 test NMSE 0.05178\n",
      "step: 5899 time 0:00:56.523646 train NMSE 0.06653 test NMSE 0.04842\n",
      "Model saved!\n",
      "step: 5949 time 0:00:53.488721 train NMSE 0.06676 test NMSE 0.05015\n",
      "step: 5999 time 0:00:56.729594 train NMSE 0.06884 test NMSE 0.05350\n",
      "step: 6049 time 0:00:53.451348 train NMSE 0.06635 test NMSE 0.05896\n",
      "step: 6099 time 0:00:56.542476 train NMSE 0.06688 test NMSE 0.05523\n",
      "step: 6149 time 0:00:53.472739 train NMSE 0.06567 test NMSE 0.04617\n",
      "Model saved!\n",
      "step: 6199 time 0:00:56.526096 train NMSE 0.06575 test NMSE 0.04844\n",
      "step: 6249 time 0:00:53.472245 train NMSE 0.06562 test NMSE 0.04751\n",
      "step: 6299 time 0:00:56.728149 train NMSE 0.06481 test NMSE 0.04931\n",
      "step: 6349 time 0:00:53.548999 train NMSE 0.06542 test NMSE 0.05400\n",
      "step: 6399 time 0:00:53.702587 train NMSE 0.06387 test NMSE 0.05329\n",
      "step: 6449 time 0:00:53.492007 train NMSE 0.06256 test NMSE 0.05985\n",
      "step: 6499 time 0:00:53.503311 train NMSE 0.06563 test NMSE 0.04831\n",
      "step: 6549 time 0:00:56.559554 train NMSE 0.06202 test NMSE 0.05236\n",
      "step: 6599 time 0:00:53.488837 train NMSE 0.06309 test NMSE 0.04898\n",
      "step: 6649 time 0:00:56.522537 train NMSE 0.06481 test NMSE 0.05390\n",
      "step: 6699 time 0:00:53.466754 train NMSE 0.06606 test NMSE 0.04330\n",
      "Model saved!\n",
      "step: 6749 time 0:00:56.532044 train NMSE 0.06418 test NMSE 0.04408\n",
      "step: 6799 time 0:00:53.455100 train NMSE 0.06441 test NMSE 0.04921\n",
      "step: 6849 time 0:00:56.516151 train NMSE 0.06138 test NMSE 0.04941\n",
      "step: 6899 time 0:00:53.478139 train NMSE 0.06207 test NMSE 0.05405\n",
      "step: 6949 time 0:00:56.543990 train NMSE 0.06319 test NMSE 0.04437\n",
      "step: 6999 time 0:00:53.467267 train NMSE 0.06303 test NMSE 0.05177\n",
      "step: 7049 time 0:00:56.508686 train NMSE 0.06228 test NMSE 0.04471\n",
      "step: 7099 time 0:00:53.471336 train NMSE 0.05949 test NMSE 0.04314\n",
      "Model saved!\n",
      "step: 7149 time 0:00:56.623315 train NMSE 0.06069 test NMSE 0.04962\n",
      "step: 7199 time 0:00:53.509102 train NMSE 0.06149 test NMSE 0.04405\n",
      "step: 7249 time 0:00:56.572949 train NMSE 0.06427 test NMSE 0.05054\n",
      "step: 7299 time 0:00:53.482151 train NMSE 0.06204 test NMSE 0.04473\n",
      "step: 7349 time 0:00:56.522657 train NMSE 0.06057 test NMSE 0.04927\n",
      "step: 7399 time 0:00:53.497164 train NMSE 0.06238 test NMSE 0.04984\n",
      "step: 7449 time 0:00:56.509163 train NMSE 0.06283 test NMSE 0.04884\n",
      "step: 7499 time 0:00:53.453989 train NMSE 0.06211 test NMSE 0.04868\n",
      "step: 7549 time 0:00:56.557067 train NMSE 0.06090 test NMSE 0.04757\n",
      "step: 7599 time 0:00:53.502791 train NMSE 0.05876 test NMSE 0.04201\n",
      "Model saved!\n",
      "step: 7649 time 0:00:56.566612 train NMSE 0.05956 test NMSE 0.04389\n",
      "step: 7699 time 0:00:53.470283 train NMSE 0.05967 test NMSE 0.04709\n",
      "step: 7749 time 0:00:56.524517 train NMSE 0.06071 test NMSE 0.03903\n",
      "Model saved!\n",
      "step: 7799 time 0:00:53.454292 train NMSE 0.05801 test NMSE 0.04916\n",
      "step: 7849 time 0:00:56.526073 train NMSE 0.06055 test NMSE 0.04380\n",
      "step: 7899 time 0:00:53.483246 train NMSE 0.06078 test NMSE 0.04593\n",
      "step: 7949 time 0:00:56.556023 train NMSE 0.05957 test NMSE 0.04944\n",
      "step: 7999 time 0:00:53.467276 train NMSE 0.05714 test NMSE 0.04660\n",
      "step: 8049 time 0:00:56.541943 train NMSE 0.05698 test NMSE 0.04516\n",
      "step: 8099 time 0:00:53.551558 train NMSE 0.06116 test NMSE 0.05282\n",
      "step: 8149 time 0:00:56.531755 train NMSE 0.05696 test NMSE 0.04324\n",
      "step: 8199 time 0:00:53.462375 train NMSE 0.05729 test NMSE 0.04354\n",
      "step: 8249 time 0:00:56.739043 train NMSE 0.05731 test NMSE 0.04931\n",
      "step: 8299 time 0:00:53.496155 train NMSE 0.05886 test NMSE 0.04140\n",
      "step: 8349 time 0:00:56.542070 train NMSE 0.05647 test NMSE 0.04429\n",
      "step: 8399 time 0:00:53.468781 train NMSE 0.05644 test NMSE 0.05136\n",
      "step: 8449 time 0:00:56.526136 train NMSE 0.05975 test NMSE 0.04718\n",
      "step: 8499 time 0:00:53.472204 train NMSE 0.05692 test NMSE 0.03961\n",
      "step: 8549 time 0:00:56.744536 train NMSE 0.05737 test NMSE 0.04688\n",
      "step: 8599 time 0:00:53.483982 train NMSE 0.05726 test NMSE 0.04686\n",
      "step: 8649 time 0:00:56.540884 train NMSE 0.05826 test NMSE 0.03962\n",
      "step: 8699 time 0:00:53.667052 train NMSE 0.06013 test NMSE 0.05198\n",
      "step: 8749 time 0:00:56.550240 train NMSE 0.05933 test NMSE 0.04319\n",
      "step: 8799 time 0:00:53.503197 train NMSE 0.05680 test NMSE 0.05106\n",
      "step: 8849 time 0:00:53.475618 train NMSE 0.05696 test NMSE 0.04042\n",
      "step: 8899 time 0:00:53.699784 train NMSE 0.05794 test NMSE 0.04174\n",
      "step: 8949 time 0:00:53.491973 train NMSE 0.06134 test NMSE 0.04184\n",
      "step: 8999 time 0:00:53.469462 train NMSE 0.05765 test NMSE 0.04205\n",
      "step: 9049 time 0:00:53.466178 train NMSE 0.05606 test NMSE 0.04026\n",
      "step: 9099 time 0:00:53.469534 train NMSE 0.05809 test NMSE 0.04150\n",
      "step: 9149 time 0:00:53.490628 train NMSE 0.05729 test NMSE 0.04939\n",
      "step: 9199 time 0:00:53.462855 train NMSE 0.05582 test NMSE 0.04357\n",
      "step: 9249 time 0:00:53.536323 train NMSE 0.05826 test NMSE 0.04690\n",
      "step: 9299 time 0:00:53.481358 train NMSE 0.05723 test NMSE 0.03928\n",
      "step: 9349 time 0:00:53.479091 train NMSE 0.05568 test NMSE 0.04917\n",
      "step: 9399 time 0:00:53.450714 train NMSE 0.05734 test NMSE 0.04868\n",
      "step: 9449 time 0:00:53.459256 train NMSE 0.05762 test NMSE 0.04499\n",
      "step: 9499 time 0:00:53.503512 train NMSE 0.05585 test NMSE 0.03983\n",
      "step: 9549 time 0:00:53.472268 train NMSE 0.05427 test NMSE 0.04294\n",
      "step: 9599 time 0:00:53.477786 train NMSE 0.05521 test NMSE 0.04576\n",
      "step: 9649 time 0:00:53.509243 train NMSE 0.05386 test NMSE 0.04057\n",
      "step: 9699 time 0:00:56.532298 train NMSE 0.05497 test NMSE 0.04111\n",
      "step: 9749 time 0:00:53.467527 train NMSE 0.05450 test NMSE 0.03797\n",
      "Model saved!\n",
      "step: 9799 time 0:00:56.533029 train NMSE 0.05430 test NMSE 0.04126\n",
      "step: 9849 time 0:00:53.466248 train NMSE 0.05418 test NMSE 0.04279\n",
      "step: 9899 time 0:00:56.545231 train NMSE 0.05454 test NMSE 0.04054\n",
      "step: 9949 time 0:00:53.448685 train NMSE 0.05530 test NMSE 0.03791\n",
      "Model saved!\n",
      "step: 9999 time 0:00:56.535209 train NMSE 0.05456 test NMSE 0.04238\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "\n",
    "batch = 64;   #用户数\n",
    "num_batch = 1\n",
    "STEPS = 10000\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "train_csiNet(param_list,num_batch,batch,STEPS,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 49 time 0:00:53.468635 train NMSE 1.02997 test NMSE 0.99813\n",
      "Model saved!\n",
      "step: 99 time 0:00:56.764568 train NMSE 0.88763 test NMSE 0.79268\n",
      "Model saved!\n",
      "step: 149 time 0:00:53.496301 train NMSE 0.69998 test NMSE 0.62065\n",
      "Model saved!\n",
      "step: 199 time 0:00:56.562249 train NMSE 0.56467 test NMSE 0.53541\n",
      "Model saved!\n",
      "step: 249 time 0:00:53.492162 train NMSE 0.47309 test NMSE 0.44884\n",
      "Model saved!\n",
      "step: 299 time 0:00:56.609471 train NMSE 0.39226 test NMSE 0.37055\n",
      "Model saved!\n",
      "step: 349 time 0:00:56.564930 train NMSE 0.31371 test NMSE 0.25140\n",
      "Model saved!\n",
      "step: 399 time 0:00:53.492155 train NMSE 0.24222 test NMSE 0.22009\n",
      "Model saved!\n",
      "step: 449 time 0:00:56.535993 train NMSE 0.19996 test NMSE 0.19062\n",
      "Model saved!\n",
      "step: 499 time 0:00:53.478099 train NMSE 0.17649 test NMSE 0.15450\n",
      "Model saved!\n",
      "step: 549 time 0:00:56.579389 train NMSE 0.15867 test NMSE 0.12606\n",
      "Model saved!\n",
      "step: 599 time 0:00:53.493465 train NMSE 0.15040 test NMSE 0.12812\n",
      "step: 649 time 0:00:56.563969 train NMSE 0.13604 test NMSE 0.12253\n",
      "Model saved!\n",
      "step: 699 time 0:00:56.552502 train NMSE 0.12562 test NMSE 0.10623\n",
      "Model saved!\n",
      "step: 749 time 0:00:53.469656 train NMSE 0.12333 test NMSE 0.12317\n",
      "step: 799 time 0:00:56.544604 train NMSE 0.11240 test NMSE 0.09495\n",
      "Model saved!\n",
      "step: 849 time 0:00:53.501229 train NMSE 0.11115 test NMSE 0.10081\n",
      "step: 899 time 0:00:56.562487 train NMSE 0.10434 test NMSE 0.09619\n",
      "step: 949 time 0:00:53.490792 train NMSE 0.10086 test NMSE 0.09638\n",
      "step: 999 time 0:00:56.597893 train NMSE 0.09475 test NMSE 0.07274\n",
      "Model saved!\n",
      "step: 1049 time 0:00:53.520188 train NMSE 0.09493 test NMSE 0.08000\n",
      "step: 1099 time 0:00:56.557070 train NMSE 0.08675 test NMSE 0.07684\n",
      "step: 1149 time 0:00:53.488228 train NMSE 0.08514 test NMSE 0.06994\n",
      "Model saved!\n",
      "step: 1199 time 0:00:53.493999 train NMSE 0.08884 test NMSE 0.07975\n",
      "step: 1249 time 0:00:56.569995 train NMSE 0.08959 test NMSE 0.07027\n",
      "step: 1299 time 0:00:53.490707 train NMSE 0.08022 test NMSE 0.06825\n",
      "Model saved!\n",
      "step: 1349 time 0:00:56.563573 train NMSE 0.08509 test NMSE 0.07636\n",
      "step: 1399 time 0:00:53.478259 train NMSE 0.07855 test NMSE 0.06673\n",
      "Model saved!\n",
      "step: 1449 time 0:00:56.785006 train NMSE 0.07614 test NMSE 0.06420\n",
      "Model saved!\n",
      "step: 1499 time 0:00:53.494776 train NMSE 0.07746 test NMSE 0.06439\n",
      "step: 1549 time 0:00:56.581789 train NMSE 0.07328 test NMSE 0.05295\n",
      "Model saved!\n",
      "step: 1599 time 0:00:53.706114 train NMSE 0.07254 test NMSE 0.05954\n",
      "step: 1649 time 0:00:53.501203 train NMSE 0.06871 test NMSE 0.05363\n",
      "step: 1699 time 0:00:56.595078 train NMSE 0.06616 test NMSE 0.05654\n",
      "step: 1749 time 0:00:53.486147 train NMSE 0.06883 test NMSE 0.05484\n",
      "step: 1799 time 0:00:56.527023 train NMSE 0.06663 test NMSE 0.05570\n",
      "step: 1849 time 0:00:53.491675 train NMSE 0.06674 test NMSE 0.06834\n",
      "step: 1899 time 0:00:56.557018 train NMSE 0.06752 test NMSE 0.05493\n",
      "step: 1949 time 0:00:53.539119 train NMSE 0.06341 test NMSE 0.05736\n",
      "step: 1999 time 0:00:56.770542 train NMSE 0.06400 test NMSE 0.06572\n",
      "step: 2049 time 0:00:53.589898 train NMSE 0.06569 test NMSE 0.04822\n",
      "Model saved!\n",
      "step: 2099 time 0:00:56.791368 train NMSE 0.06281 test NMSE 0.05431\n",
      "step: 2149 time 0:00:53.503665 train NMSE 0.06163 test NMSE 0.04620\n",
      "Model saved!\n",
      "step: 2199 time 0:00:56.547500 train NMSE 0.05993 test NMSE 0.05224\n",
      "step: 2249 time 0:00:53.513122 train NMSE 0.06000 test NMSE 0.04934\n",
      "step: 2299 time 0:00:53.497054 train NMSE 0.05685 test NMSE 0.04768\n",
      "step: 2349 time 0:00:56.613814 train NMSE 0.05787 test NMSE 0.05292\n",
      "step: 2399 time 0:00:53.489179 train NMSE 0.06062 test NMSE 0.05524\n",
      "step: 2449 time 0:00:56.677325 train NMSE 0.05964 test NMSE 0.05294\n",
      "step: 2499 time 0:00:53.510031 train NMSE 0.05702 test NMSE 0.04558\n",
      "Model saved!\n",
      "step: 2549 time 0:00:56.576467 train NMSE 0.05686 test NMSE 0.04553\n",
      "Model saved!\n",
      "step: 2599 time 0:00:53.524524 train NMSE 0.05628 test NMSE 0.05083\n",
      "step: 2649 time 0:00:56.651691 train NMSE 0.05383 test NMSE 0.04118\n",
      "Model saved!\n",
      "step: 2699 time 0:00:53.520560 train NMSE 0.05434 test NMSE 0.04067\n",
      "Model saved!\n",
      "step: 2749 time 0:00:53.501076 train NMSE 0.05392 test NMSE 0.04076\n",
      "step: 2799 time 0:00:53.493032 train NMSE 0.05360 test NMSE 0.04058\n",
      "Model saved!\n",
      "step: 2849 time 0:00:53.493107 train NMSE 0.05283 test NMSE 0.04372\n",
      "step: 2899 time 0:00:56.520239 train NMSE 0.05197 test NMSE 0.04058\n",
      "Model saved!\n",
      "step: 2949 time 0:00:53.719001 train NMSE 0.05168 test NMSE 0.04281\n",
      "step: 2999 time 0:00:56.577843 train NMSE 0.05513 test NMSE 0.05067\n",
      "step: 3049 time 0:00:53.487978 train NMSE 0.05387 test NMSE 0.04640\n",
      "step: 3099 time 0:00:56.572302 train NMSE 0.05254 test NMSE 0.04116\n",
      "step: 3149 time 0:00:53.475371 train NMSE 0.04992 test NMSE 0.03647\n",
      "Model saved!\n",
      "step: 3199 time 0:00:56.557000 train NMSE 0.05031 test NMSE 0.04242\n",
      "step: 3249 time 0:00:53.506988 train NMSE 0.05058 test NMSE 0.04343\n",
      "step: 3299 time 0:00:56.561985 train NMSE 0.05068 test NMSE 0.03976\n",
      "step: 3349 time 0:00:53.503671 train NMSE 0.04923 test NMSE 0.04259\n",
      "step: 3399 time 0:00:56.799736 train NMSE 0.04972 test NMSE 0.04177\n",
      "step: 3449 time 0:00:56.571632 train NMSE 0.04859 test NMSE 0.03443\n",
      "Model saved!\n",
      "step: 3499 time 0:00:53.506517 train NMSE 0.05046 test NMSE 0.03936\n",
      "step: 3549 time 0:00:56.547553 train NMSE 0.04922 test NMSE 0.03666\n",
      "step: 3599 time 0:00:53.499084 train NMSE 0.04911 test NMSE 0.03706\n",
      "step: 3649 time 0:00:56.556227 train NMSE 0.04556 test NMSE 0.03967\n",
      "step: 3699 time 0:00:53.504627 train NMSE 0.04705 test NMSE 0.03695\n",
      "step: 3749 time 0:00:56.573370 train NMSE 0.04409 test NMSE 0.04217\n",
      "step: 3799 time 0:00:53.544933 train NMSE 0.04612 test NMSE 0.03115\n",
      "Model saved!\n",
      "step: 3849 time 0:00:56.557538 train NMSE 0.04483 test NMSE 0.03394\n",
      "step: 3899 time 0:00:53.494994 train NMSE 0.04547 test NMSE 0.03427\n",
      "step: 3949 time 0:00:56.574158 train NMSE 0.04482 test NMSE 0.04071\n",
      "step: 3999 time 0:00:53.490500 train NMSE 0.04440 test NMSE 0.04082\n",
      "step: 4049 time 0:00:53.482962 train NMSE 0.04546 test NMSE 0.03189\n",
      "step: 4099 time 0:00:56.546430 train NMSE 0.04459 test NMSE 0.03968\n",
      "step: 4149 time 0:00:53.497992 train NMSE 0.04637 test NMSE 0.03616\n",
      "step: 4199 time 0:00:56.558047 train NMSE 0.04865 test NMSE 0.03784\n",
      "step: 4249 time 0:00:53.494153 train NMSE 0.04551 test NMSE 0.03657\n",
      "step: 4299 time 0:00:56.547754 train NMSE 0.04340 test NMSE 0.03357\n",
      "step: 4349 time 0:00:53.516956 train NMSE 0.04377 test NMSE 0.03707\n",
      "step: 4399 time 0:00:56.581264 train NMSE 0.04327 test NMSE 0.03388\n",
      "step: 4449 time 0:00:53.497388 train NMSE 0.04403 test NMSE 0.04900\n",
      "step: 4499 time 0:00:56.648119 train NMSE 0.04357 test NMSE 0.03503\n",
      "step: 4549 time 0:00:53.716732 train NMSE 0.04252 test NMSE 0.03850\n",
      "step: 4599 time 0:00:56.556656 train NMSE 0.04379 test NMSE 0.03338\n",
      "step: 4649 time 0:00:53.576776 train NMSE 0.04241 test NMSE 0.03140\n",
      "step: 4699 time 0:00:56.553367 train NMSE 0.04337 test NMSE 0.03066\n",
      "Model saved!\n",
      "step: 4749 time 0:00:56.588945 train NMSE 0.04235 test NMSE 0.03060\n",
      "Model saved!\n",
      "step: 4799 time 0:00:53.503085 train NMSE 0.04334 test NMSE 0.04121\n",
      "step: 4849 time 0:00:56.600819 train NMSE 0.04495 test NMSE 0.03519\n",
      "step: 4899 time 0:00:53.494222 train NMSE 0.04475 test NMSE 0.03352\n",
      "step: 4949 time 0:00:56.584098 train NMSE 0.04007 test NMSE 0.03400\n",
      "step: 4999 time 0:00:53.495216 train NMSE 0.04273 test NMSE 0.04345\n",
      "step: 5049 time 0:00:56.595852 train NMSE 0.04326 test NMSE 0.03117\n",
      "step: 5099 time 0:00:53.516143 train NMSE 0.04246 test NMSE 0.03160\n",
      "step: 5149 time 0:00:56.627966 train NMSE 0.04088 test NMSE 0.02917\n",
      "Model saved!\n",
      "step: 5199 time 0:00:53.489073 train NMSE 0.04102 test NMSE 0.03364\n",
      "step: 5249 time 0:00:53.513149 train NMSE 0.04066 test NMSE 0.03423\n",
      "step: 5299 time 0:00:56.562876 train NMSE 0.04110 test NMSE 0.03657\n",
      "step: 5349 time 0:00:53.530161 train NMSE 0.04000 test NMSE 0.03014\n",
      "step: 5399 time 0:00:56.763580 train NMSE 0.04130 test NMSE 0.03269\n",
      "step: 5449 time 0:00:53.509129 train NMSE 0.04313 test NMSE 0.03583\n",
      "step: 5499 time 0:00:56.588306 train NMSE 0.04149 test NMSE 0.02987\n",
      "step: 5549 time 0:00:53.515127 train NMSE 0.03825 test NMSE 0.03498\n",
      "step: 5599 time 0:00:56.547968 train NMSE 0.03953 test NMSE 0.03130\n",
      "step: 5649 time 0:00:53.489174 train NMSE 0.04032 test NMSE 0.03151\n",
      "step: 5699 time 0:00:56.589014 train NMSE 0.03867 test NMSE 0.03124\n",
      "step: 5749 time 0:00:53.537038 train NMSE 0.03919 test NMSE 0.03383\n",
      "step: 5799 time 0:00:56.567535 train NMSE 0.03953 test NMSE 0.02950\n",
      "step: 5849 time 0:00:53.503154 train NMSE 0.03882 test NMSE 0.04083\n",
      "step: 5899 time 0:00:56.792298 train NMSE 0.04099 test NMSE 0.02869\n",
      "Model saved!\n",
      "step: 5949 time 0:00:56.588389 train NMSE 0.03779 test NMSE 0.02597\n",
      "Model saved!\n",
      "step: 5999 time 0:00:53.500653 train NMSE 0.04065 test NMSE 0.03513\n",
      "step: 6049 time 0:00:56.799208 train NMSE 0.04043 test NMSE 0.03813\n",
      "step: 6099 time 0:00:53.537475 train NMSE 0.03862 test NMSE 0.02840\n",
      "step: 6149 time 0:00:56.596904 train NMSE 0.03548 test NMSE 0.02793\n",
      "step: 6199 time 0:00:53.551266 train NMSE 0.03748 test NMSE 0.03130\n",
      "step: 6249 time 0:00:56.581642 train NMSE 0.03837 test NMSE 0.03164\n",
      "step: 6299 time 0:00:53.515190 train NMSE 0.03900 test NMSE 0.02513\n",
      "Model saved!\n",
      "step: 6349 time 0:00:53.525009 train NMSE 0.03763 test NMSE 0.03179\n",
      "step: 6399 time 0:00:56.566753 train NMSE 0.03778 test NMSE 0.03029\n",
      "step: 6449 time 0:00:53.531005 train NMSE 0.03777 test NMSE 0.02847\n",
      "step: 6499 time 0:00:56.781327 train NMSE 0.03686 test NMSE 0.02887\n",
      "step: 6549 time 0:00:53.534904 train NMSE 0.03693 test NMSE 0.02844\n",
      "step: 6599 time 0:00:56.557613 train NMSE 0.03909 test NMSE 0.03252\n",
      "step: 6649 time 0:00:53.512968 train NMSE 0.03862 test NMSE 0.03274\n",
      "step: 6699 time 0:00:56.599903 train NMSE 0.03747 test NMSE 0.02770\n",
      "step: 6749 time 0:00:53.515071 train NMSE 0.03851 test NMSE 0.03549\n",
      "step: 6799 time 0:00:56.569884 train NMSE 0.03609 test NMSE 0.03388\n",
      "step: 6849 time 0:00:53.525052 train NMSE 0.03483 test NMSE 0.02903\n",
      "step: 6899 time 0:00:56.601804 train NMSE 0.03630 test NMSE 0.02686\n",
      "step: 6949 time 0:00:53.522665 train NMSE 0.03469 test NMSE 0.02832\n",
      "step: 6999 time 0:00:53.528253 train NMSE 0.03856 test NMSE 0.02942\n",
      "step: 7049 time 0:00:56.566132 train NMSE 0.03622 test NMSE 0.03088\n",
      "step: 7099 time 0:00:53.521582 train NMSE 0.03595 test NMSE 0.02463\n",
      "Model saved!\n",
      "step: 7149 time 0:00:56.570352 train NMSE 0.03663 test NMSE 0.02784\n",
      "step: 7199 time 0:00:53.517040 train NMSE 0.03424 test NMSE 0.02885\n",
      "step: 7249 time 0:00:56.569419 train NMSE 0.03582 test NMSE 0.02496\n",
      "step: 7299 time 0:00:53.552471 train NMSE 0.03649 test NMSE 0.02775\n",
      "step: 7349 time 0:00:56.788278 train NMSE 0.03611 test NMSE 0.02730\n",
      "step: 7399 time 0:00:53.712519 train NMSE 0.03484 test NMSE 0.02757\n",
      "step: 7449 time 0:00:56.576962 train NMSE 0.03496 test NMSE 0.03548\n",
      "step: 7499 time 0:00:53.500428 train NMSE 0.03582 test NMSE 0.02501\n",
      "step: 7549 time 0:00:56.837141 train NMSE 0.03523 test NMSE 0.02898\n",
      "step: 7599 time 0:00:53.718031 train NMSE 0.03725 test NMSE 0.02758\n",
      "step: 7649 time 0:00:56.770270 train NMSE 0.03538 test NMSE 0.02762\n",
      "step: 7699 time 0:00:53.600224 train NMSE 0.03569 test NMSE 0.02736\n",
      "step: 7749 time 0:00:56.580040 train NMSE 0.03415 test NMSE 0.02636\n",
      "step: 7799 time 0:00:56.562937 train NMSE 0.03499 test NMSE 0.03031\n",
      "step: 7849 time 0:00:53.553365 train NMSE 0.03615 test NMSE 0.02708\n",
      "step: 7899 time 0:00:56.589392 train NMSE 0.03358 test NMSE 0.02634\n",
      "step: 7949 time 0:00:53.521768 train NMSE 0.03366 test NMSE 0.02773\n",
      "step: 7999 time 0:00:56.580966 train NMSE 0.03377 test NMSE 0.02657\n",
      "step: 8049 time 0:00:53.514644 train NMSE 0.03435 test NMSE 0.03034\n",
      "step: 8099 time 0:00:56.547886 train NMSE 0.03558 test NMSE 0.02695\n",
      "step: 8149 time 0:00:53.525613 train NMSE 0.03350 test NMSE 0.02426\n",
      "Model saved!\n",
      "step: 8199 time 0:00:56.580523 train NMSE 0.03265 test NMSE 0.02748\n",
      "step: 8249 time 0:00:53.502352 train NMSE 0.03356 test NMSE 0.02796\n",
      "step: 8299 time 0:00:56.771073 train NMSE 0.03449 test NMSE 0.03059\n",
      "step: 8349 time 0:00:53.520043 train NMSE 0.03491 test NMSE 0.02504\n",
      "step: 8399 time 0:00:56.551864 train NMSE 0.03257 test NMSE 0.02609\n",
      "step: 8449 time 0:00:56.573836 train NMSE 0.03369 test NMSE 0.02587\n",
      "step: 8499 time 0:00:53.502678 train NMSE 0.03381 test NMSE 0.02399\n",
      "Model saved!\n",
      "step: 8549 time 0:00:56.552024 train NMSE 0.03427 test NMSE 0.02702\n",
      "step: 8599 time 0:00:53.501409 train NMSE 0.03368 test NMSE 0.02486\n",
      "step: 8649 time 0:00:56.532742 train NMSE 0.03461 test NMSE 0.02867\n",
      "step: 8699 time 0:00:53.484567 train NMSE 0.03352 test NMSE 0.02510\n",
      "step: 8749 time 0:00:56.549011 train NMSE 0.03371 test NMSE 0.02525\n",
      "step: 8799 time 0:00:53.524847 train NMSE 0.03436 test NMSE 0.02725\n",
      "step: 8849 time 0:00:56.567274 train NMSE 0.03192 test NMSE 0.02704\n",
      "step: 8899 time 0:00:53.535690 train NMSE 0.03243 test NMSE 0.02477\n",
      "step: 8949 time 0:00:56.548245 train NMSE 0.03256 test NMSE 0.02811\n",
      "step: 8999 time 0:00:53.480433 train NMSE 0.03240 test NMSE 0.02967\n",
      "step: 9049 time 0:00:56.553650 train NMSE 0.03316 test NMSE 0.02838\n",
      "step: 9099 time 0:00:56.570996 train NMSE 0.03319 test NMSE 0.02623\n",
      "step: 9149 time 0:00:53.587598 train NMSE 0.03240 test NMSE 0.02341\n",
      "Model saved!\n",
      "step: 9199 time 0:00:56.578560 train NMSE 0.03264 test NMSE 0.02480\n",
      "step: 9249 time 0:00:53.530616 train NMSE 0.03192 test NMSE 0.02576\n",
      "step: 9299 time 0:00:56.568068 train NMSE 0.03200 test NMSE 0.02481\n",
      "step: 9349 time 0:00:53.492547 train NMSE 0.03263 test NMSE 0.02229\n",
      "Model saved!\n",
      "step: 9399 time 0:00:56.550593 train NMSE 0.03370 test NMSE 0.02849\n",
      "step: 9449 time 0:00:53.722657 train NMSE 0.03240 test NMSE 0.02370\n",
      "step: 9499 time 0:00:56.579728 train NMSE 0.03162 test NMSE 0.02580\n",
      "step: 9549 time 0:00:53.503638 train NMSE 0.03286 test NMSE 0.02732\n",
      "step: 9599 time 0:00:56.571650 train NMSE 0.03216 test NMSE 0.02635\n",
      "step: 9649 time 0:00:53.546495 train NMSE 0.03095 test NMSE 0.02470\n",
      "step: 9699 time 0:00:53.513132 train NMSE 0.03027 test NMSE 0.02339\n",
      "step: 9749 time 0:00:56.552110 train NMSE 0.03066 test NMSE 0.02176\n",
      "Model saved!\n",
      "step: 9799 time 0:00:53.499845 train NMSE 0.03175 test NMSE 0.02508\n",
      "step: 9849 time 0:00:56.549469 train NMSE 0.03293 test NMSE 0.02557\n",
      "step: 9899 time 0:00:53.516266 train NMSE 0.03283 test NMSE 0.02734\n",
      "step: 9949 time 0:00:56.812754 train NMSE 0.03302 test NMSE 0.02289\n",
      "step: 9999 time 0:00:53.509935 train NMSE 0.03092 test NMSE 0.02378\n"
     ]
    }
   ],
   "source": [
    "B = 128\n",
    "\n",
    "batch = 64;   #用户数\n",
    "num_batch = 1\n",
    "STEPS = 10000\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "train_csiNet(param_list,num_batch,batch,STEPS,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 49 time 0:00:54.384621 train NMSE 1.02029 test NMSE 0.95031\n",
      "Model saved!\n",
      "step: 99 time 0:00:50.025462 train NMSE 0.78403 test NMSE 0.73160\n",
      "Model saved!\n",
      "step: 149 time 0:00:47.488076 train NMSE 0.56877 test NMSE 0.43711\n",
      "Model saved!\n",
      "step: 199 time 0:00:47.428806 train NMSE 0.36677 test NMSE 0.29545\n",
      "Model saved!\n",
      "step: 249 time 0:00:47.436331 train NMSE 0.25146 test NMSE 0.19214\n",
      "Model saved!\n",
      "step: 299 time 0:00:50.056980 train NMSE 0.17974 test NMSE 0.15279\n",
      "Model saved!\n",
      "step: 349 time 0:00:50.077803 train NMSE 0.15339 test NMSE 0.12467\n",
      "Model saved!\n",
      "step: 399 time 0:00:50.080745 train NMSE 0.13883 test NMSE 0.11672\n",
      "Model saved!\n",
      "step: 449 time 0:00:50.077856 train NMSE 0.12314 test NMSE 0.09937\n",
      "Model saved!\n",
      "step: 499 time 0:00:50.129297 train NMSE 0.11426 test NMSE 0.10259\n",
      "step: 549 time 0:00:50.093748 train NMSE 0.10902 test NMSE 0.09393\n",
      "Model saved!\n",
      "step: 599 time 0:00:50.138650 train NMSE 0.09991 test NMSE 0.08748\n",
      "Model saved!\n",
      "step: 649 time 0:00:47.505510 train NMSE 0.09861 test NMSE 0.08283\n",
      "Model saved!\n",
      "step: 699 time 0:00:47.512711 train NMSE 0.09229 test NMSE 0.08401\n",
      "step: 749 time 0:00:50.128035 train NMSE 0.08515 test NMSE 0.06605\n",
      "Model saved!\n",
      "step: 799 time 0:00:50.106084 train NMSE 0.07980 test NMSE 0.06340\n",
      "Model saved!\n",
      "step: 849 time 0:00:47.602111 train NMSE 0.07880 test NMSE 0.06985\n",
      "step: 899 time 0:00:50.112759 train NMSE 0.07595 test NMSE 0.05895\n",
      "Model saved!\n",
      "step: 949 time 0:00:50.143342 train NMSE 0.06991 test NMSE 0.05818\n",
      "Model saved!\n",
      "step: 999 time 0:00:47.528276 train NMSE 0.06743 test NMSE 0.06064\n",
      "step: 1049 time 0:00:50.138439 train NMSE 0.06557 test NMSE 0.05093\n",
      "Model saved!\n",
      "step: 1099 time 0:00:50.287835 train NMSE 0.06189 test NMSE 0.05297\n",
      "step: 1149 time 0:00:47.509725 train NMSE 0.06320 test NMSE 0.05228\n",
      "step: 1199 time 0:00:50.124336 train NMSE 0.06064 test NMSE 0.05130\n",
      "step: 1249 time 0:00:42.345536 train NMSE 0.05669 test NMSE 0.04590\n",
      "Model saved!\n",
      "step: 1299 time 0:00:14.021522 train NMSE 0.05524 test NMSE 0.04399\n",
      "Model saved!\n",
      "step: 1349 time 0:00:14.273844 train NMSE 0.05434 test NMSE 0.04157\n",
      "Model saved!\n",
      "step: 1399 time 0:00:14.320259 train NMSE 0.05434 test NMSE 0.04007\n",
      "Model saved!\n",
      "step: 1449 time 0:00:14.340678 train NMSE 0.05114 test NMSE 0.04236\n",
      "step: 1499 time 0:00:14.343748 train NMSE 0.05450 test NMSE 0.04640\n",
      "step: 1549 time 0:00:14.355195 train NMSE 0.05065 test NMSE 0.04089\n",
      "step: 1599 time 0:00:14.357650 train NMSE 0.04627 test NMSE 0.03842\n",
      "Model saved!\n",
      "step: 1649 time 0:00:14.360646 train NMSE 0.04683 test NMSE 0.03576\n",
      "Model saved!\n",
      "step: 1699 time 0:00:14.350177 train NMSE 0.04931 test NMSE 0.03692\n",
      "step: 1749 time 0:00:14.367110 train NMSE 0.04590 test NMSE 0.03676\n",
      "step: 1799 time 0:00:14.355682 train NMSE 0.04654 test NMSE 0.03897\n",
      "step: 1849 time 0:00:14.357166 train NMSE 0.04672 test NMSE 0.03707\n",
      "step: 1899 time 0:00:14.359654 train NMSE 0.04534 test NMSE 0.03357\n",
      "Model saved!\n",
      "step: 1949 time 0:00:14.340697 train NMSE 0.04377 test NMSE 0.03966\n",
      "step: 1999 time 0:00:14.350677 train NMSE 0.04656 test NMSE 0.04628\n",
      "step: 2049 time 0:00:14.345698 train NMSE 0.04259 test NMSE 0.03175\n",
      "Model saved!\n",
      "step: 2099 time 0:00:14.343696 train NMSE 0.04029 test NMSE 0.03283\n",
      "step: 2149 time 0:00:14.365667 train NMSE 0.04087 test NMSE 0.03288\n",
      "step: 2199 time 0:00:14.436851 train NMSE 0.04089 test NMSE 0.03425\n",
      "step: 2249 time 0:00:14.425144 train NMSE 0.03996 test NMSE 0.03179\n",
      "step: 2299 time 0:00:14.371214 train NMSE 0.03933 test NMSE 0.02638\n",
      "Model saved!\n",
      "step: 2349 time 0:00:14.393672 train NMSE 0.03865 test NMSE 0.03385\n",
      "step: 2399 time 0:00:34.935041 train NMSE 0.03930 test NMSE 0.02561\n",
      "Model saved!\n",
      "step: 2449 time 0:00:50.058613 train NMSE 0.04006 test NMSE 0.03259\n",
      "step: 2499 time 0:00:50.114570 train NMSE 0.03811 test NMSE 0.02990\n",
      "step: 2549 time 0:00:50.072218 train NMSE 0.03594 test NMSE 0.03074\n",
      "step: 2599 time 0:00:47.446867 train NMSE 0.03584 test NMSE 0.02886\n",
      "step: 2649 time 0:00:50.042552 train NMSE 0.03654 test NMSE 0.02726\n",
      "step: 2699 time 0:00:50.069537 train NMSE 0.03643 test NMSE 0.03031\n",
      "step: 2749 time 0:00:50.151041 train NMSE 0.03613 test NMSE 0.02723\n",
      "step: 2799 time 0:00:47.516184 train NMSE 0.03459 test NMSE 0.02395\n",
      "Model saved!\n",
      "step: 2849 time 0:00:50.224856 train NMSE 0.03444 test NMSE 0.02720\n",
      "step: 2899 time 0:00:50.104261 train NMSE 0.03348 test NMSE 0.03189\n",
      "step: 2949 time 0:00:47.474866 train NMSE 0.03371 test NMSE 0.02336\n",
      "Model saved!\n",
      "step: 2999 time 0:00:50.094494 train NMSE 0.03239 test NMSE 0.02475\n",
      "step: 3049 time 0:00:50.081870 train NMSE 0.03422 test NMSE 0.02961\n",
      "step: 3099 time 0:00:47.645757 train NMSE 0.03454 test NMSE 0.02273\n",
      "Model saved!\n",
      "step: 3149 time 0:00:50.087178 train NMSE 0.03284 test NMSE 0.02341\n",
      "step: 3199 time 0:00:50.095824 train NMSE 0.03242 test NMSE 0.02590\n",
      "step: 3249 time 0:00:47.490514 train NMSE 0.03348 test NMSE 0.02533\n",
      "step: 3299 time 0:00:50.092798 train NMSE 0.03128 test NMSE 0.02856\n",
      "step: 3349 time 0:00:50.058251 train NMSE 0.03184 test NMSE 0.02373\n",
      "step: 3399 time 0:00:47.497631 train NMSE 0.03063 test NMSE 0.02336\n",
      "step: 3449 time 0:00:50.087212 train NMSE 0.03006 test NMSE 0.02174\n",
      "Model saved!\n",
      "step: 3499 time 0:00:47.476005 train NMSE 0.03055 test NMSE 0.02153\n",
      "Model saved!\n",
      "step: 3549 time 0:00:47.497263 train NMSE 0.03109 test NMSE 0.02393\n",
      "step: 3599 time 0:00:50.088542 train NMSE 0.03062 test NMSE 0.02207\n",
      "step: 3649 time 0:00:47.634804 train NMSE 0.03058 test NMSE 0.02603\n",
      "step: 3699 time 0:00:50.091616 train NMSE 0.03009 test NMSE 0.02274\n",
      "step: 3749 time 0:00:50.110178 train NMSE 0.02912 test NMSE 0.02206\n",
      "step: 3799 time 0:00:47.508958 train NMSE 0.02861 test NMSE 0.02399\n",
      "step: 3849 time 0:00:50.064410 train NMSE 0.03070 test NMSE 0.02340\n",
      "step: 3899 time 0:00:47.467830 train NMSE 0.02933 test NMSE 0.03030\n",
      "step: 3949 time 0:00:47.483064 train NMSE 0.02884 test NMSE 0.02196\n",
      "step: 3999 time 0:00:50.091816 train NMSE 0.02937 test NMSE 0.02116\n",
      "Model saved!\n",
      "step: 4049 time 0:00:47.516053 train NMSE 0.02931 test NMSE 0.02155\n",
      "step: 4099 time 0:00:50.058399 train NMSE 0.02731 test NMSE 0.01995\n",
      "Model saved!\n",
      "step: 4149 time 0:00:50.098756 train NMSE 0.02793 test NMSE 0.02030\n",
      "step: 4199 time 0:00:47.469192 train NMSE 0.02771 test NMSE 0.02030\n",
      "step: 4249 time 0:00:50.063934 train NMSE 0.02698 test NMSE 0.02004\n",
      "step: 4299 time 0:00:50.082890 train NMSE 0.02633 test NMSE 0.01986\n",
      "Model saved!\n",
      "step: 4349 time 0:00:47.452275 train NMSE 0.02798 test NMSE 0.02147\n",
      "step: 4399 time 0:00:50.066238 train NMSE 0.02785 test NMSE 0.02045\n",
      "step: 4449 time 0:00:50.079761 train NMSE 0.02674 test NMSE 0.01819\n",
      "Model saved!\n",
      "step: 4499 time 0:00:47.498207 train NMSE 0.02726 test NMSE 0.02339\n",
      "step: 4549 time 0:00:50.078767 train NMSE 0.02850 test NMSE 0.02274\n",
      "step: 4599 time 0:00:47.435375 train NMSE 0.02729 test NMSE 0.02159\n",
      "step: 4649 time 0:00:50.075250 train NMSE 0.02771 test NMSE 0.02186\n",
      "step: 4699 time 0:00:50.051303 train NMSE 0.02672 test NMSE 0.01901\n",
      "step: 4749 time 0:00:47.506904 train NMSE 0.02599 test NMSE 0.02020\n",
      "step: 4799 time 0:00:50.236847 train NMSE 0.02634 test NMSE 0.01866\n",
      "step: 4849 time 0:00:50.231846 train NMSE 0.02583 test NMSE 0.02196\n",
      "step: 4899 time 0:00:47.476904 train NMSE 0.02588 test NMSE 0.01857\n",
      "step: 4949 time 0:00:50.044325 train NMSE 0.02694 test NMSE 0.02036\n",
      "step: 4999 time 0:00:47.456223 train NMSE 0.02676 test NMSE 0.02168\n",
      "step: 5049 time 0:00:50.050032 train NMSE 0.02634 test NMSE 0.02085\n",
      "step: 5099 time 0:00:50.054828 train NMSE 0.02555 test NMSE 0.01980\n",
      "step: 5149 time 0:00:47.701079 train NMSE 0.02570 test NMSE 0.01979\n",
      "step: 5199 time 0:00:50.091860 train NMSE 0.02508 test NMSE 0.01754\n",
      "Model saved!\n",
      "step: 5249 time 0:00:50.090610 train NMSE 0.02455 test NMSE 0.02111\n",
      "step: 5299 time 0:00:47.477322 train NMSE 0.02396 test NMSE 0.01851\n",
      "step: 5349 time 0:00:50.085771 train NMSE 0.02450 test NMSE 0.02136\n",
      "step: 5399 time 0:00:47.452325 train NMSE 0.02449 test NMSE 0.01682\n",
      "Model saved!\n",
      "step: 5449 time 0:00:47.471985 train NMSE 0.02504 test NMSE 0.01920\n",
      "step: 5499 time 0:00:50.081155 train NMSE 0.02606 test NMSE 0.01799\n",
      "step: 5549 time 0:00:47.456011 train NMSE 0.02353 test NMSE 0.01826\n",
      "step: 5599 time 0:00:50.077262 train NMSE 0.02433 test NMSE 0.01881\n",
      "step: 5649 time 0:00:50.048428 train NMSE 0.02395 test NMSE 0.02056\n",
      "step: 5699 time 0:00:47.532969 train NMSE 0.02345 test NMSE 0.01859\n",
      "step: 5749 time 0:00:50.077340 train NMSE 0.02379 test NMSE 0.01698\n",
      "step: 5799 time 0:00:50.077746 train NMSE 0.02345 test NMSE 0.01770\n",
      "step: 5849 time 0:00:47.476328 train NMSE 0.02423 test NMSE 0.01914\n",
      "step: 5899 time 0:00:50.080707 train NMSE 0.02331 test NMSE 0.01795\n",
      "step: 5949 time 0:00:47.499143 train NMSE 0.02344 test NMSE 0.01853\n",
      "step: 5999 time 0:00:50.077837 train NMSE 0.02492 test NMSE 0.01865\n",
      "step: 6049 time 0:00:50.056828 train NMSE 0.02418 test NMSE 0.01857\n",
      "step: 6099 time 0:00:47.476735 train NMSE 0.02386 test NMSE 0.01940\n",
      "step: 6149 time 0:00:50.240409 train NMSE 0.02267 test NMSE 0.01911\n",
      "step: 6199 time 0:00:50.067489 train NMSE 0.02270 test NMSE 0.01566\n",
      "Model saved!\n",
      "step: 6249 time 0:00:47.475324 train NMSE 0.02276 test NMSE 0.01742\n",
      "step: 6299 time 0:00:50.080064 train NMSE 0.02335 test NMSE 0.01620\n",
      "step: 6349 time 0:00:50.075850 train NMSE 0.02191 test NMSE 0.02081\n",
      "step: 6399 time 0:00:47.651721 train NMSE 0.02229 test NMSE 0.01868\n",
      "step: 6449 time 0:00:50.080797 train NMSE 0.02479 test NMSE 0.01828\n",
      "step: 6499 time 0:00:47.470859 train NMSE 0.02309 test NMSE 0.01767\n",
      "step: 6549 time 0:00:50.249639 train NMSE 0.02171 test NMSE 0.01607\n",
      "step: 6599 time 0:00:50.063868 train NMSE 0.02267 test NMSE 0.01656\n",
      "step: 6649 time 0:00:47.497099 train NMSE 0.02183 test NMSE 0.01821\n",
      "step: 6699 time 0:00:50.076777 train NMSE 0.02321 test NMSE 0.01613\n",
      "step: 6749 time 0:00:50.140278 train NMSE 0.02175 test NMSE 0.01819\n",
      "step: 6799 time 0:00:47.477947 train NMSE 0.02222 test NMSE 0.01632\n",
      "step: 6849 time 0:00:50.049937 train NMSE 0.02139 test NMSE 0.01713\n",
      "step: 6899 time 0:00:47.476286 train NMSE 0.02219 test NMSE 0.01642\n",
      "step: 6949 time 0:00:50.086363 train NMSE 0.02198 test NMSE 0.01692\n",
      "step: 6999 time 0:00:50.051845 train NMSE 0.02376 test NMSE 0.01793\n",
      "step: 7049 time 0:00:47.484531 train NMSE 0.02259 test NMSE 0.01566\n",
      "Model saved!\n",
      "step: 7099 time 0:00:50.120139 train NMSE 0.02226 test NMSE 0.01695\n",
      "step: 7149 time 0:00:50.062428 train NMSE 0.02171 test NMSE 0.01738\n",
      "step: 7199 time 0:00:47.484443 train NMSE 0.02156 test NMSE 0.01785\n",
      "step: 7249 time 0:00:50.222815 train NMSE 0.02214 test NMSE 0.01617\n",
      "step: 7299 time 0:00:47.418957 train NMSE 0.02267 test NMSE 0.01573\n",
      "step: 7349 time 0:00:50.184271 train NMSE 0.02017 test NMSE 0.01543\n",
      "Model saved!\n",
      "step: 7399 time 0:00:50.057797 train NMSE 0.02068 test NMSE 0.01601\n",
      "step: 7449 time 0:00:47.500068 train NMSE 0.02071 test NMSE 0.01674\n",
      "step: 7499 time 0:00:50.063393 train NMSE 0.02072 test NMSE 0.01582\n",
      "step: 7549 time 0:00:47.443766 train NMSE 0.02100 test NMSE 0.01969\n",
      "step: 7599 time 0:00:47.473806 train NMSE 0.02198 test NMSE 0.01576\n",
      "step: 7649 time 0:00:49.989964 train NMSE 0.02159 test NMSE 0.01696\n",
      "step: 7699 time 0:00:47.373811 train NMSE 0.02121 test NMSE 0.01704\n",
      "step: 7749 time 0:00:49.997529 train NMSE 0.02105 test NMSE 0.01424\n",
      "Model saved!\n",
      "step: 7799 time 0:00:47.484781 train NMSE 0.02081 test NMSE 0.01446\n",
      "step: 7849 time 0:00:47.462375 train NMSE 0.02092 test NMSE 0.01800\n",
      "step: 7899 time 0:00:50.238240 train NMSE 0.02103 test NMSE 0.01540\n",
      "step: 7949 time 0:00:47.609445 train NMSE 0.02140 test NMSE 0.01634\n",
      "step: 7999 time 0:00:50.049487 train NMSE 0.02154 test NMSE 0.01627\n",
      "step: 8049 time 0:00:50.127194 train NMSE 0.02023 test NMSE 0.01461\n",
      "step: 8099 time 0:00:47.493238 train NMSE 0.02120 test NMSE 0.01547\n",
      "step: 8149 time 0:00:50.053045 train NMSE 0.02016 test NMSE 0.01563\n",
      "step: 8199 time 0:00:47.483802 train NMSE 0.02078 test NMSE 0.01618\n",
      "step: 8249 time 0:00:50.052766 train NMSE 0.01831 test NMSE 0.01550\n",
      "step: 8299 time 0:00:50.214597 train NMSE 0.02101 test NMSE 0.01652\n",
      "step: 8349 time 0:00:47.413898 train NMSE 0.01961 test NMSE 0.01507\n",
      "step: 8399 time 0:00:50.174590 train NMSE 0.02054 test NMSE 0.01452\n",
      "step: 8449 time 0:00:47.479323 train NMSE 0.02036 test NMSE 0.01416\n",
      "Model saved!\n",
      "step: 8499 time 0:00:47.455379 train NMSE 0.01920 test NMSE 0.01430\n",
      "step: 8549 time 0:00:50.178829 train NMSE 0.01912 test NMSE 0.01660\n",
      "step: 8599 time 0:00:47.462310 train NMSE 0.02052 test NMSE 0.01544\n",
      "step: 8649 time 0:00:50.068393 train NMSE 0.01989 test NMSE 0.01365\n",
      "Model saved!\n",
      "step: 8699 time 0:00:50.047960 train NMSE 0.02059 test NMSE 0.01624\n",
      "step: 8749 time 0:00:47.492740 train NMSE 0.02029 test NMSE 0.01364\n",
      "Model saved!\n",
      "step: 8799 time 0:00:50.044293 train NMSE 0.01982 test NMSE 0.01487\n",
      "step: 8849 time 0:00:50.034315 train NMSE 0.01970 test NMSE 0.01442\n",
      "step: 8899 time 0:00:47.453222 train NMSE 0.02052 test NMSE 0.01402\n",
      "step: 8949 time 0:00:50.003463 train NMSE 0.01966 test NMSE 0.01517\n",
      "step: 8999 time 0:00:47.456580 train NMSE 0.01954 test NMSE 0.01553\n",
      "step: 9049 time 0:00:50.059812 train NMSE 0.01895 test NMSE 0.01542\n",
      "step: 9099 time 0:00:50.069330 train NMSE 0.01926 test NMSE 0.01550\n",
      "step: 9149 time 0:00:47.472794 train NMSE 0.01963 test NMSE 0.01378\n",
      "step: 9199 time 0:00:50.027748 train NMSE 0.01921 test NMSE 0.01386\n",
      "step: 9249 time 0:00:47.466144 train NMSE 0.01990 test NMSE 0.01981\n",
      "step: 9299 time 0:00:47.622375 train NMSE 0.02031 test NMSE 0.01528\n",
      "step: 9349 time 0:00:50.040163 train NMSE 0.01936 test NMSE 0.01566\n",
      "step: 9399 time 0:00:47.421810 train NMSE 0.02028 test NMSE 0.01297\n",
      "Model saved!\n",
      "step: 9449 time 0:00:50.076569 train NMSE 0.02024 test NMSE 0.01832\n",
      "step: 9499 time 0:00:50.052866 train NMSE 0.02062 test NMSE 0.01390\n",
      "step: 9549 time 0:00:47.459494 train NMSE 0.01814 test NMSE 0.01410\n",
      "step: 9599 time 0:00:50.047548 train NMSE 0.01861 test NMSE 0.01450\n",
      "step: 9649 time 0:00:47.621445 train NMSE 0.01987 test NMSE 0.01418\n",
      "step: 9699 time 0:00:50.074965 train NMSE 0.01824 test NMSE 0.01384\n",
      "step: 9749 time 0:00:50.071320 train NMSE 0.01917 test NMSE 0.01480\n",
      "step: 9799 time 0:00:47.630032 train NMSE 0.01896 test NMSE 0.01476\n",
      "step: 9849 time 0:00:50.040716 train NMSE 0.01890 test NMSE 0.01463\n",
      "step: 9899 time 0:00:47.624132 train NMSE 0.01826 test NMSE 0.01591\n",
      "step: 9949 time 0:00:47.505268 train NMSE 0.01852 test NMSE 0.01423\n",
      "step: 9999 time 0:00:50.065337 train NMSE 0.01932 test NMSE 0.01466\n"
     ]
    }
   ],
   "source": [
    "B = 256\n",
    "\n",
    "batch = 64;   #用户数\n",
    "num_batch = 1\n",
    "STEPS = 10000\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "train_csiNet(param_list,num_batch,batch,STEPS,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10000000000.0"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "BW"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test csiNet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04008\n",
      "SOMP道估计+完美反馈 NMSE 0.99900\n",
      "BSOMP道估计+完美反馈 NMSE 0.99871\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 1.09190\n",
      "BSOMP道估计+csiNet反馈 NMSE 1.09106\n",
      "AMP道估计+csiNet反馈 NMSE 1.08704\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 1\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04227\n",
      "SOMP道估计+完美反馈 NMSE 0.99971\n",
      "BSOMP道估计+完美反馈 NMSE 0.99041\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 1.09132\n",
      "BSOMP道估计+csiNet反馈 NMSE 1.07413\n",
      "AMP道估计+csiNet反馈 NMSE 1.09231\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 2\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04107\n",
      "SOMP道估计+完美反馈 NMSE 1.00000\n",
      "BSOMP道估计+完美反馈 NMSE 0.98601\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 1.09377\n",
      "BSOMP道估计+csiNet反馈 NMSE 1.05731\n",
      "AMP道估计+csiNet反馈 NMSE 1.08874\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 4\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04044\n",
      "SOMP道估计+完美反馈 NMSE 0.95751\n",
      "BSOMP道估计+完美反馈 NMSE 0.45456\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 1.05400\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.35872\n",
      "AMP道估计+csiNet反馈 NMSE 1.08972\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 8\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04553\n",
      "SOMP道估计+完美反馈 NMSE 0.31151\n",
      "BSOMP道估计+完美反馈 NMSE 0.13828\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.32229\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.11716\n",
      "AMP道估计+csiNet反馈 NMSE 1.08931\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04085\n",
      "SOMP道估计+完美反馈 NMSE 0.11240\n",
      "BSOMP道估计+完美反馈 NMSE 0.08451\n",
      "AMP道估计+完美反馈 NMSE 0.13510\n",
      "SOMP道估计+csiNet反馈 NMSE 0.09713\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.08445\n",
      "AMP道估计+csiNet反馈 NMSE 0.15815\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 32\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04205\n",
      "SOMP道估计+完美反馈 NMSE 0.07626\n",
      "BSOMP道估计+完美反馈 NMSE 0.06469\n",
      "AMP道估计+完美反馈 NMSE 0.01724\n",
      "SOMP道估计+csiNet反馈 NMSE 0.07618\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.07355\n",
      "AMP道估计+csiNet反馈 NMSE 0.04480\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 64\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.72169\n",
      "SOMP道估计+完美反馈 NMSE 0.32255\n",
      "BSOMP道估计+完美反馈 NMSE 0.12935\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.83182\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.74119\n",
      "AMP道估计+csiNet反馈 NMSE 1.02660\n"
     ]
    }
   ],
   "source": [
    "B = 8\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.36793\n",
      "SOMP道估计+完美反馈 NMSE 0.33611\n",
      "BSOMP道估计+完美反馈 NMSE 0.12922\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.65856\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.41504\n",
      "AMP道估计+csiNet反馈 NMSE 1.19024\n"
     ]
    }
   ],
   "source": [
    "B = 16\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09593\n",
      "SOMP道估计+完美反馈 NMSE 0.34198\n",
      "BSOMP道估计+完美反馈 NMSE 0.12862\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.47044\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.17412\n",
      "AMP道估计+csiNet反馈 NMSE 1.24075\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.03944\n",
      "SOMP道估计+完美反馈 NMSE 0.31139\n",
      "BSOMP道估计+完美反馈 NMSE 0.13136\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.31320\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.11795\n",
      "AMP道估计+csiNet反馈 NMSE 1.08776\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.02304\n",
      "SOMP道估计+完美反馈 NMSE 0.33141\n",
      "BSOMP道估计+完美反馈 NMSE 0.14613\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.26410\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.08320\n",
      "AMP道估计+csiNet反馈 NMSE 1.05489\n"
     ]
    }
   ],
   "source": [
    "B = 128\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.01356\n",
      "SOMP道估计+完美反馈 NMSE 0.34043\n",
      "BSOMP道估计+完美反馈 NMSE 0.12739\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.24218\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.05778\n",
      "AMP道估计+csiNet反馈 NMSE 1.01043\n"
     ]
    }
   ],
   "source": [
    "B = 256\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09625\n",
      "SOMP道估计+完美反馈 NMSE 0.30849\n",
      "BSOMP道估计+完美反馈 NMSE 0.12572\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.48722\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.17967\n",
      "AMP道估计+csiNet反馈 NMSE 1.22755\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e1\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09202\n",
      "SOMP道估计+完美反馈 NMSE 0.31364\n",
      "BSOMP道估计+完美反馈 NMSE 0.13797\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.43396\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.18291\n",
      "AMP道估计+csiNet反馈 NMSE 1.24841\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09132\n",
      "SOMP道估计+完美反馈 NMSE 0.39242\n",
      "BSOMP道估计+完美反馈 NMSE 0.17413\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.53521\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.18207\n",
      "AMP道估计+csiNet反馈 NMSE 1.23704\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e3\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09930\n",
      "SOMP道估计+完美反馈 NMSE 0.87246\n",
      "BSOMP道估计+完美反馈 NMSE 0.57901\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 0.95569\n",
      "BSOMP道估计+csiNet反馈 NMSE 0.24812\n",
      "AMP道估计+csiNet反馈 NMSE 1.23209\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e4\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09002\n",
      "SOMP道估计+完美反馈 NMSE 1.00000\n",
      "BSOMP道估计+完美反馈 NMSE 1.00000\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 1.22897\n",
      "BSOMP道估计+csiNet反馈 NMSE 1.22611\n",
      "AMP道估计+csiNet反馈 NMSE 1.24133\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e5\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09521\n",
      "SOMP道估计+完美反馈 NMSE 1.00000\n",
      "BSOMP道估计+完美反馈 NMSE 1.00000\n",
      "AMP道估计+完美反馈 NMSE 1.00000\n",
      "SOMP道估计+csiNet反馈 NMSE 1.24329\n",
      "BSOMP道估计+csiNet反馈 NMSE 1.23227\n",
      "AMP道估计+csiNet反馈 NMSE 1.23144\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e6\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_channelAcquisition(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### GMMV-LAMP + csiNet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04097\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.97291\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 1.09622\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 1\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.03944\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.96958\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 1.09345\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 2\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04249\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.75001\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.91940\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 4\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.03869\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.21550\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.13425\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 8\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04075\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.11457\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.07496\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04108\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.03315\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.04758\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 32\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04124\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.01669\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.04282\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 64\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.71706\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.12258\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.73390\n"
     ]
    }
   ],
   "source": [
    "B = 8\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.37059\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.12474\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.41655\n"
     ]
    }
   ],
   "source": [
    "B = 16\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09177\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.12206\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.15328\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.04209\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.12145\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.07362\n"
     ]
    }
   ],
   "source": [
    "B = 64\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.02335\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.12628\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.04772\n"
     ]
    }
   ],
   "source": [
    "B = 128\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.01413\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.12149\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.03579\n"
     ]
    }
   ],
   "source": [
    "B = 256\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09706\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.06252\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.12598\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e1\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.10048\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.11971\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.14442\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e2\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09730\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.18183\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.16330\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e3\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09158\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.46768\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 0.41230\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e4\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.09650\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.87797\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 1.03291\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e5\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "完美信道估计+csiNet反馈 NMSE 0.10033\n",
      "GMMV-LAMP道估计+完美反馈 NMSE 0.99690\n",
      "GMMV-LAMP道估计+csiNet反馈 NMSE 1.24200\n"
     ]
    }
   ],
   "source": [
    "B = 32\n",
    "L = 16\n",
    "sigma = 1e6\n",
    "batch = 256;   #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "model_name_CF = './models/RIS_CF_'+str(B)+'bits_'+str(Lp)+'paths'+'.pth'\n",
    "test_CF_MDDL(param_list,batch,model_name_CF)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### GMMV-LAMP 训练"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "pho = 2 #过采样倍数\n",
    "N_2 = pho*N \n",
    "\n",
    "\n",
    "A = torch.zeros(N*N,N_2*N_2).cuda() + 0j\n",
    "A_t = torch.zeros(M_ant,N_2*N_2*M*M).cuda() + 0j\n",
    "for i in range(N_2):\n",
    "    for j in range(N_2):\n",
    "        a_h = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*i*2*pi\n",
    "        a_v = torch.range(0,N-1).reshape(1,N).cuda()/N/pho*j*2*pi\n",
    "        a = kron_add(a_h,a_v).reshape(N*N)\n",
    "        A[:,i*N_2+j] = torch.exp(1j*a)\n",
    "for i in range(M*M):\n",
    "    A_t[i*N*N:(i+1)*N*N,i*N_2*N_2:(i+1)*N_2*N_2] = A"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:02.037092 train NMSE 10.65716 test NMSE 8.66925\n",
      "step: 199 time 0:00:01.773785 train NMSE 7.54252 test NMSE 6.44187\n",
      "step: 299 time 0:00:01.773853 train NMSE 5.37646 test NMSE 4.50775\n",
      "step: 399 time 0:00:01.786821 train NMSE 3.93455 test NMSE 3.61277\n",
      "step: 499 time 0:00:01.788264 train NMSE 2.99628 test NMSE 2.55816\n",
      "step: 599 time 0:00:01.778394 train NMSE 2.37531 test NMSE 2.16657\n",
      "step: 699 time 0:00:01.802039 train NMSE 1.96395 test NMSE 1.73962\n",
      "Model saved!\n",
      "step: 799 time 0:00:01.763833 train NMSE 1.67828 test NMSE 1.54369\n",
      "Model saved!\n",
      "step: 899 time 0:00:01.773304 train NMSE 1.50649 test NMSE 1.46771\n",
      "Model saved!\n",
      "step: 999 time 0:00:01.797696 train NMSE 1.37474 test NMSE 1.31922\n",
      "Model saved!\n",
      "step: 1099 time 0:00:01.791713 train NMSE 1.28245 test NMSE 1.27431\n",
      "Model saved!\n",
      "step: 1199 time 0:00:01.766779 train NMSE 1.21408 test NMSE 1.21348\n",
      "Model saved!\n",
      "step: 1299 time 0:00:01.778253 train NMSE 1.16072 test NMSE 1.15776\n",
      "Model saved!\n",
      "step: 1399 time 0:00:01.772258 train NMSE 1.11995 test NMSE 1.10206\n",
      "Model saved!\n",
      "step: 1499 time 0:00:01.762309 train NMSE 1.09169 test NMSE 1.08818\n",
      "Model saved!\n",
      "step: 1599 time 0:00:01.770768 train NMSE 1.07136 test NMSE 1.05061\n",
      "Model saved!\n",
      "step: 1699 time 0:00:01.791443 train NMSE 1.05132 test NMSE 1.03128\n",
      "Model saved!\n",
      "step: 1799 time 0:00:01.771272 train NMSE 1.04062 test NMSE 1.02264\n",
      "Model saved!\n",
      "step: 1899 time 0:00:01.766668 train NMSE 1.03056 test NMSE 1.01651\n",
      "Model saved!\n",
      "step: 1999 time 0:00:01.761578 train NMSE 1.02224 test NMSE 1.01849\n",
      "step: 2099 time 0:00:01.799197 train NMSE 1.01515 test NMSE 1.01252\n",
      "Model saved!\n",
      "step: 2199 time 0:00:01.825143 train NMSE 1.01136 test NMSE 1.00699\n",
      "Model saved!\n",
      "step: 2299 time 0:00:01.796708 train NMSE 1.00627 test NMSE 1.00139\n",
      "Model saved!\n",
      "step: 2399 time 0:00:01.772268 train NMSE 1.00365 test NMSE 1.00532\n",
      "step: 2499 time 0:00:01.783264 train NMSE 1.00039 test NMSE 0.99943\n",
      "Model saved!\n",
      "step: 2599 time 0:00:01.800701 train NMSE 0.99896 test NMSE 0.99650\n",
      "Model saved!\n",
      "step: 2699 time 0:00:01.809768 train NMSE 0.99649 test NMSE 0.99561\n",
      "Model saved!\n",
      "step: 2799 time 0:00:01.815662 train NMSE 0.99525 test NMSE 0.99263\n",
      "Model saved!\n",
      "step: 2899 time 0:00:01.840633 train NMSE 0.99465 test NMSE 0.99235\n",
      "Model saved!\n",
      "step: 2999 time 0:00:01.815195 train NMSE 0.99346 test NMSE 0.99527\n",
      "step: 99 time 0:00:02.069021 train NMSE 1.10698 test NMSE 1.07263\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.049544 train NMSE 1.04924 test NMSE 1.03245\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.053582 train NMSE 1.02425 test NMSE 1.01379\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.125335 train NMSE 1.01013 test NMSE 1.00599\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.053380 train NMSE 1.00207 test NMSE 1.00061\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.032291 train NMSE 0.99665 test NMSE 0.99338\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.056540 train NMSE 0.99303 test NMSE 0.99227\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.039584 train NMSE 0.98959 test NMSE 0.98716\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.053648 train NMSE 0.98725 test NMSE 0.98800\n",
      "step: 999 time 0:00:02.047176 train NMSE 0.98540 test NMSE 0.98611\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.044569 train NMSE 0.98379 test NMSE 0.98174\n",
      "Model saved!\n",
      "step: 1199 time 0:00:02.043224 train NMSE 0.98239 test NMSE 0.98033\n",
      "Model saved!\n",
      "step: 1299 time 0:00:02.049569 train NMSE 0.98125 test NMSE 0.98145\n",
      "step: 1399 time 0:00:02.055025 train NMSE 0.98025 test NMSE 0.97636\n",
      "Model saved!\n",
      "step: 1499 time 0:00:02.072855 train NMSE 0.97963 test NMSE 0.98028\n",
      "step: 1599 time 0:00:02.046543 train NMSE 0.97851 test NMSE 0.97906\n",
      "step: 1699 time 0:00:02.053467 train NMSE 0.97796 test NMSE 0.97479\n",
      "Model saved!\n",
      "step: 1799 time 0:00:02.088455 train NMSE 0.97734 test NMSE 0.97619\n",
      "step: 1899 time 0:00:02.098859 train NMSE 0.97628 test NMSE 0.98091\n",
      "step: 1999 time 0:00:02.094452 train NMSE 0.97583 test NMSE 0.97555\n",
      "step: 2099 time 0:00:02.130197 train NMSE 0.97518 test NMSE 0.97352\n",
      "Model saved!\n",
      "step: 2199 time 0:00:02.089938 train NMSE 0.97446 test NMSE 0.97598\n",
      "step: 2299 time 0:00:02.107905 train NMSE 0.97436 test NMSE 0.97561\n",
      "step: 2399 time 0:00:02.075512 train NMSE 0.97362 test NMSE 0.96828\n",
      "Model saved!\n",
      "step: 2499 time 0:00:02.071002 train NMSE 0.97316 test NMSE 0.97108\n",
      "step: 2599 time 0:00:02.063536 train NMSE 0.97268 test NMSE 0.96620\n",
      "Model saved!\n",
      "step: 2699 time 0:00:02.065888 train NMSE 0.97325 test NMSE 0.96950\n",
      "step: 2799 time 0:00:02.088449 train NMSE 0.97139 test NMSE 0.97261\n",
      "step: 2899 time 0:00:02.073752 train NMSE 0.97199 test NMSE 0.96542\n",
      "Model saved!\n",
      "step: 2999 time 0:00:02.056021 train NMSE 0.97071 test NMSE 0.97110\n",
      "step: 99 time 0:00:02.338743 train NMSE 1.01622 test NMSE 1.00861\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.343183 train NMSE 1.00349 test NMSE 1.00096\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.343498 train NMSE 0.99829 test NMSE 0.99490\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.364768 train NMSE 0.99430 test NMSE 0.99302\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.343802 train NMSE 0.99192 test NMSE 0.98983\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.341801 train NMSE 0.99047 test NMSE 0.98912\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.458326 train NMSE 0.98887 test NMSE 0.98904\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.320830 train NMSE 0.98764 test NMSE 0.98493\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.302839 train NMSE 0.98623 test NMSE 0.98553\n",
      "step: 999 time 0:00:02.287880 train NMSE 0.98567 test NMSE 0.98454\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.301843 train NMSE 0.98509 test NMSE 0.98372\n",
      "Model saved!\n",
      "step: 1199 time 0:00:02.320790 train NMSE 0.98421 test NMSE 0.98149\n",
      "Model saved!\n",
      "step: 1299 time 0:00:02.315804 train NMSE 0.98372 test NMSE 0.98194\n",
      "step: 1399 time 0:00:02.295858 train NMSE 0.98316 test NMSE 0.98055\n",
      "Model saved!\n",
      "step: 1499 time 0:00:02.312813 train NMSE 0.98293 test NMSE 0.98175\n",
      "step: 1599 time 0:00:02.304835 train NMSE 0.98217 test NMSE 0.98060\n",
      "step: 1699 time 0:00:02.304833 train NMSE 0.98234 test NMSE 0.98086\n",
      "step: 1799 time 0:00:02.295859 train NMSE 0.98120 test NMSE 0.97938\n",
      "Model saved!\n",
      "step: 1899 time 0:00:02.310817 train NMSE 0.98129 test NMSE 0.97905\n",
      "Model saved!\n",
      "step: 1999 time 0:00:02.307826 train NMSE 0.98071 test NMSE 0.97844\n",
      "Model saved!\n",
      "step: 2099 time 0:00:02.313849 train NMSE 0.98037 test NMSE 0.98032\n",
      "step: 2199 time 0:00:02.300891 train NMSE 0.97982 test NMSE 0.97907\n",
      "step: 2299 time 0:00:02.294886 train NMSE 0.97986 test NMSE 0.97912\n",
      "step: 2399 time 0:00:02.309821 train NMSE 0.97920 test NMSE 0.97876\n",
      "step: 2499 time 0:00:02.307841 train NMSE 0.97925 test NMSE 0.97940\n",
      "step: 2599 time 0:00:02.304881 train NMSE 0.97842 test NMSE 0.97597\n",
      "Model saved!\n",
      "step: 2699 time 0:00:02.314840 train NMSE 0.97879 test NMSE 0.97748\n",
      "step: 2799 time 0:00:02.321803 train NMSE 0.97840 test NMSE 0.97838\n",
      "step: 2899 time 0:00:02.311814 train NMSE 0.97745 test NMSE 0.97700\n",
      "step: 2999 time 0:00:02.317811 train NMSE 0.97734 test NMSE 0.97842\n",
      "step: 99 time 0:00:02.588118 train NMSE 0.99841 test NMSE 0.99587\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.586106 train NMSE 0.99467 test NMSE 0.99376\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.590072 train NMSE 0.99267 test NMSE 0.99172\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.590071 train NMSE 0.99103 test NMSE 0.99043\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.595083 train NMSE 0.98939 test NMSE 0.98873\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.591080 train NMSE 0.98750 test NMSE 0.98778\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.593062 train NMSE 0.98574 test NMSE 0.98555\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.587111 train NMSE 0.98398 test NMSE 0.98109\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.585128 train NMSE 0.98213 test NMSE 0.98139\n",
      "step: 999 time 0:00:02.573145 train NMSE 0.98077 test NMSE 0.97864\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.585098 train NMSE 0.97938 test NMSE 0.98036\n",
      "step: 1199 time 0:00:02.586118 train NMSE 0.97856 test NMSE 0.97623\n",
      "Model saved!\n",
      "step: 1299 time 0:00:02.578159 train NMSE 0.97780 test NMSE 0.97652\n",
      "step: 1399 time 0:00:02.583103 train NMSE 0.97716 test NMSE 0.98042\n",
      "step: 1499 time 0:00:02.569127 train NMSE 0.97629 test NMSE 0.97752\n",
      "step: 1599 time 0:00:02.579125 train NMSE 0.97610 test NMSE 0.97628\n",
      "step: 1699 time 0:00:02.645361 train NMSE 0.97556 test NMSE 0.97523\n",
      "Model saved!\n",
      "step: 1799 time 0:00:02.599090 train NMSE 0.97500 test NMSE 0.97320\n",
      "Model saved!\n",
      "step: 1899 time 0:00:02.579099 train NMSE 0.97455 test NMSE 0.97316\n",
      "Model saved!\n",
      "step: 1999 time 0:00:02.589114 train NMSE 0.97426 test NMSE 0.97169\n",
      "Model saved!\n",
      "step: 2099 time 0:00:02.591067 train NMSE 0.97363 test NMSE 0.97420\n",
      "step: 2199 time 0:00:02.576123 train NMSE 0.97397 test NMSE 0.97178\n",
      "step: 2299 time 0:00:02.581120 train NMSE 0.97342 test NMSE 0.97395\n",
      "step: 2399 time 0:00:02.593623 train NMSE 0.97359 test NMSE 0.97549\n",
      "step: 2499 time 0:00:02.590070 train NMSE 0.97231 test NMSE 0.97599\n",
      "step: 2599 time 0:00:02.587599 train NMSE 0.97234 test NMSE 0.97703\n",
      "step: 2699 time 0:00:02.603076 train NMSE 0.97236 test NMSE 0.97359\n",
      "step: 2799 time 0:00:02.591130 train NMSE 0.97152 test NMSE 0.97591\n",
      "step: 2899 time 0:00:02.589073 train NMSE 0.97138 test NMSE 0.97008\n",
      "Model saved!\n",
      "step: 2999 time 0:00:02.596079 train NMSE 0.97126 test NMSE 0.97230\n",
      "step: 99 time 0:00:02.868382 train NMSE 0.99889 test NMSE 0.99678\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.885322 train NMSE 0.99500 test NMSE 0.99345\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.845388 train NMSE 0.99279 test NMSE 0.99187\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.848455 train NMSE 0.99117 test NMSE 0.99039\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.865079 train NMSE 0.98962 test NMSE 0.98923\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.874195 train NMSE 0.98806 test NMSE 0.98614\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.876305 train NMSE 0.98640 test NMSE 0.98569\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.884548 train NMSE 0.98459 test NMSE 0.98279\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.885642 train NMSE 0.98322 test NMSE 0.98450\n",
      "step: 999 time 0:00:02.861863 train NMSE 0.98184 test NMSE 0.98010\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.885705 train NMSE 0.98085 test NMSE 0.98158\n",
      "step: 1199 time 0:00:02.869337 train NMSE 0.97968 test NMSE 0.97824\n",
      "Model saved!\n",
      "step: 1299 time 0:00:02.870758 train NMSE 0.97906 test NMSE 0.97965\n",
      "step: 1399 time 0:00:02.875814 train NMSE 0.97796 test NMSE 0.97730\n",
      "Model saved!\n",
      "step: 1499 time 0:00:02.875308 train NMSE 0.97756 test NMSE 0.97578\n",
      "Model saved!\n",
      "step: 1599 time 0:00:02.894297 train NMSE 0.97743 test NMSE 0.97712\n",
      "step: 1699 time 0:00:02.887996 train NMSE 0.97672 test NMSE 0.97244\n",
      "Model saved!\n",
      "step: 1799 time 0:00:02.892262 train NMSE 0.97627 test NMSE 0.97794\n",
      "step: 1899 time 0:00:02.885805 train NMSE 0.97621 test NMSE 0.97620\n",
      "step: 1999 time 0:00:02.863619 train NMSE 0.97556 test NMSE 0.97864\n",
      "step: 2099 time 0:00:02.871468 train NMSE 0.97540 test NMSE 0.97655\n",
      "step: 2199 time 0:00:02.877303 train NMSE 0.97527 test NMSE 0.97401\n",
      "step: 2299 time 0:00:02.905734 train NMSE 0.97500 test NMSE 0.97430\n",
      "step: 2399 time 0:00:02.931766 train NMSE 0.97474 test NMSE 0.97287\n",
      "step: 2499 time 0:00:02.866343 train NMSE 0.97479 test NMSE 0.97547\n",
      "step: 2599 time 0:00:02.880318 train NMSE 0.97413 test NMSE 0.97397\n",
      "step: 2699 time 0:00:02.888301 train NMSE 0.97408 test NMSE 0.97388\n",
      "step: 2799 time 0:00:02.879471 train NMSE 0.97407 test NMSE 0.97392\n",
      "step: 2899 time 0:00:02.868327 train NMSE 0.97403 test NMSE 0.97789\n",
      "step: 2999 time 0:00:02.891277 train NMSE 0.97367 test NMSE 0.97205\n",
      "Model saved!\n"
     ]
    }
   ],
   "source": [
    "L  = 1  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:01.987217 train NMSE 20.48805 test NMSE 14.95176\n",
      "step: 199 time 0:00:01.963768 train NMSE 15.27663 test NMSE 13.55973\n",
      "step: 299 time 0:00:01.962750 train NMSE 11.36668 test NMSE 9.97800\n",
      "step: 399 time 0:00:01.956786 train NMSE 8.74011 test NMSE 7.42560\n",
      "step: 499 time 0:00:01.972304 train NMSE 6.83870 test NMSE 6.46052\n",
      "step: 599 time 0:00:01.958979 train NMSE 5.35346 test NMSE 4.97472\n",
      "step: 699 time 0:00:01.968665 train NMSE 4.21605 test NMSE 3.40709\n",
      "step: 799 time 0:00:01.953773 train NMSE 3.39833 test NMSE 3.23069\n",
      "step: 899 time 0:00:01.964744 train NMSE 2.79392 test NMSE 2.43245\n",
      "step: 999 time 0:00:01.976220 train NMSE 2.28228 test NMSE 2.12983\n",
      "step: 1099 time 0:00:01.952070 train NMSE 1.90396 test NMSE 1.69563\n",
      "Model saved!\n",
      "step: 1199 time 0:00:01.958785 train NMSE 1.61969 test NMSE 1.51776\n",
      "Model saved!\n",
      "step: 1299 time 0:00:01.965955 train NMSE 1.42097 test NMSE 1.39534\n",
      "Model saved!\n",
      "step: 1399 time 0:00:01.947789 train NMSE 1.26503 test NMSE 1.18319\n",
      "Model saved!\n",
      "step: 1499 time 0:00:01.966749 train NMSE 1.16316 test NMSE 1.10916\n",
      "Model saved!\n",
      "step: 1599 time 0:00:01.980219 train NMSE 1.08698 test NMSE 1.05001\n",
      "Model saved!\n",
      "step: 1699 time 0:00:01.962074 train NMSE 1.04272 test NMSE 1.02513\n",
      "Model saved!\n",
      "step: 1799 time 0:00:01.954954 train NMSE 1.01590 test NMSE 1.01193\n",
      "Model saved!\n",
      "step: 1899 time 0:00:01.966739 train NMSE 1.00040 test NMSE 0.99599\n",
      "Model saved!\n",
      "step: 1999 time 0:00:01.966738 train NMSE 0.99100 test NMSE 0.99268\n",
      "Model saved!\n",
      "step: 2099 time 0:00:01.974717 train NMSE 0.98623 test NMSE 0.98357\n",
      "Model saved!\n",
      "step: 2199 time 0:00:01.962402 train NMSE 0.98198 test NMSE 0.98219\n",
      "Model saved!\n",
      "step: 2299 time 0:00:01.968541 train NMSE 0.98029 test NMSE 0.97480\n",
      "Model saved!\n",
      "step: 2399 time 0:00:01.965740 train NMSE 0.97750 test NMSE 0.97626\n",
      "step: 2499 time 0:00:01.978707 train NMSE 0.97670 test NMSE 0.97113\n",
      "Model saved!\n",
      "step: 2599 time 0:00:01.950588 train NMSE 0.97533 test NMSE 0.97280\n",
      "step: 2699 time 0:00:01.965874 train NMSE 0.97402 test NMSE 0.97535\n",
      "step: 2799 time 0:00:01.970545 train NMSE 0.97276 test NMSE 0.97763\n",
      "step: 2899 time 0:00:01.956712 train NMSE 0.97201 test NMSE 0.97292\n",
      "step: 2999 time 0:00:01.970728 train NMSE 0.97077 test NMSE 0.97270\n",
      "step: 99 time 0:00:02.256962 train NMSE 4.91168 test NMSE 3.80655\n",
      "step: 199 time 0:00:02.265054 train NMSE 3.44247 test NMSE 2.52565\n",
      "step: 299 time 0:00:02.253771 train NMSE 2.63265 test NMSE 2.37727\n",
      "step: 399 time 0:00:02.258957 train NMSE 2.08613 test NMSE 1.92870\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.256962 train NMSE 1.76390 test NMSE 1.53396\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.245658 train NMSE 1.52940 test NMSE 1.38500\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.274115 train NMSE 1.36452 test NMSE 1.23229\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.263508 train NMSE 1.26139 test NMSE 1.16001\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.259954 train NMSE 1.17770 test NMSE 1.13644\n",
      "Model saved!\n",
      "step: 999 time 0:00:02.262947 train NMSE 1.11744 test NMSE 0.99355\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.254981 train NMSE 1.07384 test NMSE 1.00672\n",
      "step: 1199 time 0:00:02.246506 train NMSE 1.04164 test NMSE 1.02483\n",
      "step: 1299 time 0:00:02.274682 train NMSE 1.01375 test NMSE 0.99021\n",
      "Model saved!\n",
      "step: 1399 time 0:00:02.247986 train NMSE 0.99721 test NMSE 0.99525\n",
      "step: 1499 time 0:00:02.271430 train NMSE 0.98493 test NMSE 0.98058\n",
      "Model saved!\n",
      "step: 1599 time 0:00:02.239037 train NMSE 0.97311 test NMSE 0.96073\n",
      "Model saved!\n",
      "step: 1699 time 0:00:02.242538 train NMSE 0.96303 test NMSE 0.96551\n",
      "step: 1799 time 0:00:02.245008 train NMSE 0.95771 test NMSE 0.96189\n",
      "step: 1899 time 0:00:02.244994 train NMSE 0.95274 test NMSE 0.95002\n",
      "Model saved!\n",
      "step: 1999 time 0:00:02.264512 train NMSE 0.94765 test NMSE 0.93884\n",
      "Model saved!\n",
      "step: 2099 time 0:00:02.246523 train NMSE 0.94570 test NMSE 0.94258\n",
      "step: 2199 time 0:00:02.297895 train NMSE 0.94255 test NMSE 0.93551\n",
      "Model saved!\n",
      "step: 2299 time 0:00:02.246988 train NMSE 0.93959 test NMSE 0.94413\n",
      "step: 2399 time 0:00:02.249981 train NMSE 0.93775 test NMSE 0.95096\n",
      "step: 2499 time 0:00:02.267961 train NMSE 0.93535 test NMSE 0.93062\n",
      "Model saved!\n",
      "step: 2599 time 0:00:02.251032 train NMSE 0.93497 test NMSE 0.93061\n",
      "Model saved!\n",
      "step: 2699 time 0:00:02.253499 train NMSE 0.93104 test NMSE 0.92931\n",
      "Model saved!\n",
      "step: 2799 time 0:00:02.251975 train NMSE 0.93134 test NMSE 0.92839\n",
      "Model saved!\n",
      "step: 2899 time 0:00:02.255965 train NMSE 0.92858 test NMSE 0.93001\n",
      "step: 2999 time 0:00:02.263960 train NMSE 0.92768 test NMSE 0.92520\n",
      "Model saved!\n",
      "step: 99 time 0:00:02.580613 train NMSE 1.18890 test NMSE 1.08505\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.567154 train NMSE 1.08967 test NMSE 1.04592\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.552172 train NMSE 1.05144 test NMSE 1.05435\n",
      "step: 399 time 0:00:02.560151 train NMSE 1.03162 test NMSE 1.08207\n",
      "step: 499 time 0:00:02.557689 train NMSE 1.01096 test NMSE 0.98184\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.560686 train NMSE 1.00142 test NMSE 0.98323\n",
      "step: 699 time 0:00:02.572106 train NMSE 0.99535 test NMSE 0.98964\n",
      "step: 799 time 0:00:02.544232 train NMSE 0.98840 test NMSE 0.99736\n",
      "step: 899 time 0:00:02.548182 train NMSE 0.98218 test NMSE 0.97035\n",
      "Model saved!\n",
      "step: 999 time 0:00:02.558188 train NMSE 0.98160 test NMSE 0.96864\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.563648 train NMSE 0.97741 test NMSE 0.98568\n",
      "step: 1199 time 0:00:02.543237 train NMSE 0.97641 test NMSE 0.97283\n",
      "step: 1299 time 0:00:02.555706 train NMSE 0.97421 test NMSE 0.97411\n",
      "step: 1399 time 0:00:02.546232 train NMSE 0.97259 test NMSE 0.96389\n",
      "Model saved!\n",
      "step: 1499 time 0:00:02.565862 train NMSE 0.97060 test NMSE 0.96388\n",
      "Model saved!\n",
      "step: 1599 time 0:00:02.546232 train NMSE 0.97120 test NMSE 0.96195\n",
      "Model saved!\n",
      "step: 1699 time 0:00:02.558386 train NMSE 0.97027 test NMSE 0.96539\n",
      "step: 1799 time 0:00:02.560698 train NMSE 0.96817 test NMSE 0.96237\n",
      "step: 1899 time 0:00:02.556172 train NMSE 0.96818 test NMSE 0.96079\n",
      "Model saved!\n",
      "step: 1999 time 0:00:02.549770 train NMSE 0.96753 test NMSE 0.95976\n",
      "Model saved!\n",
      "step: 2099 time 0:00:02.554207 train NMSE 0.96700 test NMSE 0.95991\n",
      "step: 2199 time 0:00:02.546723 train NMSE 0.96650 test NMSE 0.96521\n",
      "step: 2299 time 0:00:02.548726 train NMSE 0.96572 test NMSE 0.97287\n",
      "step: 2399 time 0:00:02.553718 train NMSE 0.96467 test NMSE 0.97690\n",
      "step: 2499 time 0:00:02.553205 train NMSE 0.96484 test NMSE 0.96033\n",
      "step: 2599 time 0:00:02.549241 train NMSE 0.96368 test NMSE 0.96217\n",
      "step: 2699 time 0:00:02.551199 train NMSE 0.96461 test NMSE 0.96341\n",
      "step: 2799 time 0:00:02.552606 train NMSE 0.96346 test NMSE 0.96247\n",
      "step: 2899 time 0:00:02.550193 train NMSE 0.96324 test NMSE 0.96436\n",
      "step: 2999 time 0:00:02.559181 train NMSE 0.96345 test NMSE 0.95932\n",
      "Model saved!\n",
      "step: 99 time 0:00:02.866848 train NMSE 1.40552 test NMSE 1.39622\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.867370 train NMSE 1.21915 test NMSE 1.15612\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.866351 train NMSE 1.14061 test NMSE 1.09550\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.849903 train NMSE 1.06412 test NMSE 1.00816\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.868831 train NMSE 1.02885 test NMSE 0.99506\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.865399 train NMSE 1.00767 test NMSE 0.97777\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.874348 train NMSE 0.99706 test NMSE 0.96572\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.871355 train NMSE 0.98124 test NMSE 0.95559\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.870362 train NMSE 0.98072 test NMSE 0.96199\n",
      "step: 999 time 0:00:02.850855 train NMSE 0.96347 test NMSE 0.95196\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.891888 train NMSE 0.96435 test NMSE 0.94787\n",
      "Model saved!\n",
      "step: 1199 time 0:00:02.866906 train NMSE 0.95851 test NMSE 0.95849\n",
      "step: 1299 time 0:00:02.855373 train NMSE 0.95268 test NMSE 0.94220\n",
      "Model saved!\n",
      "step: 1399 time 0:00:02.860899 train NMSE 0.95186 test NMSE 0.92236\n",
      "Model saved!\n",
      "step: 1499 time 0:00:02.865371 train NMSE 0.94803 test NMSE 0.93410\n",
      "step: 1599 time 0:00:02.867359 train NMSE 0.94608 test NMSE 0.93707\n",
      "step: 1699 time 0:00:02.868383 train NMSE 0.94329 test NMSE 0.94544\n",
      "step: 1799 time 0:00:02.850430 train NMSE 0.94194 test NMSE 0.93416\n",
      "step: 1899 time 0:00:02.849910 train NMSE 0.93836 test NMSE 0.94764\n",
      "step: 1999 time 0:00:02.861397 train NMSE 0.93663 test NMSE 0.94082\n",
      "step: 2099 time 0:00:02.849421 train NMSE 0.93722 test NMSE 0.92406\n",
      "step: 2199 time 0:00:02.863340 train NMSE 0.93551 test NMSE 0.93382\n",
      "step: 2299 time 0:00:02.864368 train NMSE 0.93701 test NMSE 0.92819\n",
      "step: 2399 time 0:00:02.897249 train NMSE 0.93503 test NMSE 0.92672\n",
      "step: 2499 time 0:00:02.864348 train NMSE 0.93080 test NMSE 0.92779\n",
      "step: 2599 time 0:00:02.854929 train NMSE 0.93218 test NMSE 0.93102\n",
      "step: 2699 time 0:00:02.853366 train NMSE 0.93034 test NMSE 0.93031\n",
      "step: 2799 time 0:00:02.865772 train NMSE 0.92957 test NMSE 0.91236\n",
      "Model saved!\n",
      "step: 2899 time 0:00:02.864354 train NMSE 0.92756 test NMSE 0.92590\n",
      "step: 2999 time 0:00:02.854870 train NMSE 0.92787 test NMSE 0.91943\n",
      "step: 99 time 0:00:03.152566 train NMSE 1.32458 test NMSE 1.28559\n",
      "Model saved!\n",
      "step: 199 time 0:00:03.168223 train NMSE 1.18007 test NMSE 1.27295\n",
      "Model saved!\n",
      "step: 299 time 0:00:03.166583 train NMSE 1.13138 test NMSE 1.00698\n",
      "Model saved!\n",
      "step: 399 time 0:00:03.171516 train NMSE 1.07868 test NMSE 1.22958\n",
      "step: 499 time 0:00:03.160069 train NMSE 1.04075 test NMSE 1.14159\n",
      "step: 599 time 0:00:03.162613 train NMSE 1.01567 test NMSE 0.97659\n",
      "Model saved!\n",
      "step: 699 time 0:00:03.168534 train NMSE 1.01518 test NMSE 0.97545\n",
      "Model saved!\n",
      "step: 799 time 0:00:03.149619 train NMSE 0.99429 test NMSE 0.96900\n",
      "Model saved!\n",
      "step: 899 time 0:00:03.162567 train NMSE 0.99225 test NMSE 0.96576\n",
      "Model saved!\n",
      "step: 999 time 0:00:03.182019 train NMSE 0.99573 test NMSE 1.05659\n",
      "step: 1099 time 0:00:03.161541 train NMSE 0.98536 test NMSE 0.95756\n",
      "Model saved!\n",
      "step: 1199 time 0:00:03.167569 train NMSE 0.97856 test NMSE 0.96457\n",
      "step: 1299 time 0:00:03.168559 train NMSE 0.98925 test NMSE 0.96541\n",
      "step: 1399 time 0:00:03.166529 train NMSE 0.97088 test NMSE 0.97299\n",
      "step: 1499 time 0:00:03.171525 train NMSE 0.98154 test NMSE 0.95866\n",
      "step: 1599 time 0:00:03.175547 train NMSE 0.97224 test NMSE 1.01392\n",
      "step: 1699 time 0:00:03.185503 train NMSE 0.97495 test NMSE 0.95995\n",
      "step: 1799 time 0:00:03.159297 train NMSE 0.96938 test NMSE 0.96258\n",
      "step: 1899 time 0:00:03.196488 train NMSE 0.97231 test NMSE 1.00827\n",
      "step: 1999 time 0:00:03.173044 train NMSE 0.97007 test NMSE 0.95652\n",
      "Model saved!\n",
      "step: 2099 time 0:00:03.158550 train NMSE 0.97013 test NMSE 0.95948\n",
      "step: 2199 time 0:00:03.169584 train NMSE 0.96777 test NMSE 0.96074\n",
      "step: 2299 time 0:00:03.167557 train NMSE 0.96505 test NMSE 0.95337\n",
      "Model saved!\n",
      "step: 2399 time 0:00:03.168523 train NMSE 0.96324 test NMSE 0.96051\n",
      "step: 2499 time 0:00:03.177515 train NMSE 0.96654 test NMSE 0.96063\n",
      "step: 2599 time 0:00:03.170286 train NMSE 0.96431 test NMSE 0.95741\n",
      "step: 2699 time 0:00:03.170529 train NMSE 0.96442 test NMSE 0.95882\n",
      "step: 2799 time 0:00:03.170789 train NMSE 0.96251 test NMSE 0.96001\n",
      "step: 2899 time 0:00:03.183529 train NMSE 0.96341 test NMSE 0.96170\n",
      "step: 2999 time 0:00:03.182685 train NMSE 0.96084 test NMSE 0.98871\n"
     ]
    }
   ],
   "source": [
    "L  = 2  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:02.328769 train NMSE 3.04297 test NMSE 2.24585\n",
      "step: 199 time 0:00:02.322796 train NMSE 2.00323 test NMSE 1.63368\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.319487 train NMSE 1.55926 test NMSE 1.36092\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.333771 train NMSE 1.28918 test NMSE 1.20427\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.328346 train NMSE 1.12187 test NMSE 1.03400\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.329811 train NMSE 1.01700 test NMSE 0.99576\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.329788 train NMSE 0.94784 test NMSE 0.97601\n",
      "Model saved!\n",
      "step: 799 time 0:00:02.325322 train NMSE 0.88198 test NMSE 0.87721\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.336735 train NMSE 0.84502 test NMSE 0.83206\n",
      "Model saved!\n",
      "step: 999 time 0:00:02.322811 train NMSE 0.81416 test NMSE 0.79796\n",
      "Model saved!\n",
      "step: 1099 time 0:00:02.326745 train NMSE 0.79024 test NMSE 0.76933\n",
      "Model saved!\n",
      "step: 1199 time 0:00:02.322950 train NMSE 0.76853 test NMSE 0.74773\n",
      "Model saved!\n",
      "step: 1299 time 0:00:02.329299 train NMSE 0.75286 test NMSE 0.73851\n",
      "Model saved!\n",
      "step: 1399 time 0:00:02.323284 train NMSE 0.74003 test NMSE 0.75483\n",
      "step: 1499 time 0:00:02.329099 train NMSE 0.72752 test NMSE 0.69625\n",
      "Model saved!\n",
      "step: 1599 time 0:00:02.324848 train NMSE 0.71853 test NMSE 0.70152\n",
      "step: 1699 time 0:00:02.326330 train NMSE 0.70520 test NMSE 0.69867\n",
      "step: 1799 time 0:00:02.316818 train NMSE 0.69451 test NMSE 0.68747\n",
      "Model saved!\n",
      "step: 1899 time 0:00:02.323782 train NMSE 0.68602 test NMSE 0.68431\n",
      "Model saved!\n",
      "step: 1999 time 0:00:02.319980 train NMSE 0.67937 test NMSE 0.67377\n",
      "Model saved!\n",
      "step: 2099 time 0:00:02.320987 train NMSE 0.67463 test NMSE 0.66332\n",
      "Model saved!\n",
      "step: 2199 time 0:00:02.318336 train NMSE 0.66982 test NMSE 0.66028\n",
      "Model saved!\n",
      "step: 2299 time 0:00:02.332307 train NMSE 0.66161 test NMSE 0.65578\n",
      "Model saved!\n",
      "step: 2399 time 0:00:02.317873 train NMSE 0.65864 test NMSE 0.65824\n",
      "step: 2499 time 0:00:02.317816 train NMSE 0.65122 test NMSE 0.64463\n",
      "Model saved!\n",
      "step: 2599 time 0:00:02.318845 train NMSE 0.64666 test NMSE 0.64466\n",
      "step: 2699 time 0:00:02.347273 train NMSE 0.64194 test NMSE 0.64602\n",
      "step: 2799 time 0:00:02.323312 train NMSE 0.63587 test NMSE 0.64128\n",
      "Model saved!\n",
      "step: 2899 time 0:00:02.319849 train NMSE 0.62959 test NMSE 0.61196\n",
      "Model saved!\n",
      "step: 2999 time 0:00:02.342743 train NMSE 0.62495 test NMSE 0.61782\n",
      "step: 99 time 0:00:02.692337 train NMSE 1.79316 test NMSE 1.35319\n",
      "Model saved!\n",
      "step: 199 time 0:00:02.688978 train NMSE 1.20651 test NMSE 0.92015\n",
      "Model saved!\n",
      "step: 299 time 0:00:02.749174 train NMSE 0.91828 test NMSE 0.79130\n",
      "Model saved!\n",
      "step: 399 time 0:00:02.693301 train NMSE 0.78604 test NMSE 0.70356\n",
      "Model saved!\n",
      "step: 499 time 0:00:02.692328 train NMSE 0.71888 test NMSE 0.66821\n",
      "Model saved!\n",
      "step: 599 time 0:00:02.707280 train NMSE 0.67703 test NMSE 0.63219\n",
      "Model saved!\n",
      "step: 699 time 0:00:02.701826 train NMSE 0.65245 test NMSE 0.65124\n",
      "step: 799 time 0:00:02.704787 train NMSE 0.63395 test NMSE 0.59626\n",
      "Model saved!\n",
      "step: 899 time 0:00:02.694323 train NMSE 0.62387 test NMSE 0.57863\n",
      "Model saved!\n",
      "step: 999 time 0:00:02.692337 train NMSE 0.61191 test NMSE 0.62283\n",
      "step: 1099 time 0:00:02.689289 train NMSE 0.59890 test NMSE 0.60691\n",
      "step: 1199 time 0:00:02.697797 train NMSE 0.59079 test NMSE 0.58626\n",
      "step: 1299 time 0:00:02.693389 train NMSE 0.57758 test NMSE 0.63717\n",
      "step: 1399 time 0:00:02.691233 train NMSE 0.57458 test NMSE 0.55952\n",
      "Model saved!\n",
      "step: 1499 time 0:00:02.701434 train NMSE 0.56770 test NMSE 0.59465\n",
      "step: 1599 time 0:00:02.696310 train NMSE 0.56224 test NMSE 0.56733\n",
      "step: 1699 time 0:00:02.698320 train NMSE 0.55832 test NMSE 0.53429\n",
      "Model saved!\n",
      "step: 1799 time 0:00:02.714761 train NMSE 0.55263 test NMSE 0.54925\n",
      "step: 1899 time 0:00:02.692322 train NMSE 0.54820 test NMSE 0.54391\n",
      "step: 1999 time 0:00:02.700295 train NMSE 0.54336 test NMSE 0.55183\n",
      "step: 2099 time 0:00:02.687849 train NMSE 0.54320 test NMSE 0.53443\n",
      "step: 2199 time 0:00:02.694317 train NMSE 0.54180 test NMSE 0.54377\n",
      "step: 2299 time 0:00:02.685859 train NMSE 0.53584 test NMSE 0.53128\n",
      "Model saved!\n",
      "step: 2399 time 0:00:02.699804 train NMSE 0.53522 test NMSE 0.54523\n",
      "step: 2499 time 0:00:02.697810 train NMSE 0.52876 test NMSE 0.51815\n",
      "Model saved!\n",
      "step: 2599 time 0:00:02.715751 train NMSE 0.52566 test NMSE 0.51571\n",
      "Model saved!\n",
      "step: 2699 time 0:00:02.711746 train NMSE 0.52067 test NMSE 0.52519\n",
      "step: 2799 time 0:00:02.718726 train NMSE 0.51727 test NMSE 0.50407\n",
      "Model saved!\n",
      "step: 2899 time 0:00:02.703367 train NMSE 0.51545 test NMSE 0.53256\n",
      "step: 2999 time 0:00:02.702769 train NMSE 0.50781 test NMSE 0.49494\n",
      "Model saved!\n",
      "step: 99 time 0:00:03.085931 train NMSE 1.34800 test NMSE 1.21557\n",
      "Model saved!\n",
      "step: 199 time 0:00:03.089733 train NMSE 1.01713 test NMSE 0.95679\n",
      "Model saved!\n",
      "step: 299 time 0:00:03.094413 train NMSE 0.83015 test NMSE 0.72282\n",
      "Model saved!\n",
      "step: 399 time 0:00:03.066796 train NMSE 0.73189 test NMSE 0.63912\n",
      "Model saved!\n",
      "step: 499 time 0:00:03.084775 train NMSE 0.67335 test NMSE 0.72175\n",
      "step: 599 time 0:00:03.091101 train NMSE 0.65107 test NMSE 0.67837\n",
      "step: 699 time 0:00:03.083752 train NMSE 0.62405 test NMSE 0.69262\n",
      "step: 799 time 0:00:03.084285 train NMSE 0.61410 test NMSE 0.60317\n",
      "Model saved!\n",
      "step: 899 time 0:00:03.070784 train NMSE 0.59916 test NMSE 0.58571\n",
      "Model saved!\n",
      "step: 999 time 0:00:03.086756 train NMSE 0.58862 test NMSE 0.62697\n",
      "step: 1099 time 0:00:03.061810 train NMSE 0.57903 test NMSE 0.55862\n",
      "Model saved!\n",
      "step: 1199 time 0:00:03.086068 train NMSE 0.56513 test NMSE 0.51372\n",
      "Model saved!\n",
      "step: 1299 time 0:00:03.085264 train NMSE 0.56105 test NMSE 0.54585\n",
      "step: 1399 time 0:00:03.064800 train NMSE 0.55464 test NMSE 0.63852\n",
      "step: 1499 time 0:00:03.074786 train NMSE 0.55198 test NMSE 0.53657\n",
      "step: 1599 time 0:00:03.086742 train NMSE 0.54143 test NMSE 0.52822\n",
      "step: 1699 time 0:00:03.080773 train NMSE 0.54072 test NMSE 0.51211\n",
      "Model saved!\n",
      "step: 1799 time 0:00:03.087740 train NMSE 0.53337 test NMSE 0.54019\n",
      "step: 1899 time 0:00:03.070805 train NMSE 0.52588 test NMSE 0.51265\n",
      "step: 1999 time 0:00:03.084758 train NMSE 0.52467 test NMSE 0.51905\n",
      "step: 2099 time 0:00:03.091721 train NMSE 0.51945 test NMSE 0.51550\n",
      "step: 2199 time 0:00:03.094720 train NMSE 0.51587 test NMSE 0.53639\n",
      "step: 2299 time 0:00:03.083287 train NMSE 0.51447 test NMSE 0.51268\n",
      "step: 2399 time 0:00:03.083606 train NMSE 0.50742 test NMSE 0.49880\n",
      "Model saved!\n",
      "step: 2499 time 0:00:03.086743 train NMSE 0.49982 test NMSE 0.47974\n",
      "Model saved!\n",
      "step: 2599 time 0:00:03.087247 train NMSE 0.49911 test NMSE 0.46412\n",
      "Model saved!\n",
      "step: 2699 time 0:00:03.104694 train NMSE 0.49631 test NMSE 0.47350\n",
      "step: 2799 time 0:00:03.085757 train NMSE 0.49462 test NMSE 0.52053\n",
      "step: 2899 time 0:00:03.072780 train NMSE 0.48716 test NMSE 0.51711\n",
      "step: 2999 time 0:00:03.088766 train NMSE 0.48346 test NMSE 0.48746\n",
      "step: 99 time 0:00:03.453776 train NMSE 1.11631 test NMSE 1.46069\n",
      "Model saved!\n",
      "step: 199 time 0:00:03.447082 train NMSE 0.99202 test NMSE 0.92000\n",
      "Model saved!\n",
      "step: 299 time 0:00:03.435851 train NMSE 0.90395 test NMSE 0.86004\n",
      "Model saved!\n",
      "step: 399 time 0:00:03.432352 train NMSE 0.87008 test NMSE 0.80939\n",
      "Model saved!\n",
      "step: 499 time 0:00:03.426847 train NMSE 0.81318 test NMSE 0.80794\n",
      "Model saved!\n",
      "step: 599 time 0:00:03.449771 train NMSE 0.78016 test NMSE 0.73759\n",
      "Model saved!\n",
      "step: 699 time 0:00:03.428860 train NMSE 0.75525 test NMSE 0.70874\n",
      "Model saved!\n",
      "step: 799 time 0:00:03.432817 train NMSE 0.74101 test NMSE 0.70989\n",
      "step: 899 time 0:00:03.423839 train NMSE 0.72368 test NMSE 0.71567\n",
      "step: 999 time 0:00:03.431863 train NMSE 0.70508 test NMSE 0.84447\n",
      "step: 1099 time 0:00:03.429824 train NMSE 0.69314 test NMSE 0.75007\n",
      "step: 1199 time 0:00:03.436746 train NMSE 0.68274 test NMSE 0.69721\n",
      "Model saved!\n",
      "step: 1299 time 0:00:03.438309 train NMSE 0.68081 test NMSE 0.73810\n",
      "step: 1399 time 0:00:03.435809 train NMSE 0.66084 test NMSE 0.70322\n",
      "step: 1499 time 0:00:03.451791 train NMSE 0.66465 test NMSE 0.63359\n",
      "Model saved!\n",
      "step: 1599 time 0:00:03.423876 train NMSE 0.65921 test NMSE 0.63839\n",
      "step: 1699 time 0:00:03.431819 train NMSE 0.65214 test NMSE 0.63382\n",
      "step: 1799 time 0:00:03.435836 train NMSE 0.63497 test NMSE 0.63237\n",
      "Model saved!\n",
      "step: 1899 time 0:00:03.432829 train NMSE 0.63119 test NMSE 0.62529\n",
      "Model saved!\n",
      "step: 1999 time 0:00:03.458746 train NMSE 0.61732 test NMSE 0.60721\n",
      "Model saved!\n",
      "step: 2099 time 0:00:03.442815 train NMSE 0.61063 test NMSE 0.57960\n",
      "Model saved!\n",
      "step: 2199 time 0:00:03.439422 train NMSE 0.60770 test NMSE 0.58280\n",
      "step: 2299 time 0:00:03.474703 train NMSE 0.59675 test NMSE 0.64672\n",
      "step: 2399 time 0:00:03.414485 train NMSE 0.59351 test NMSE 0.57486\n",
      "Model saved!\n",
      "step: 2499 time 0:00:03.449318 train NMSE 0.58412 test NMSE 0.59721\n",
      "step: 2599 time 0:00:03.426833 train NMSE 0.57795 test NMSE 0.59264\n",
      "step: 2699 time 0:00:03.444238 train NMSE 0.57581 test NMSE 0.56933\n",
      "Model saved!\n",
      "step: 2799 time 0:00:03.418388 train NMSE 0.56949 test NMSE 0.54348\n",
      "Model saved!\n",
      "step: 2899 time 0:00:03.452763 train NMSE 0.55993 test NMSE 0.54987\n",
      "step: 2999 time 0:00:03.444206 train NMSE 0.56041 test NMSE 0.55871\n",
      "step: 99 time 0:00:03.811801 train NMSE 1.06285 test NMSE 1.06221\n",
      "Model saved!\n",
      "step: 199 time 0:00:03.827769 train NMSE 0.99574 test NMSE 0.96513\n",
      "Model saved!\n",
      "step: 299 time 0:00:03.822900 train NMSE 0.96463 test NMSE 1.05040\n",
      "step: 399 time 0:00:03.832789 train NMSE 0.94018 test NMSE 0.90062\n",
      "Model saved!\n",
      "step: 499 time 0:00:03.823785 train NMSE 0.91506 test NMSE 0.87037\n",
      "Model saved!\n",
      "step: 599 time 0:00:03.829769 train NMSE 0.90125 test NMSE 0.88139\n",
      "step: 699 time 0:00:03.830261 train NMSE 0.89409 test NMSE 0.86341\n",
      "Model saved!\n",
      "step: 799 time 0:00:03.852716 train NMSE 0.87544 test NMSE 0.86472\n",
      "step: 899 time 0:00:03.821557 train NMSE 0.86410 test NMSE 0.86423\n",
      "step: 999 time 0:00:03.815791 train NMSE 0.85502 test NMSE 0.84821\n",
      "Model saved!\n",
      "step: 1099 time 0:00:03.818316 train NMSE 0.84662 test NMSE 0.83239\n",
      "Model saved!\n",
      "step: 1199 time 0:00:03.826516 train NMSE 0.84078 test NMSE 0.82690\n",
      "Model saved!\n",
      "step: 1299 time 0:00:03.811815 train NMSE 0.83084 test NMSE 0.81517\n",
      "Model saved!\n",
      "step: 1399 time 0:00:03.824328 train NMSE 0.82691 test NMSE 0.81432\n",
      "Model saved!\n",
      "step: 1499 time 0:00:03.812308 train NMSE 0.82424 test NMSE 0.80461\n",
      "Model saved!\n",
      "step: 1599 time 0:00:03.808318 train NMSE 0.81087 test NMSE 0.91715\n",
      "step: 1699 time 0:00:03.826272 train NMSE 0.81320 test NMSE 0.80110\n",
      "Model saved!\n",
      "step: 1799 time 0:00:03.826271 train NMSE 0.80255 test NMSE 0.78369\n",
      "Model saved!\n",
      "step: 1899 time 0:00:03.831777 train NMSE 0.80384 test NMSE 0.82645\n",
      "step: 1999 time 0:00:03.822454 train NMSE 0.79843 test NMSE 0.78149\n",
      "Model saved!\n",
      "step: 2099 time 0:00:03.834741 train NMSE 0.78853 test NMSE 0.80418\n",
      "step: 2199 time 0:00:03.829004 train NMSE 0.78341 test NMSE 0.78676\n",
      "step: 2299 time 0:00:03.803359 train NMSE 0.77729 test NMSE 0.75703\n",
      "Model saved!\n",
      "step: 2399 time 0:00:03.810728 train NMSE 0.77126 test NMSE 0.76842\n",
      "step: 2499 time 0:00:03.819827 train NMSE 0.76826 test NMSE 0.78181\n",
      "step: 2599 time 0:00:03.810335 train NMSE 0.75792 test NMSE 0.78161\n",
      "step: 2699 time 0:00:03.818866 train NMSE 0.75436 test NMSE 0.75723\n",
      "step: 2799 time 0:00:03.830777 train NMSE 0.74819 test NMSE 0.77611\n",
      "step: 2899 time 0:00:03.826812 train NMSE 0.74784 test NMSE 0.73838\n",
      "Model saved!\n",
      "step: 2999 time 0:00:03.837763 train NMSE 0.73860 test NMSE 0.71508\n",
      "Model saved!\n"
     ]
    }
   ],
   "source": [
    "L  = 4  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:03.084787 train NMSE 0.94930 test NMSE 0.87908\n",
      "Model saved!\n",
      "step: 199 time 0:00:03.078808 train NMSE 0.83990 test NMSE 0.78845\n",
      "Model saved!\n",
      "step: 299 time 0:00:03.069813 train NMSE 0.76206 test NMSE 0.72547\n",
      "Model saved!\n",
      "step: 399 time 0:00:03.075320 train NMSE 0.70028 test NMSE 0.66696\n",
      "Model saved!\n",
      "step: 499 time 0:00:03.079411 train NMSE 0.65307 test NMSE 0.64348\n",
      "Model saved!\n",
      "step: 599 time 0:00:03.094481 train NMSE 0.61704 test NMSE 0.57569\n",
      "Model saved!\n",
      "step: 699 time 0:00:03.101301 train NMSE 0.58300 test NMSE 0.56728\n",
      "Model saved!\n",
      "step: 799 time 0:00:03.084297 train NMSE 0.55587 test NMSE 0.56548\n",
      "Model saved!\n",
      "step: 899 time 0:00:03.096241 train NMSE 0.53801 test NMSE 0.52273\n",
      "Model saved!\n",
      "step: 999 time 0:00:03.083763 train NMSE 0.50988 test NMSE 0.49684\n",
      "Model saved!\n",
      "step: 1099 time 0:00:03.090745 train NMSE 0.47908 test NMSE 0.45228\n",
      "Model saved!\n",
      "step: 1199 time 0:00:03.078091 train NMSE 0.45149 test NMSE 0.43615\n",
      "Model saved!\n",
      "step: 1299 time 0:00:03.095235 train NMSE 0.42760 test NMSE 0.40622\n",
      "Model saved!\n",
      "step: 1399 time 0:00:03.081755 train NMSE 0.41246 test NMSE 0.41043\n",
      "step: 1499 time 0:00:03.076291 train NMSE 0.40081 test NMSE 0.39777\n",
      "Model saved!\n",
      "step: 1599 time 0:00:03.079807 train NMSE 0.39112 test NMSE 0.39647\n",
      "Model saved!\n",
      "step: 1699 time 0:00:03.095602 train NMSE 0.38104 test NMSE 0.39571\n",
      "Model saved!\n",
      "step: 1799 time 0:00:03.081819 train NMSE 0.37270 test NMSE 0.37060\n",
      "Model saved!\n",
      "step: 1899 time 0:00:03.078306 train NMSE 0.36635 test NMSE 0.35981\n",
      "Model saved!\n",
      "step: 1999 time 0:00:03.079802 train NMSE 0.36195 test NMSE 0.36665\n",
      "step: 2099 time 0:00:03.090742 train NMSE 0.35630 test NMSE 0.36568\n",
      "step: 2199 time 0:00:03.079908 train NMSE 0.35374 test NMSE 0.32673\n",
      "Model saved!\n",
      "step: 2299 time 0:00:03.072791 train NMSE 0.34562 test NMSE 0.35438\n",
      "step: 2399 time 0:00:03.074814 train NMSE 0.34468 test NMSE 0.34869\n",
      "step: 2499 time 0:00:03.095795 train NMSE 0.34027 test NMSE 0.33150\n",
      "step: 2599 time 0:00:03.075990 train NMSE 0.33803 test NMSE 0.34817\n",
      "step: 2699 time 0:00:03.083444 train NMSE 0.33651 test NMSE 0.33881\n",
      "step: 2799 time 0:00:03.084792 train NMSE 0.33561 test NMSE 0.33055\n",
      "step: 2899 time 0:00:03.102721 train NMSE 0.32968 test NMSE 0.35055\n",
      "step: 2999 time 0:00:03.086139 train NMSE 0.32709 test NMSE 0.32988\n",
      "step: 99 time 0:00:03.612802 train NMSE 0.44364 test NMSE 0.39833\n",
      "Model saved!\n",
      "step: 199 time 0:00:03.628217 train NMSE 0.38481 test NMSE 0.35998\n",
      "Model saved!\n",
      "step: 299 time 0:00:03.625337 train NMSE 0.35780 test NMSE 0.36944\n",
      "step: 399 time 0:00:03.622340 train NMSE 0.33674 test NMSE 0.31145\n",
      "Model saved!\n",
      "step: 499 time 0:00:03.624341 train NMSE 0.32048 test NMSE 0.30962\n",
      "Model saved!\n",
      "step: 599 time 0:00:03.618451 train NMSE 0.30741 test NMSE 0.29893\n",
      "Model saved!\n",
      "step: 699 time 0:00:03.626327 train NMSE 0.29742 test NMSE 0.27627\n",
      "Model saved!\n",
      "step: 799 time 0:00:03.632338 train NMSE 0.28901 test NMSE 0.31172\n",
      "step: 899 time 0:00:03.612895 train NMSE 0.28173 test NMSE 0.27382\n",
      "Model saved!\n",
      "step: 999 time 0:00:03.614379 train NMSE 0.27269 test NMSE 0.25374\n",
      "Model saved!\n",
      "step: 1099 time 0:00:03.678799 train NMSE 0.27083 test NMSE 0.27394\n",
      "step: 1199 time 0:00:03.619257 train NMSE 0.26595 test NMSE 0.25533\n",
      "step: 1299 time 0:00:03.614391 train NMSE 0.25892 test NMSE 0.25770\n",
      "step: 1399 time 0:00:03.615369 train NMSE 0.25972 test NMSE 0.25604\n",
      "step: 1499 time 0:00:03.613391 train NMSE 0.25414 test NMSE 0.25403\n",
      "step: 1599 time 0:00:03.627726 train NMSE 0.25656 test NMSE 0.24993\n",
      "Model saved!\n",
      "step: 1699 time 0:00:03.606891 train NMSE 0.24825 test NMSE 0.25409\n",
      "step: 1799 time 0:00:03.616959 train NMSE 0.24882 test NMSE 0.23819\n",
      "Model saved!\n",
      "step: 1899 time 0:00:03.603943 train NMSE 0.24563 test NMSE 0.25070\n",
      "step: 1999 time 0:00:03.618372 train NMSE 0.24322 test NMSE 0.25765\n",
      "step: 2099 time 0:00:03.609203 train NMSE 0.24388 test NMSE 0.22623\n",
      "Model saved!\n",
      "step: 2199 time 0:00:03.617813 train NMSE 0.24141 test NMSE 0.25518\n",
      "step: 2299 time 0:00:03.643378 train NMSE 0.24099 test NMSE 0.26062\n",
      "step: 2399 time 0:00:03.619663 train NMSE 0.24006 test NMSE 0.24151\n",
      "step: 2499 time 0:00:03.618368 train NMSE 0.23652 test NMSE 0.25096\n",
      "step: 2599 time 0:00:03.626314 train NMSE 0.23799 test NMSE 0.25353\n",
      "step: 2699 time 0:00:03.617972 train NMSE 0.23828 test NMSE 0.22871\n",
      "step: 2799 time 0:00:03.629339 train NMSE 0.23469 test NMSE 0.24783\n",
      "step: 2899 time 0:00:03.616919 train NMSE 0.23409 test NMSE 0.22053\n",
      "Model saved!\n",
      "step: 2999 time 0:00:03.617592 train NMSE 0.23477 test NMSE 0.22740\n",
      "step: 99 time 0:00:04.141334 train NMSE 0.34345 test NMSE 0.30815\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.153449 train NMSE 0.31806 test NMSE 0.28056\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.158667 train NMSE 0.28710 test NMSE 0.27494\n",
      "Model saved!\n",
      "step: 399 time 0:00:04.163874 train NMSE 0.27589 test NMSE 0.25534\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.162862 train NMSE 0.27796 test NMSE 0.26757\n",
      "step: 599 time 0:00:04.157305 train NMSE 0.26654 test NMSE 0.26195\n",
      "step: 699 time 0:00:04.165403 train NMSE 0.26128 test NMSE 0.25582\n",
      "step: 799 time 0:00:04.162467 train NMSE 0.25016 test NMSE 0.24116\n",
      "Model saved!\n",
      "step: 899 time 0:00:04.162692 train NMSE 0.24737 test NMSE 0.25756\n",
      "step: 999 time 0:00:04.183978 train NMSE 0.24701 test NMSE 0.25129\n",
      "step: 1099 time 0:00:04.140005 train NMSE 0.24765 test NMSE 0.22777\n",
      "Model saved!\n",
      "step: 1199 time 0:00:04.163917 train NMSE 0.23928 test NMSE 0.23064\n",
      "step: 1299 time 0:00:04.156358 train NMSE 0.24059 test NMSE 0.23566\n",
      "step: 1399 time 0:00:04.161961 train NMSE 0.23592 test NMSE 0.23939\n",
      "step: 1499 time 0:00:04.143428 train NMSE 0.23393 test NMSE 0.21882\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.167441 train NMSE 0.23304 test NMSE 0.22750\n",
      "step: 1699 time 0:00:04.153256 train NMSE 0.23316 test NMSE 0.23134\n",
      "step: 1799 time 0:00:04.145909 train NMSE 0.23659 test NMSE 0.24375\n",
      "step: 1899 time 0:00:04.161921 train NMSE 0.23208 test NMSE 0.23148\n",
      "step: 1999 time 0:00:04.151988 train NMSE 0.23010 test NMSE 0.22787\n",
      "step: 2099 time 0:00:04.155028 train NMSE 0.22796 test NMSE 0.23380\n",
      "step: 2199 time 0:00:04.167886 train NMSE 0.22704 test NMSE 0.21829\n",
      "Model saved!\n",
      "step: 2299 time 0:00:04.163919 train NMSE 0.22488 test NMSE 0.23553\n",
      "step: 2399 time 0:00:04.165385 train NMSE 0.22839 test NMSE 0.21712\n",
      "Model saved!\n",
      "step: 2499 time 0:00:04.160829 train NMSE 0.22807 test NMSE 0.21447\n",
      "Model saved!\n",
      "step: 2599 time 0:00:04.155223 train NMSE 0.22686 test NMSE 0.22880\n",
      "step: 2699 time 0:00:04.154946 train NMSE 0.22281 test NMSE 0.21307\n",
      "Model saved!\n",
      "step: 2799 time 0:00:04.157876 train NMSE 0.22613 test NMSE 0.21649\n",
      "step: 2899 time 0:00:04.149911 train NMSE 0.22548 test NMSE 0.22177\n",
      "step: 2999 time 0:00:04.157892 train NMSE 0.22299 test NMSE 0.20324\n",
      "Model saved!\n",
      "step: 99 time 0:00:04.687480 train NMSE 0.32530 test NMSE 0.30558\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.706409 train NMSE 0.30425 test NMSE 0.28305\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.672500 train NMSE 0.28144 test NMSE 0.28518\n",
      "step: 399 time 0:00:04.694001 train NMSE 0.28150 test NMSE 0.26161\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.679502 train NMSE 0.26350 test NMSE 0.27677\n",
      "step: 599 time 0:00:04.693265 train NMSE 0.26630 test NMSE 0.26329\n",
      "step: 699 time 0:00:04.695451 train NMSE 0.25305 test NMSE 0.28015\n",
      "step: 799 time 0:00:04.703035 train NMSE 0.24924 test NMSE 0.26462\n",
      "step: 899 time 0:00:04.691461 train NMSE 0.24979 test NMSE 0.23595\n",
      "Model saved!\n",
      "step: 999 time 0:00:04.679829 train NMSE 0.24905 test NMSE 0.24169\n",
      "step: 1099 time 0:00:04.909142 train NMSE 0.24911 test NMSE 0.23322\n",
      "Model saved!\n",
      "step: 1199 time 0:00:04.794216 train NMSE 0.23916 test NMSE 0.25849\n",
      "step: 1299 time 0:00:04.761820 train NMSE 0.23768 test NMSE 0.22655\n",
      "Model saved!\n",
      "step: 1399 time 0:00:04.780216 train NMSE 0.23521 test NMSE 0.23840\n",
      "step: 1499 time 0:00:04.730876 train NMSE 0.23883 test NMSE 0.23140\n",
      "step: 1599 time 0:00:04.755856 train NMSE 0.23453 test NMSE 0.21163\n",
      "Model saved!\n",
      "step: 1699 time 0:00:04.679541 train NMSE 0.23284 test NMSE 0.24041\n",
      "step: 1799 time 0:00:04.684468 train NMSE 0.23750 test NMSE 0.22943\n",
      "step: 1899 time 0:00:04.680816 train NMSE 0.22878 test NMSE 0.22598\n",
      "step: 1999 time 0:00:04.687761 train NMSE 0.23614 test NMSE 0.21824\n",
      "step: 2099 time 0:00:04.676489 train NMSE 0.22643 test NMSE 0.22961\n",
      "step: 2199 time 0:00:04.671503 train NMSE 0.23245 test NMSE 0.23155\n",
      "step: 2299 time 0:00:04.692446 train NMSE 0.22022 test NMSE 0.21915\n",
      "step: 2399 time 0:00:04.668510 train NMSE 0.22310 test NMSE 0.21662\n",
      "step: 2499 time 0:00:04.681476 train NMSE 0.22392 test NMSE 0.20934\n",
      "Model saved!\n",
      "step: 2599 time 0:00:04.659534 train NMSE 0.22071 test NMSE 0.22507\n",
      "step: 2699 time 0:00:04.665518 train NMSE 0.22354 test NMSE 0.22226\n",
      "step: 2799 time 0:00:04.680480 train NMSE 0.22039 test NMSE 0.23039\n",
      "step: 2899 time 0:00:04.684467 train NMSE 0.22224 test NMSE 0.22267\n",
      "step: 2999 time 0:00:04.671502 train NMSE 0.21716 test NMSE 0.21952\n",
      "step: 99 time 0:00:05.204076 train NMSE 0.32218 test NMSE 0.29773\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.263918 train NMSE 0.29520 test NMSE 0.27724\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.195101 train NMSE 0.27388 test NMSE 0.25892\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.190115 train NMSE 0.27957 test NMSE 0.27460\n",
      "step: 499 time 0:00:05.205074 train NMSE 0.26322 test NMSE 0.25420\n",
      "Model saved!\n",
      "step: 599 time 0:00:05.183147 train NMSE 0.25073 test NMSE 0.24941\n",
      "Model saved!\n",
      "step: 699 time 0:00:05.195622 train NMSE 0.24918 test NMSE 0.24140\n",
      "Model saved!\n",
      "step: 799 time 0:00:05.203108 train NMSE 0.24562 test NMSE 0.24896\n",
      "step: 899 time 0:00:05.190635 train NMSE 0.24614 test NMSE 0.23998\n",
      "Model saved!\n",
      "step: 999 time 0:00:05.178701 train NMSE 0.25237 test NMSE 0.23758\n",
      "Model saved!\n",
      "step: 1099 time 0:00:05.190131 train NMSE 0.23500 test NMSE 0.23256\n",
      "Model saved!\n",
      "step: 1199 time 0:00:05.203736 train NMSE 0.24342 test NMSE 0.25059\n",
      "step: 1299 time 0:00:05.186137 train NMSE 0.23210 test NMSE 0.24054\n",
      "step: 1399 time 0:00:05.191632 train NMSE 0.23153 test NMSE 0.22573\n",
      "Model saved!\n",
      "step: 1499 time 0:00:05.188624 train NMSE 0.23226 test NMSE 0.24009\n",
      "step: 1599 time 0:00:05.183696 train NMSE 0.23749 test NMSE 0.24261\n",
      "step: 1699 time 0:00:05.210073 train NMSE 0.23038 test NMSE 0.23344\n",
      "step: 1799 time 0:00:05.191125 train NMSE 0.23251 test NMSE 0.23288\n",
      "step: 1899 time 0:00:05.207575 train NMSE 0.22363 test NMSE 0.19990\n",
      "Model saved!\n",
      "step: 1999 time 0:00:05.189161 train NMSE 0.22781 test NMSE 0.24436\n",
      "step: 2099 time 0:00:05.180156 train NMSE 0.22642 test NMSE 0.23842\n",
      "step: 2199 time 0:00:05.182208 train NMSE 0.22190 test NMSE 0.23438\n",
      "step: 2299 time 0:00:05.195415 train NMSE 0.22044 test NMSE 0.21067\n",
      "step: 2399 time 0:00:05.197121 train NMSE 0.22031 test NMSE 0.22420\n",
      "step: 2499 time 0:00:05.195130 train NMSE 0.22484 test NMSE 0.22347\n",
      "step: 2599 time 0:00:05.196109 train NMSE 0.21824 test NMSE 0.21034\n",
      "step: 2699 time 0:00:05.213086 train NMSE 0.21897 test NMSE 0.21559\n",
      "step: 2799 time 0:00:05.210062 train NMSE 0.21816 test NMSE 0.21266\n",
      "step: 2899 time 0:00:05.204076 train NMSE 0.21919 test NMSE 0.23828\n",
      "step: 2999 time 0:00:05.204077 train NMSE 0.21888 test NMSE 0.22666\n"
     ]
    }
   ],
   "source": [
    "L  = 8  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:04.600760 train NMSE 0.87585 test NMSE 0.85593\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.506629 train NMSE 0.77393 test NMSE 0.72020\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.492539 train NMSE 0.68227 test NMSE 0.65930\n",
      "Model saved!\n",
      "step: 399 time 0:00:04.514949 train NMSE 0.62027 test NMSE 0.59680\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.529095 train NMSE 0.58431 test NMSE 0.56634\n",
      "Model saved!\n",
      "step: 599 time 0:00:04.553711 train NMSE 0.55803 test NMSE 0.57227\n",
      "step: 699 time 0:00:04.544385 train NMSE 0.53645 test NMSE 0.52696\n",
      "Model saved!\n",
      "step: 799 time 0:00:04.557846 train NMSE 0.51707 test NMSE 0.50179\n",
      "Model saved!\n",
      "step: 899 time 0:00:04.528928 train NMSE 0.50771 test NMSE 0.48460\n",
      "Model saved!\n",
      "step: 999 time 0:00:04.642258 train NMSE 0.48943 test NMSE 0.50364\n",
      "step: 1099 time 0:00:04.579769 train NMSE 0.48001 test NMSE 0.45559\n",
      "Model saved!\n",
      "step: 1199 time 0:00:04.653627 train NMSE 0.47004 test NMSE 0.47633\n",
      "step: 1299 time 0:00:04.663640 train NMSE 0.46220 test NMSE 0.44282\n",
      "Model saved!\n",
      "step: 1399 time 0:00:04.649142 train NMSE 0.45317 test NMSE 0.43710\n",
      "Model saved!\n",
      "step: 1499 time 0:00:04.608694 train NMSE 0.44468 test NMSE 0.43437\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.650620 train NMSE 0.43924 test NMSE 0.43602\n",
      "step: 1699 time 0:00:04.632656 train NMSE 0.43337 test NMSE 0.44702\n",
      "step: 1799 time 0:00:04.663798 train NMSE 0.42682 test NMSE 0.42632\n",
      "Model saved!\n",
      "step: 1899 time 0:00:04.663951 train NMSE 0.42093 test NMSE 0.41958\n",
      "Model saved!\n",
      "step: 1999 time 0:00:04.684827 train NMSE 0.41329 test NMSE 0.41367\n",
      "Model saved!\n",
      "step: 2099 time 0:00:04.691311 train NMSE 0.40634 test NMSE 0.40848\n",
      "Model saved!\n",
      "step: 2199 time 0:00:04.660142 train NMSE 0.39935 test NMSE 0.37988\n",
      "Model saved!\n",
      "step: 2299 time 0:00:04.682208 train NMSE 0.39015 test NMSE 0.40996\n",
      "step: 2399 time 0:00:04.693045 train NMSE 0.38527 test NMSE 0.38865\n",
      "step: 2499 time 0:00:04.702040 train NMSE 0.37877 test NMSE 0.38760\n",
      "step: 2599 time 0:00:04.705571 train NMSE 0.37615 test NMSE 0.36782\n",
      "Model saved!\n",
      "step: 2699 time 0:00:04.725504 train NMSE 0.36798 test NMSE 0.37136\n",
      "step: 2799 time 0:00:04.675093 train NMSE 0.35795 test NMSE 0.33966\n",
      "Model saved!\n",
      "step: 2899 time 0:00:04.615192 train NMSE 0.35102 test NMSE 0.32884\n",
      "Model saved!\n",
      "step: 2999 time 0:00:04.683494 train NMSE 0.34539 test NMSE 0.34529\n",
      "step: 99 time 0:00:05.458928 train NMSE 0.35498 test NMSE 0.34598\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.406099 train NMSE 0.32601 test NMSE 0.31115\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.431390 train NMSE 0.29894 test NMSE 0.28402\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.408592 train NMSE 0.28495 test NMSE 0.28775\n",
      "step: 499 time 0:00:05.390634 train NMSE 0.27502 test NMSE 0.28143\n",
      "Model saved!\n",
      "step: 599 time 0:00:05.417174 train NMSE 0.26438 test NMSE 0.26355\n",
      "Model saved!\n",
      "step: 699 time 0:00:05.407594 train NMSE 0.26009 test NMSE 0.25539\n",
      "Model saved!\n",
      "step: 799 time 0:00:05.406586 train NMSE 0.25222 test NMSE 0.26179\n",
      "step: 899 time 0:00:05.420538 train NMSE 0.24665 test NMSE 0.24364\n",
      "Model saved!\n",
      "step: 999 time 0:00:05.423558 train NMSE 0.24323 test NMSE 0.23866\n",
      "Model saved!\n",
      "step: 1099 time 0:00:05.419594 train NMSE 0.23318 test NMSE 0.23665\n",
      "Model saved!\n",
      "step: 1199 time 0:00:05.417549 train NMSE 0.22512 test NMSE 0.19676\n",
      "Model saved!\n",
      "step: 1299 time 0:00:05.423729 train NMSE 0.21807 test NMSE 0.20716\n",
      "step: 1399 time 0:00:05.419614 train NMSE 0.20453 test NMSE 0.20234\n",
      "step: 1499 time 0:00:05.406609 train NMSE 0.19540 test NMSE 0.19757\n",
      "step: 1599 time 0:00:05.433048 train NMSE 0.19158 test NMSE 0.18765\n",
      "Model saved!\n",
      "step: 1699 time 0:00:05.422969 train NMSE 0.18764 test NMSE 0.18999\n",
      "step: 1799 time 0:00:05.471944 train NMSE 0.18300 test NMSE 0.16655\n",
      "Model saved!\n",
      "step: 1899 time 0:00:05.477568 train NMSE 0.17520 test NMSE 0.16444\n",
      "Model saved!\n",
      "step: 1999 time 0:00:05.505578 train NMSE 0.17088 test NMSE 0.18636\n",
      "step: 2099 time 0:00:05.502635 train NMSE 0.16680 test NMSE 0.16942\n",
      "step: 2199 time 0:00:05.473936 train NMSE 0.16200 test NMSE 0.16001\n",
      "Model saved!\n",
      "step: 2299 time 0:00:05.459981 train NMSE 0.16203 test NMSE 0.16194\n",
      "step: 2399 time 0:00:05.457517 train NMSE 0.15787 test NMSE 0.15362\n",
      "Model saved!\n",
      "step: 2499 time 0:00:05.453742 train NMSE 0.15735 test NMSE 0.16034\n",
      "step: 2599 time 0:00:05.443403 train NMSE 0.15565 test NMSE 0.15320\n",
      "Model saved!\n",
      "step: 2699 time 0:00:05.447186 train NMSE 0.15252 test NMSE 0.15714\n",
      "step: 2799 time 0:00:05.488439 train NMSE 0.15307 test NMSE 0.15800\n",
      "step: 2899 time 0:00:05.441746 train NMSE 0.15084 test NMSE 0.15684\n",
      "step: 2999 time 0:00:05.403106 train NMSE 0.15021 test NMSE 0.14062\n",
      "Model saved!\n",
      "step: 99 time 0:00:06.209466 train NMSE 0.17699 test NMSE 0.17970\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.255337 train NMSE 0.16136 test NMSE 0.15414\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.207445 train NMSE 0.15377 test NMSE 0.15819\n",
      "step: 399 time 0:00:06.207476 train NMSE 0.14877 test NMSE 0.14976\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.242439 train NMSE 0.14583 test NMSE 0.14282\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.211452 train NMSE 0.14428 test NMSE 0.12762\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.205121 train NMSE 0.14346 test NMSE 0.15123\n",
      "step: 799 time 0:00:06.231067 train NMSE 0.14094 test NMSE 0.14459\n",
      "step: 899 time 0:00:06.216019 train NMSE 0.14167 test NMSE 0.12954\n",
      "step: 999 time 0:00:06.189515 train NMSE 0.14097 test NMSE 0.14223\n",
      "step: 1099 time 0:00:06.195993 train NMSE 0.13788 test NMSE 0.12650\n",
      "Model saved!\n",
      "step: 1199 time 0:00:06.214454 train NMSE 0.13724 test NMSE 0.12975\n",
      "step: 1299 time 0:00:06.218427 train NMSE 0.13766 test NMSE 0.12816\n",
      "step: 1399 time 0:00:06.211331 train NMSE 0.13695 test NMSE 0.13607\n",
      "step: 1499 time 0:00:06.229445 train NMSE 0.13621 test NMSE 0.12987\n",
      "step: 1599 time 0:00:06.246024 train NMSE 0.13551 test NMSE 0.14067\n",
      "step: 1699 time 0:00:06.220369 train NMSE 0.13466 test NMSE 0.13744\n",
      "step: 1799 time 0:00:06.214749 train NMSE 0.13378 test NMSE 0.13558\n",
      "step: 1899 time 0:00:06.218402 train NMSE 0.13265 test NMSE 0.14286\n",
      "step: 1999 time 0:00:06.224960 train NMSE 0.13289 test NMSE 0.13160\n",
      "step: 2099 time 0:00:06.208961 train NMSE 0.13247 test NMSE 0.12440\n",
      "Model saved!\n",
      "step: 2199 time 0:00:06.212432 train NMSE 0.13200 test NMSE 0.12996\n",
      "step: 2299 time 0:00:06.227080 train NMSE 0.13009 test NMSE 0.12702\n",
      "step: 2399 time 0:00:06.201915 train NMSE 0.13039 test NMSE 0.12669\n",
      "step: 2499 time 0:00:06.213000 train NMSE 0.12898 test NMSE 0.12791\n",
      "step: 2599 time 0:00:06.197950 train NMSE 0.12827 test NMSE 0.13664\n",
      "step: 2699 time 0:00:06.216925 train NMSE 0.12863 test NMSE 0.13098\n",
      "step: 2799 time 0:00:06.197045 train NMSE 0.13068 test NMSE 0.12824\n",
      "step: 2899 time 0:00:06.201503 train NMSE 0.12768 test NMSE 0.12479\n",
      "step: 2999 time 0:00:06.219424 train NMSE 0.12829 test NMSE 0.11710\n",
      "Model saved!\n",
      "step: 99 time 0:00:06.965376 train NMSE 0.15666 test NMSE 0.15417\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.975402 train NMSE 0.14525 test NMSE 0.14315\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.971900 train NMSE 0.13752 test NMSE 0.14167\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.964426 train NMSE 0.13715 test NMSE 0.13556\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.962393 train NMSE 0.13578 test NMSE 0.12984\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.969400 train NMSE 0.13415 test NMSE 0.13773\n",
      "step: 699 time 0:00:06.983390 train NMSE 0.13353 test NMSE 0.13415\n",
      "step: 799 time 0:00:06.962918 train NMSE 0.13194 test NMSE 0.13914\n",
      "step: 899 time 0:00:06.974387 train NMSE 0.13328 test NMSE 0.14116\n",
      "step: 999 time 0:00:06.972072 train NMSE 0.13171 test NMSE 0.11423\n",
      "Model saved!\n",
      "step: 1099 time 0:00:06.946466 train NMSE 0.13186 test NMSE 0.12142\n",
      "step: 1199 time 0:00:06.974416 train NMSE 0.12945 test NMSE 0.12958\n",
      "step: 1299 time 0:00:06.978874 train NMSE 0.12915 test NMSE 0.11975\n",
      "step: 1399 time 0:00:06.967256 train NMSE 0.13002 test NMSE 0.12316\n",
      "step: 1499 time 0:00:06.966464 train NMSE 0.12915 test NMSE 0.13743\n",
      "step: 1599 time 0:00:06.980428 train NMSE 0.12755 test NMSE 0.13209\n",
      "step: 1699 time 0:00:06.976898 train NMSE 0.12813 test NMSE 0.12001\n",
      "step: 1799 time 0:00:06.976399 train NMSE 0.12810 test NMSE 0.12559\n",
      "step: 1899 time 0:00:06.978331 train NMSE 0.12669 test NMSE 0.12907\n",
      "step: 1999 time 0:00:06.979385 train NMSE 0.12722 test NMSE 0.10975\n",
      "Model saved!\n",
      "step: 2099 time 0:00:07.017269 train NMSE 0.12723 test NMSE 0.13499\n",
      "step: 2199 time 0:00:06.970386 train NMSE 0.12696 test NMSE 0.12623\n",
      "step: 2299 time 0:00:06.967395 train NMSE 0.12601 test NMSE 0.13170\n",
      "step: 2399 time 0:00:06.986854 train NMSE 0.12658 test NMSE 0.11859\n",
      "step: 2499 time 0:00:06.961859 train NMSE 0.12455 test NMSE 0.12095\n",
      "step: 2599 time 0:00:06.976833 train NMSE 0.12554 test NMSE 0.12068\n",
      "step: 2699 time 0:00:06.968078 train NMSE 0.12380 test NMSE 0.12038\n",
      "step: 2799 time 0:00:06.980954 train NMSE 0.12347 test NMSE 0.12317\n",
      "step: 2899 time 0:00:06.984972 train NMSE 0.12275 test NMSE 0.12265\n",
      "step: 2999 time 0:00:06.984020 train NMSE 0.12228 test NMSE 0.12358\n",
      "step: 99 time 0:00:07.775798 train NMSE 0.15142 test NMSE 0.15201\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.772232 train NMSE 0.14078 test NMSE 0.14183\n",
      "Model saved!\n",
      "step: 299 time 0:00:07.774791 train NMSE 0.13499 test NMSE 0.13536\n",
      "Model saved!\n",
      "step: 399 time 0:00:07.758316 train NMSE 0.13272 test NMSE 0.13119\n",
      "Model saved!\n",
      "step: 499 time 0:00:07.827469 train NMSE 0.13089 test NMSE 0.12964\n",
      "Model saved!\n",
      "step: 599 time 0:00:07.797555 train NMSE 0.13510 test NMSE 0.14647\n",
      "step: 699 time 0:00:07.830590 train NMSE 0.12816 test NMSE 0.12950\n",
      "Model saved!\n",
      "step: 799 time 0:00:07.783641 train NMSE 0.12742 test NMSE 0.13091\n",
      "step: 899 time 0:00:07.760325 train NMSE 0.12739 test NMSE 0.12537\n",
      "Model saved!\n",
      "step: 999 time 0:00:07.805209 train NMSE 0.12696 test NMSE 0.11243\n",
      "Model saved!\n",
      "step: 1099 time 0:00:07.919516 train NMSE 0.12521 test NMSE 0.12407\n",
      "step: 1199 time 0:00:07.732876 train NMSE 0.12541 test NMSE 0.12648\n",
      "step: 1299 time 0:00:07.778412 train NMSE 0.12490 test NMSE 0.13245\n",
      "step: 1399 time 0:00:07.735905 train NMSE 0.12498 test NMSE 0.11310\n",
      "step: 1499 time 0:00:07.752808 train NMSE 0.12410 test NMSE 0.13081\n",
      "step: 1599 time 0:00:07.741384 train NMSE 0.12503 test NMSE 0.12877\n",
      "step: 1699 time 0:00:07.740914 train NMSE 0.12369 test NMSE 0.12271\n",
      "step: 1799 time 0:00:07.740237 train NMSE 0.12366 test NMSE 0.12525\n",
      "step: 1899 time 0:00:07.755847 train NMSE 0.12325 test NMSE 0.11427\n",
      "step: 1999 time 0:00:07.741345 train NMSE 0.12313 test NMSE 0.12184\n",
      "step: 2099 time 0:00:07.741830 train NMSE 0.12170 test NMSE 0.12130\n",
      "step: 2199 time 0:00:07.749352 train NMSE 0.12301 test NMSE 0.12392\n",
      "step: 2299 time 0:00:07.750417 train NMSE 0.12156 test NMSE 0.12898\n",
      "step: 2399 time 0:00:07.745588 train NMSE 0.12196 test NMSE 0.11987\n",
      "step: 2499 time 0:00:07.748982 train NMSE 0.12181 test NMSE 0.11627\n",
      "step: 2599 time 0:00:07.749403 train NMSE 0.12096 test NMSE 0.10799\n",
      "Model saved!\n",
      "step: 2699 time 0:00:07.749852 train NMSE 0.12148 test NMSE 0.12166\n",
      "step: 2799 time 0:00:07.749378 train NMSE 0.12094 test NMSE 0.12543\n",
      "step: 2899 time 0:00:07.775264 train NMSE 0.12048 test NMSE 0.12086\n",
      "step: 2999 time 0:00:07.740249 train NMSE 0.12055 test NMSE 0.12801\n"
     ]
    }
   ],
   "source": [
    "L  = 16  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:07.677506 train NMSE 0.59002 test NMSE 0.44429\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.652954 train NMSE 0.38871 test NMSE 0.33578\n",
      "Model saved!\n",
      "step: 299 time 0:00:07.736872 train NMSE 0.30523 test NMSE 0.28776\n",
      "Model saved!\n",
      "step: 399 time 0:00:07.681484 train NMSE 0.26723 test NMSE 0.26172\n",
      "Model saved!\n",
      "step: 499 time 0:00:07.661174 train NMSE 0.24056 test NMSE 0.23530\n",
      "Model saved!\n",
      "step: 599 time 0:00:07.669872 train NMSE 0.21825 test NMSE 0.20245\n",
      "Model saved!\n",
      "step: 699 time 0:00:07.709923 train NMSE 0.20254 test NMSE 0.20055\n",
      "Model saved!\n",
      "step: 799 time 0:00:07.648065 train NMSE 0.19189 test NMSE 0.17669\n",
      "Model saved!\n",
      "step: 899 time 0:00:07.662649 train NMSE 0.18295 test NMSE 0.18618\n",
      "step: 999 time 0:00:07.677504 train NMSE 0.17436 test NMSE 0.18004\n",
      "step: 1099 time 0:00:07.659565 train NMSE 0.16568 test NMSE 0.15169\n",
      "Model saved!\n",
      "step: 1199 time 0:00:07.677528 train NMSE 0.15845 test NMSE 0.16320\n",
      "step: 1299 time 0:00:07.675433 train NMSE 0.15396 test NMSE 0.14672\n",
      "Model saved!\n",
      "step: 1399 time 0:00:07.651711 train NMSE 0.14886 test NMSE 0.15428\n",
      "step: 1499 time 0:00:07.681530 train NMSE 0.14418 test NMSE 0.14983\n",
      "step: 1599 time 0:00:07.665573 train NMSE 0.14271 test NMSE 0.13951\n",
      "Model saved!\n",
      "step: 1699 time 0:00:07.661244 train NMSE 0.14059 test NMSE 0.14865\n",
      "step: 1799 time 0:00:07.691866 train NMSE 0.13773 test NMSE 0.14591\n",
      "step: 1899 time 0:00:07.681838 train NMSE 0.13532 test NMSE 0.12681\n",
      "Model saved!\n",
      "step: 1999 time 0:00:07.667564 train NMSE 0.13461 test NMSE 0.13719\n",
      "step: 2099 time 0:00:07.696259 train NMSE 0.13115 test NMSE 0.13182\n",
      "step: 2199 time 0:00:07.683576 train NMSE 0.12648 test NMSE 0.12327\n",
      "Model saved!\n",
      "step: 2299 time 0:00:07.687489 train NMSE 0.12513 test NMSE 0.12995\n",
      "step: 2399 time 0:00:07.693493 train NMSE 0.12066 test NMSE 0.11764\n",
      "Model saved!\n",
      "step: 2499 time 0:00:07.703962 train NMSE 0.11801 test NMSE 0.10690\n",
      "Model saved!\n",
      "step: 2599 time 0:00:07.652120 train NMSE 0.11322 test NMSE 0.11386\n",
      "step: 2699 time 0:00:07.691758 train NMSE 0.10776 test NMSE 0.09529\n",
      "Model saved!\n",
      "step: 2799 time 0:00:07.640883 train NMSE 0.10128 test NMSE 0.09251\n",
      "Model saved!\n",
      "step: 2899 time 0:00:07.644839 train NMSE 0.09660 test NMSE 0.08876\n",
      "Model saved!\n",
      "step: 2999 time 0:00:07.635156 train NMSE 0.09062 test NMSE 0.08906\n",
      "step: 99 time 0:00:08.969231 train NMSE 0.13038 test NMSE 0.10379\n",
      "Model saved!\n",
      "step: 199 time 0:00:08.982321 train NMSE 0.09639 test NMSE 0.09180\n",
      "Model saved!\n",
      "step: 299 time 0:00:08.983961 train NMSE 0.08519 test NMSE 0.08320\n",
      "Model saved!\n",
      "step: 399 time 0:00:08.983935 train NMSE 0.07994 test NMSE 0.07621\n",
      "Model saved!\n",
      "step: 499 time 0:00:08.968727 train NMSE 0.07717 test NMSE 0.08312\n",
      "step: 599 time 0:00:08.983081 train NMSE 0.07499 test NMSE 0.07608\n",
      "Model saved!\n",
      "step: 699 time 0:00:08.966126 train NMSE 0.07310 test NMSE 0.06867\n",
      "Model saved!\n",
      "step: 799 time 0:00:08.988085 train NMSE 0.07160 test NMSE 0.07552\n",
      "step: 899 time 0:00:08.996103 train NMSE 0.07016 test NMSE 0.06864\n",
      "Model saved!\n",
      "step: 999 time 0:00:08.997117 train NMSE 0.06953 test NMSE 0.06913\n",
      "step: 1099 time 0:00:08.975118 train NMSE 0.06782 test NMSE 0.06363\n",
      "Model saved!\n",
      "step: 1199 time 0:00:08.995544 train NMSE 0.06705 test NMSE 0.06342\n",
      "Model saved!\n",
      "step: 1299 time 0:00:09.003067 train NMSE 0.06590 test NMSE 0.06863\n",
      "step: 1399 time 0:00:08.988077 train NMSE 0.06519 test NMSE 0.06519\n",
      "step: 1499 time 0:00:08.988629 train NMSE 0.06418 test NMSE 0.06187\n",
      "Model saved!\n",
      "step: 1599 time 0:00:08.996047 train NMSE 0.06200 test NMSE 0.06433\n",
      "step: 1699 time 0:00:08.996540 train NMSE 0.06059 test NMSE 0.05712\n",
      "Model saved!\n",
      "step: 1799 time 0:00:09.001034 train NMSE 0.05880 test NMSE 0.06099\n",
      "step: 1899 time 0:00:09.265886 train NMSE 0.05793 test NMSE 0.05868\n",
      "step: 1999 time 0:00:09.154571 train NMSE 0.05691 test NMSE 0.05916\n",
      "step: 2099 time 0:00:09.085177 train NMSE 0.05609 test NMSE 0.05629\n",
      "Model saved!\n",
      "step: 2199 time 0:00:09.099628 train NMSE 0.05514 test NMSE 0.05733\n",
      "step: 2299 time 0:00:09.081289 train NMSE 0.05458 test NMSE 0.04994\n",
      "Model saved!\n",
      "step: 2399 time 0:00:09.032359 train NMSE 0.05366 test NMSE 0.05137\n",
      "step: 2499 time 0:00:09.056819 train NMSE 0.05253 test NMSE 0.05286\n",
      "step: 2599 time 0:00:09.061352 train NMSE 0.05215 test NMSE 0.05010\n",
      "step: 2699 time 0:00:09.059870 train NMSE 0.05181 test NMSE 0.05211\n",
      "step: 2799 time 0:00:09.053808 train NMSE 0.05095 test NMSE 0.04914\n",
      "Model saved!\n",
      "step: 2899 time 0:00:09.053862 train NMSE 0.05113 test NMSE 0.05219\n",
      "step: 2999 time 0:00:09.060833 train NMSE 0.05043 test NMSE 0.05077\n",
      "step: 99 time 0:00:10.307013 train NMSE 0.06063 test NMSE 0.05330\n",
      "Model saved!\n",
      "step: 199 time 0:00:10.315001 train NMSE 0.05300 test NMSE 0.05007\n",
      "Model saved!\n",
      "step: 299 time 0:00:10.329436 train NMSE 0.04985 test NMSE 0.04984\n",
      "Model saved!\n",
      "step: 399 time 0:00:10.330430 train NMSE 0.04829 test NMSE 0.05026\n",
      "step: 499 time 0:00:10.314460 train NMSE 0.05988 test NMSE 0.04968\n",
      "Model saved!\n",
      "step: 599 time 0:00:10.313482 train NMSE 0.06342 test NMSE 0.04629\n",
      "Model saved!\n",
      "step: 699 time 0:00:10.310931 train NMSE 0.04723 test NMSE 0.04652\n",
      "step: 799 time 0:00:10.326906 train NMSE 0.04627 test NMSE 0.04563\n",
      "Model saved!\n",
      "step: 899 time 0:00:10.308483 train NMSE 0.06261 test NMSE 0.04458\n",
      "Model saved!\n",
      "step: 999 time 0:00:10.319506 train NMSE 0.04573 test NMSE 0.04477\n",
      "step: 1099 time 0:00:10.330424 train NMSE 0.04746 test NMSE 0.04548\n",
      "step: 1199 time 0:00:10.318040 train NMSE 0.04553 test NMSE 0.04443\n",
      "Model saved!\n",
      "step: 1299 time 0:00:10.335423 train NMSE 0.04550 test NMSE 0.04566\n",
      "step: 1399 time 0:00:10.329471 train NMSE 0.04495 test NMSE 0.04498\n",
      "step: 1499 time 0:00:10.341452 train NMSE 0.05447 test NMSE 0.04571\n",
      "step: 1599 time 0:00:10.313463 train NMSE 0.04483 test NMSE 0.04374\n",
      "Model saved!\n",
      "step: 1699 time 0:00:10.305014 train NMSE 0.04498 test NMSE 0.04412\n",
      "step: 1799 time 0:00:10.317963 train NMSE 0.04485 test NMSE 0.04489\n",
      "step: 1899 time 0:00:10.331463 train NMSE 0.05685 test NMSE 0.04190\n",
      "Model saved!\n",
      "step: 1999 time 0:00:10.321932 train NMSE 0.04468 test NMSE 0.04310\n",
      "step: 2099 time 0:00:10.337387 train NMSE 0.04458 test NMSE 0.04527\n",
      "step: 2199 time 0:00:10.347424 train NMSE 0.04654 test NMSE 0.04326\n",
      "step: 2299 time 0:00:10.372662 train NMSE 0.05198 test NMSE 0.04481\n",
      "step: 2399 time 0:00:10.330233 train NMSE 0.04435 test NMSE 0.04513\n",
      "step: 2499 time 0:00:10.330409 train NMSE 0.04472 test NMSE 0.04630\n",
      "step: 2599 time 0:00:10.321308 train NMSE 0.04415 test NMSE 0.04338\n",
      "step: 2699 time 0:00:10.334884 train NMSE 0.05163 test NMSE 0.04472\n",
      "step: 2799 time 0:00:10.320064 train NMSE 0.04419 test NMSE 0.04518\n",
      "step: 2899 time 0:00:10.319907 train NMSE 0.04422 test NMSE 0.04412\n",
      "step: 2999 time 0:00:10.320492 train NMSE 0.04401 test NMSE 0.04440\n",
      "step: 99 time 0:00:11.651279 train NMSE 0.04595 test NMSE 0.04375\n",
      "Model saved!\n",
      "step: 199 time 0:00:11.673843 train NMSE 0.04186 test NMSE 0.04053\n",
      "Model saved!\n",
      "step: 299 time 0:00:11.630259 train NMSE 0.04041 test NMSE 0.04120\n",
      "step: 399 time 0:00:11.664904 train NMSE 0.04370 test NMSE 0.04037\n",
      "Model saved!\n",
      "step: 499 time 0:00:11.655288 train NMSE 0.03904 test NMSE 0.03926\n",
      "Model saved!\n",
      "step: 599 time 0:00:11.652527 train NMSE 0.03909 test NMSE 0.03755\n",
      "Model saved!\n",
      "step: 699 time 0:00:11.653186 train NMSE 0.04045 test NMSE 0.03955\n",
      "step: 799 time 0:00:11.652635 train NMSE 0.03856 test NMSE 0.03815\n",
      "step: 899 time 0:00:11.657793 train NMSE 0.03814 test NMSE 0.03708\n",
      "Model saved!\n",
      "step: 999 time 0:00:11.666388 train NMSE 0.03813 test NMSE 0.03768\n",
      "step: 1099 time 0:00:11.663780 train NMSE 0.03906 test NMSE 0.03731\n",
      "step: 1199 time 0:00:11.653045 train NMSE 0.03791 test NMSE 0.03815\n",
      "step: 1299 time 0:00:11.654629 train NMSE 0.03783 test NMSE 0.03936\n",
      "step: 1399 time 0:00:11.648138 train NMSE 0.03765 test NMSE 0.03877\n",
      "step: 1499 time 0:00:11.663960 train NMSE 0.03768 test NMSE 0.03699\n",
      "Model saved!\n",
      "step: 1599 time 0:00:11.685222 train NMSE 0.03780 test NMSE 0.03670\n",
      "Model saved!\n",
      "step: 1699 time 0:00:11.657576 train NMSE 0.03787 test NMSE 0.03777\n",
      "step: 1799 time 0:00:11.679964 train NMSE 0.03755 test NMSE 0.03772\n",
      "step: 1899 time 0:00:11.684142 train NMSE 0.03758 test NMSE 0.03760\n",
      "step: 1999 time 0:00:11.677007 train NMSE 0.03748 test NMSE 0.03712\n",
      "step: 2099 time 0:00:11.634906 train NMSE 0.03755 test NMSE 0.03707\n",
      "step: 2199 time 0:00:11.658908 train NMSE 0.03732 test NMSE 0.03986\n",
      "step: 2299 time 0:00:11.656185 train NMSE 0.03728 test NMSE 0.03609\n",
      "Model saved!\n",
      "step: 2399 time 0:00:11.669381 train NMSE 0.03742 test NMSE 0.03544\n",
      "Model saved!\n",
      "step: 2499 time 0:00:11.670404 train NMSE 0.03786 test NMSE 0.03705\n",
      "step: 2599 time 0:00:11.675829 train NMSE 0.03727 test NMSE 0.03717\n",
      "step: 2699 time 0:00:11.663327 train NMSE 0.03703 test NMSE 0.03771\n",
      "step: 2799 time 0:00:11.652884 train NMSE 0.03719 test NMSE 0.03804\n",
      "step: 2899 time 0:00:11.681721 train NMSE 0.04061 test NMSE 0.03785\n",
      "step: 2999 time 0:00:11.641482 train NMSE 0.03684 test NMSE 0.03486\n",
      "Model saved!\n",
      "step: 99 time 0:00:12.961595 train NMSE 0.03929 test NMSE 0.03631\n",
      "Model saved!\n",
      "step: 199 time 0:00:12.985468 train NMSE 0.03664 test NMSE 0.03539\n",
      "Model saved!\n",
      "step: 299 time 0:00:13.008950 train NMSE 0.05682 test NMSE 0.03632\n",
      "step: 399 time 0:00:12.956787 train NMSE 0.03512 test NMSE 0.03452\n",
      "Model saved!\n",
      "step: 499 time 0:00:12.962892 train NMSE 0.03437 test NMSE 0.03561\n",
      "step: 599 time 0:00:12.977373 train NMSE 0.04861 test NMSE 0.03361\n",
      "Model saved!\n",
      "step: 699 time 0:00:12.969889 train NMSE 0.03404 test NMSE 0.03270\n",
      "Model saved!\n",
      "step: 799 time 0:00:12.966854 train NMSE 0.03396 test NMSE 0.03200\n",
      "Model saved!\n",
      "step: 899 time 0:00:12.967464 train NMSE 0.03386 test NMSE 0.03431\n",
      "step: 999 time 0:00:12.962142 train NMSE 0.03379 test NMSE 0.03380\n",
      "step: 1099 time 0:00:12.980471 train NMSE 0.03378 test NMSE 0.03362\n",
      "step: 1199 time 0:00:12.965499 train NMSE 0.03368 test NMSE 0.03361\n",
      "step: 1299 time 0:00:12.965242 train NMSE 0.03396 test NMSE 0.03459\n",
      "step: 1399 time 0:00:12.966138 train NMSE 0.06541 test NMSE 0.03326\n",
      "step: 1499 time 0:00:12.963403 train NMSE 0.04118 test NMSE 0.03289\n",
      "step: 1599 time 0:00:12.970195 train NMSE 0.03555 test NMSE 0.03316\n",
      "step: 1699 time 0:00:12.960556 train NMSE 0.03361 test NMSE 0.03359\n",
      "step: 1799 time 0:00:12.974327 train NMSE 0.03350 test NMSE 0.03319\n",
      "step: 1899 time 0:00:12.991910 train NMSE 0.03352 test NMSE 0.03297\n",
      "step: 1999 time 0:00:12.995811 train NMSE 0.06001 test NMSE 0.03341\n",
      "step: 2099 time 0:00:12.991759 train NMSE 0.03378 test NMSE 0.03415\n",
      "step: 2199 time 0:00:12.978449 train NMSE 0.03337 test NMSE 0.03231\n",
      "step: 2299 time 0:00:12.989260 train NMSE 0.03355 test NMSE 0.03466\n",
      "step: 2399 time 0:00:12.983682 train NMSE 0.03391 test NMSE 0.03484\n",
      "step: 2499 time 0:00:12.998833 train NMSE 0.03587 test NMSE 0.03320\n",
      "step: 2599 time 0:00:12.986323 train NMSE 0.03380 test NMSE 0.03269\n",
      "step: 2699 time 0:00:12.998161 train NMSE 0.04300 test NMSE 0.03272\n",
      "step: 2799 time 0:00:12.977508 train NMSE 0.03510 test NMSE 0.03405\n",
      "step: 2899 time 0:00:12.992744 train NMSE 0.03342 test NMSE 0.03404\n",
      "step: 2999 time 0:00:12.974938 train NMSE 0.03320 test NMSE 0.03302\n"
     ]
    }
   ],
   "source": [
    "L  = 32  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:13.988841 train NMSE 0.22225 test NMSE 0.15287\n",
      "Model saved!\n",
      "step: 199 time 0:00:13.876553 train NMSE 0.13638 test NMSE 0.12155\n",
      "Model saved!\n",
      "step: 299 time 0:00:13.917938 train NMSE 0.10989 test NMSE 0.10457\n",
      "Model saved!\n",
      "step: 399 time 0:00:13.974770 train NMSE 0.09774 test NMSE 0.08955\n",
      "Model saved!\n",
      "step: 499 time 0:00:13.937353 train NMSE 0.08947 test NMSE 0.09227\n",
      "step: 599 time 0:00:13.915989 train NMSE 0.08534 test NMSE 0.07588\n",
      "Model saved!\n",
      "step: 699 time 0:00:13.888962 train NMSE 0.08117 test NMSE 0.08396\n",
      "step: 799 time 0:00:13.926921 train NMSE 0.07895 test NMSE 0.07348\n",
      "Model saved!\n",
      "step: 899 time 0:00:13.863019 train NMSE 0.07649 test NMSE 0.07576\n",
      "step: 999 time 0:00:13.876021 train NMSE 0.07530 test NMSE 0.07103\n",
      "Model saved!\n",
      "step: 1099 time 0:00:13.920522 train NMSE 0.07297 test NMSE 0.07798\n",
      "step: 1199 time 0:00:13.905488 train NMSE 0.07215 test NMSE 0.07288\n",
      "step: 1299 time 0:00:13.878123 train NMSE 0.07073 test NMSE 0.07176\n",
      "step: 1399 time 0:00:13.891640 train NMSE 0.06933 test NMSE 0.06405\n",
      "Model saved!\n",
      "step: 1499 time 0:00:13.882031 train NMSE 0.06826 test NMSE 0.06484\n",
      "step: 1599 time 0:00:13.919959 train NMSE 0.06745 test NMSE 0.06497\n",
      "step: 1699 time 0:00:13.915502 train NMSE 0.06528 test NMSE 0.06567\n",
      "step: 1799 time 0:00:13.919883 train NMSE 0.06285 test NMSE 0.06138\n",
      "Model saved!\n",
      "step: 1899 time 0:00:13.934517 train NMSE 0.06065 test NMSE 0.05929\n",
      "Model saved!\n",
      "step: 1999 time 0:00:13.915908 train NMSE 0.05945 test NMSE 0.05745\n",
      "Model saved!\n",
      "step: 2099 time 0:00:13.946852 train NMSE 0.05778 test NMSE 0.05877\n",
      "step: 2199 time 0:00:13.937142 train NMSE 0.05707 test NMSE 0.05273\n",
      "Model saved!\n",
      "step: 2299 time 0:00:13.922833 train NMSE 0.05569 test NMSE 0.05491\n",
      "step: 2399 time 0:00:13.919846 train NMSE 0.05496 test NMSE 0.05208\n",
      "Model saved!\n",
      "step: 2499 time 0:00:13.933831 train NMSE 0.05416 test NMSE 0.05243\n",
      "step: 2599 time 0:00:13.914890 train NMSE 0.05450 test NMSE 0.05513\n",
      "step: 2699 time 0:00:13.923130 train NMSE 0.05386 test NMSE 0.05310\n",
      "step: 2799 time 0:00:13.937839 train NMSE 0.05318 test NMSE 0.05337\n",
      "step: 2899 time 0:00:13.944494 train NMSE 0.05274 test NMSE 0.05132\n",
      "Model saved!\n",
      "step: 2999 time 0:00:13.901077 train NMSE 0.05244 test NMSE 0.05161\n",
      "step: 99 time 0:00:16.406266 train NMSE 0.06582 test NMSE 0.05391\n",
      "Model saved!\n",
      "step: 199 time 0:00:16.374284 train NMSE 0.04985 test NMSE 0.04832\n",
      "Model saved!\n",
      "step: 299 time 0:00:16.387358 train NMSE 0.04675 test NMSE 0.04760\n",
      "Model saved!\n",
      "step: 399 time 0:00:16.418730 train NMSE 0.04503 test NMSE 0.04319\n",
      "Model saved!\n",
      "step: 499 time 0:00:16.448354 train NMSE 0.04437 test NMSE 0.04457\n",
      "step: 599 time 0:00:16.394901 train NMSE 0.04351 test NMSE 0.04389\n",
      "step: 699 time 0:00:16.436201 train NMSE 0.04337 test NMSE 0.04229\n",
      "Model saved!\n",
      "step: 799 time 0:00:16.406177 train NMSE 0.04230 test NMSE 0.04009\n",
      "Model saved!\n",
      "step: 899 time 0:00:16.407696 train NMSE 0.04214 test NMSE 0.04391\n",
      "step: 999 time 0:00:16.432869 train NMSE 0.04118 test NMSE 0.04063\n",
      "step: 1099 time 0:00:16.432767 train NMSE 0.03956 test NMSE 0.04022\n",
      "step: 1199 time 0:00:16.430892 train NMSE 0.03761 test NMSE 0.03618\n",
      "Model saved!\n",
      "step: 1299 time 0:00:16.428610 train NMSE 0.03552 test NMSE 0.03361\n",
      "Model saved!\n",
      "step: 1399 time 0:00:16.469937 train NMSE 0.03340 test NMSE 0.03203\n",
      "Model saved!\n",
      "step: 1499 time 0:00:16.449159 train NMSE 0.03130 test NMSE 0.03080\n",
      "Model saved!\n",
      "step: 1599 time 0:00:16.408372 train NMSE 0.02958 test NMSE 0.02914\n",
      "Model saved!\n",
      "step: 1699 time 0:00:16.423998 train NMSE 0.02841 test NMSE 0.02757\n",
      "Model saved!\n",
      "step: 1799 time 0:00:16.462655 train NMSE 0.02743 test NMSE 0.02597\n",
      "Model saved!\n",
      "step: 1899 time 0:00:16.420267 train NMSE 0.02661 test NMSE 0.02628\n",
      "step: 1999 time 0:00:16.457234 train NMSE 0.02570 test NMSE 0.02719\n",
      "step: 2099 time 0:00:16.454249 train NMSE 0.02509 test NMSE 0.02419\n",
      "Model saved!\n",
      "step: 2199 time 0:00:16.444696 train NMSE 0.02469 test NMSE 0.02495\n",
      "step: 2299 time 0:00:16.441950 train NMSE 0.02443 test NMSE 0.02527\n",
      "step: 2399 time 0:00:16.432220 train NMSE 0.02407 test NMSE 0.02489\n",
      "step: 2499 time 0:00:16.431306 train NMSE 0.02376 test NMSE 0.02491\n",
      "step: 2599 time 0:00:16.436698 train NMSE 0.02352 test NMSE 0.02385\n",
      "Model saved!\n",
      "step: 2699 time 0:00:16.420218 train NMSE 0.02329 test NMSE 0.02344\n",
      "Model saved!\n",
      "step: 2799 time 0:00:16.425944 train NMSE 0.02299 test NMSE 0.02362\n",
      "step: 2899 time 0:00:16.463110 train NMSE 0.02303 test NMSE 0.02172\n",
      "Model saved!\n",
      "step: 2999 time 0:00:16.470126 train NMSE 0.02289 test NMSE 0.02315\n",
      "step: 99 time 0:00:18.924009 train NMSE 0.03243 test NMSE 0.02873\n",
      "Model saved!\n",
      "step: 199 time 0:00:18.842423 train NMSE 0.02764 test NMSE 0.02721\n",
      "Model saved!\n",
      "step: 299 time 0:00:18.804363 train NMSE 0.02595 test NMSE 0.02561\n",
      "Model saved!\n",
      "step: 399 time 0:00:18.811826 train NMSE 0.02501 test NMSE 0.02442\n",
      "Model saved!\n",
      "step: 499 time 0:00:18.819412 train NMSE 0.02437 test NMSE 0.02538\n",
      "step: 599 time 0:00:18.796710 train NMSE 0.02391 test NMSE 0.02403\n",
      "Model saved!\n",
      "step: 699 time 0:00:18.816863 train NMSE 0.02325 test NMSE 0.02184\n",
      "Model saved!\n",
      "step: 799 time 0:00:18.813834 train NMSE 0.02193 test NMSE 0.02014\n",
      "Model saved!\n",
      "step: 899 time 0:00:18.871252 train NMSE 0.02099 test NMSE 0.02067\n",
      "step: 999 time 0:00:18.870895 train NMSE 0.02054 test NMSE 0.01959\n",
      "Model saved!\n",
      "step: 1099 time 0:00:18.763949 train NMSE 0.02022 test NMSE 0.01928\n",
      "Model saved!\n",
      "step: 1199 time 0:00:18.856725 train NMSE 0.02012 test NMSE 0.02039\n",
      "step: 1299 time 0:00:18.832768 train NMSE 0.01987 test NMSE 0.02010\n",
      "step: 1399 time 0:00:18.809870 train NMSE 0.01971 test NMSE 0.01928\n",
      "step: 1499 time 0:00:18.909608 train NMSE 0.01960 test NMSE 0.02046\n",
      "step: 1599 time 0:00:18.838852 train NMSE 0.01951 test NMSE 0.01909\n",
      "Model saved!\n",
      "step: 1699 time 0:00:18.815312 train NMSE 0.01950 test NMSE 0.01924\n",
      "step: 1799 time 0:00:18.874238 train NMSE 0.01948 test NMSE 0.01850\n",
      "Model saved!\n",
      "step: 1899 time 0:00:18.832895 train NMSE 0.01929 test NMSE 0.02081\n",
      "step: 1999 time 0:00:18.843792 train NMSE 0.01929 test NMSE 0.01786\n",
      "Model saved!\n",
      "step: 2099 time 0:00:18.810838 train NMSE 0.01913 test NMSE 0.01917\n",
      "step: 2199 time 0:00:18.814792 train NMSE 0.01921 test NMSE 0.01959\n",
      "step: 2299 time 0:00:18.885793 train NMSE 0.01921 test NMSE 0.01808\n",
      "step: 2399 time 0:00:18.891628 train NMSE 0.01906 test NMSE 0.01911\n",
      "step: 2499 time 0:00:18.884687 train NMSE 0.01910 test NMSE 0.02008\n",
      "step: 2599 time 0:00:18.856845 train NMSE 0.01909 test NMSE 0.02035\n",
      "step: 2699 time 0:00:18.860151 train NMSE 0.01904 test NMSE 0.01956\n",
      "step: 2799 time 0:00:18.919109 train NMSE 0.01916 test NMSE 0.01874\n",
      "step: 2899 time 0:00:18.886696 train NMSE 0.01908 test NMSE 0.01928\n",
      "step: 2999 time 0:00:18.886882 train NMSE 0.01915 test NMSE 0.01946\n",
      "step: 99 time 0:00:21.341174 train NMSE 0.02407 test NMSE 0.02256\n",
      "Model saved!\n",
      "step: 199 time 0:00:21.332099 train NMSE 0.02207 test NMSE 0.02238\n",
      "Model saved!\n",
      "step: 299 time 0:00:21.296278 train NMSE 0.02088 test NMSE 0.02145\n",
      "Model saved!\n",
      "step: 399 time 0:00:21.330406 train NMSE 0.02042 test NMSE 0.01968\n",
      "Model saved!\n",
      "step: 499 time 0:00:21.303729 train NMSE 0.02002 test NMSE 0.01991\n",
      "step: 599 time 0:00:21.379614 train NMSE 0.01944 test NMSE 0.01863\n",
      "Model saved!\n",
      "step: 699 time 0:00:21.353696 train NMSE 0.01871 test NMSE 0.01952\n",
      "step: 799 time 0:00:21.317823 train NMSE 0.01812 test NMSE 0.01790\n",
      "Model saved!\n",
      "step: 899 time 0:00:21.415940 train NMSE 0.01795 test NMSE 0.01797\n",
      "step: 999 time 0:00:21.326275 train NMSE 0.01771 test NMSE 0.01719\n",
      "Model saved!\n",
      "step: 1099 time 0:00:21.312210 train NMSE 0.01772 test NMSE 0.01728\n",
      "step: 1199 time 0:00:21.351155 train NMSE 0.01765 test NMSE 0.01839\n",
      "step: 1299 time 0:00:21.340541 train NMSE 0.01748 test NMSE 0.01726\n",
      "step: 1399 time 0:00:21.366083 train NMSE 0.01725 test NMSE 0.01831\n",
      "step: 1499 time 0:00:21.346094 train NMSE 0.01735 test NMSE 0.01712\n",
      "Model saved!\n",
      "step: 1599 time 0:00:21.312646 train NMSE 0.01738 test NMSE 0.01773\n",
      "step: 1699 time 0:00:21.353875 train NMSE 0.01727 test NMSE 0.01720\n",
      "step: 1799 time 0:00:21.356815 train NMSE 0.01719 test NMSE 0.01627\n",
      "Model saved!\n",
      "step: 1899 time 0:00:21.388206 train NMSE 0.01716 test NMSE 0.01702\n",
      "step: 1999 time 0:00:21.363605 train NMSE 0.01728 test NMSE 0.01870\n",
      "step: 2099 time 0:00:21.331083 train NMSE 0.01711 test NMSE 0.01684\n",
      "step: 2199 time 0:00:21.365587 train NMSE 0.01716 test NMSE 0.01779\n",
      "step: 2299 time 0:00:21.340962 train NMSE 0.01716 test NMSE 0.01826\n",
      "step: 2399 time 0:00:21.322259 train NMSE 0.01713 test NMSE 0.01723\n",
      "step: 2499 time 0:00:21.375281 train NMSE 0.01713 test NMSE 0.01692\n",
      "step: 2599 time 0:00:21.367287 train NMSE 0.01716 test NMSE 0.01726\n",
      "step: 2699 time 0:00:21.373891 train NMSE 0.01712 test NMSE 0.01762\n",
      "step: 2799 time 0:00:21.371526 train NMSE 0.01713 test NMSE 0.01735\n",
      "step: 2899 time 0:00:21.384972 train NMSE 0.01712 test NMSE 0.01797\n",
      "step: 2999 time 0:00:21.379350 train NMSE 0.01709 test NMSE 0.01680\n",
      "step: 99 time 0:00:23.874216 train NMSE 0.02131 test NMSE 0.01935\n",
      "Model saved!\n",
      "step: 199 time 0:00:23.800527 train NMSE 0.02006 test NMSE 0.01797\n",
      "Model saved!\n",
      "step: 299 time 0:00:23.863967 train NMSE 0.01941 test NMSE 0.01936\n",
      "step: 399 time 0:00:23.944660 train NMSE 0.01899 test NMSE 0.01830\n",
      "step: 499 time 0:00:23.869293 train NMSE 0.01842 test NMSE 0.01774\n",
      "Model saved!\n",
      "step: 599 time 0:00:23.822018 train NMSE 0.01798 test NMSE 0.01768\n",
      "Model saved!\n",
      "step: 699 time 0:00:23.823349 train NMSE 0.01742 test NMSE 0.01757\n",
      "Model saved!\n",
      "step: 799 time 0:00:23.881710 train NMSE 0.01706 test NMSE 0.01729\n",
      "Model saved!\n",
      "step: 899 time 0:00:23.778546 train NMSE 0.01691 test NMSE 0.01685\n",
      "Model saved!\n",
      "step: 999 time 0:00:23.820906 train NMSE 0.01697 test NMSE 0.01679\n",
      "Model saved!\n",
      "step: 1099 time 0:00:23.830941 train NMSE 0.01685 test NMSE 0.01709\n",
      "step: 1199 time 0:00:23.875756 train NMSE 0.01689 test NMSE 0.01683\n",
      "step: 1299 time 0:00:23.877664 train NMSE 0.01667 test NMSE 0.01668\n",
      "Model saved!\n",
      "step: 1399 time 0:00:23.858227 train NMSE 0.01671 test NMSE 0.01715\n",
      "step: 1499 time 0:00:23.873313 train NMSE 0.01668 test NMSE 0.01827\n",
      "step: 1599 time 0:00:23.879208 train NMSE 0.01670 test NMSE 0.01614\n",
      "Model saved!\n",
      "step: 1699 time 0:00:23.862298 train NMSE 0.01671 test NMSE 0.01598\n",
      "Model saved!\n",
      "step: 1799 time 0:00:23.859460 train NMSE 0.01660 test NMSE 0.01651\n",
      "step: 1899 time 0:00:23.893483 train NMSE 0.01674 test NMSE 0.01682\n",
      "step: 1999 time 0:00:23.871547 train NMSE 0.01668 test NMSE 0.01722\n",
      "step: 2099 time 0:00:23.880938 train NMSE 0.01665 test NMSE 0.01678\n",
      "step: 2199 time 0:00:23.864483 train NMSE 0.01664 test NMSE 0.01484\n",
      "Model saved!\n",
      "step: 2299 time 0:00:23.807515 train NMSE 0.01663 test NMSE 0.01683\n",
      "step: 2399 time 0:00:23.856507 train NMSE 0.01663 test NMSE 0.01627\n",
      "step: 2499 time 0:00:23.873403 train NMSE 0.01666 test NMSE 0.01674\n",
      "step: 2599 time 0:00:23.905384 train NMSE 0.01666 test NMSE 0.01678\n",
      "step: 2699 time 0:00:23.883867 train NMSE 0.01675 test NMSE 0.01634\n",
      "step: 2799 time 0:00:23.884954 train NMSE 0.01675 test NMSE 0.01699\n",
      "step: 2899 time 0:00:23.900420 train NMSE 0.01682 test NMSE 0.01684\n",
      "step: 2999 time 0:00:23.905931 train NMSE 0.01674 test NMSE 0.01720\n"
     ]
    }
   ],
   "source": [
    "L  = 64  #观测时隙数\n",
    "sigma = 1e2\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:04.595704 train NMSE 0.78027 test NMSE 0.68867\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.559801 train NMSE 0.62366 test NMSE 0.58220\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.574761 train NMSE 0.55053 test NMSE 0.52232\n",
      "Model saved!\n",
      "step: 399 time 0:00:04.578750 train NMSE 0.49761 test NMSE 0.47654\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.567780 train NMSE 0.46112 test NMSE 0.42281\n",
      "Model saved!\n",
      "step: 599 time 0:00:04.581742 train NMSE 0.43270 test NMSE 0.42757\n",
      "step: 699 time 0:00:04.566783 train NMSE 0.40850 test NMSE 0.38997\n",
      "Model saved!\n",
      "step: 799 time 0:00:04.582752 train NMSE 0.38754 test NMSE 0.37850\n",
      "Model saved!\n",
      "step: 899 time 0:00:04.565800 train NMSE 0.37074 test NMSE 0.37318\n",
      "Model saved!\n",
      "step: 999 time 0:00:04.575773 train NMSE 0.35428 test NMSE 0.34035\n",
      "Model saved!\n",
      "step: 1099 time 0:00:04.561806 train NMSE 0.33918 test NMSE 0.33332\n",
      "Model saved!\n",
      "step: 1199 time 0:00:04.557806 train NMSE 0.32254 test NMSE 0.30660\n",
      "Model saved!\n",
      "step: 1299 time 0:00:04.556346 train NMSE 0.30551 test NMSE 0.29806\n",
      "Model saved!\n",
      "step: 1399 time 0:00:04.547330 train NMSE 0.29552 test NMSE 0.30983\n",
      "step: 1499 time 0:00:04.559816 train NMSE 0.27587 test NMSE 0.28185\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.564468 train NMSE 0.26125 test NMSE 0.25915\n",
      "Model saved!\n",
      "step: 1699 time 0:00:04.571782 train NMSE 0.26040 test NMSE 0.22291\n",
      "Model saved!\n",
      "step: 1799 time 0:00:04.559382 train NMSE 0.20036 test NMSE 0.19185\n",
      "Model saved!\n",
      "step: 1899 time 0:00:04.563294 train NMSE 0.17739 test NMSE 0.16977\n",
      "Model saved!\n",
      "step: 1999 time 0:00:04.564348 train NMSE 0.16439 test NMSE 0.17237\n",
      "step: 2099 time 0:00:04.572789 train NMSE 0.15707 test NMSE 0.15035\n",
      "Model saved!\n",
      "step: 2199 time 0:00:04.571171 train NMSE 0.14876 test NMSE 0.14237\n",
      "Model saved!\n",
      "step: 2299 time 0:00:04.571027 train NMSE 0.14363 test NMSE 0.13242\n",
      "Model saved!\n",
      "step: 2399 time 0:00:04.584775 train NMSE 0.14030 test NMSE 0.13634\n",
      "step: 2499 time 0:00:04.587168 train NMSE 0.13599 test NMSE 0.13340\n",
      "step: 2599 time 0:00:04.567482 train NMSE 0.13282 test NMSE 0.12489\n",
      "Model saved!\n",
      "step: 2699 time 0:00:04.563802 train NMSE 0.12996 test NMSE 0.11959\n",
      "Model saved!\n",
      "step: 2799 time 0:00:04.554854 train NMSE 0.12738 test NMSE 0.11917\n",
      "Model saved!\n",
      "step: 2899 time 0:00:04.560246 train NMSE 0.12590 test NMSE 0.12120\n",
      "step: 2999 time 0:00:04.543374 train NMSE 0.12232 test NMSE 0.12460\n",
      "step: 99 time 0:00:05.322440 train NMSE 0.26360 test NMSE 0.21809\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.386829 train NMSE 0.18901 test NMSE 0.17305\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.367198 train NMSE 0.15562 test NMSE 0.14898\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.356932 train NMSE 0.13852 test NMSE 0.14315\n",
      "Model saved!\n",
      "step: 499 time 0:00:05.360045 train NMSE 0.12757 test NMSE 0.11432\n",
      "Model saved!\n",
      "step: 599 time 0:00:05.364744 train NMSE 0.12171 test NMSE 0.12543\n",
      "step: 699 time 0:00:05.360221 train NMSE 0.11673 test NMSE 0.12870\n",
      "step: 799 time 0:00:05.372209 train NMSE 0.11262 test NMSE 0.10535\n",
      "Model saved!\n",
      "step: 899 time 0:00:05.374711 train NMSE 0.10935 test NMSE 0.11240\n",
      "step: 999 time 0:00:05.363497 train NMSE 0.10626 test NMSE 0.09938\n",
      "Model saved!\n",
      "step: 1099 time 0:00:05.369865 train NMSE 0.10444 test NMSE 0.10046\n",
      "step: 1199 time 0:00:05.379651 train NMSE 0.10263 test NMSE 0.09901\n",
      "Model saved!\n",
      "step: 1299 time 0:00:05.372893 train NMSE 0.10052 test NMSE 0.09509\n",
      "Model saved!\n",
      "step: 1399 time 0:00:05.357491 train NMSE 0.09952 test NMSE 0.09393\n",
      "Model saved!\n",
      "step: 1499 time 0:00:05.383909 train NMSE 0.09739 test NMSE 0.09221\n",
      "Model saved!\n",
      "step: 1599 time 0:00:05.370657 train NMSE 0.09637 test NMSE 0.09995\n",
      "step: 1699 time 0:00:05.370176 train NMSE 0.09474 test NMSE 0.09191\n",
      "Model saved!\n",
      "step: 1799 time 0:00:05.363845 train NMSE 0.09363 test NMSE 0.09278\n",
      "step: 1899 time 0:00:05.356104 train NMSE 0.09217 test NMSE 0.09136\n",
      "Model saved!\n",
      "step: 1999 time 0:00:05.349762 train NMSE 0.09057 test NMSE 0.08767\n",
      "Model saved!\n",
      "step: 2099 time 0:00:05.357872 train NMSE 0.09031 test NMSE 0.09604\n",
      "step: 2199 time 0:00:05.352793 train NMSE 0.08918 test NMSE 0.08852\n",
      "step: 2299 time 0:00:05.370171 train NMSE 0.08830 test NMSE 0.08487\n",
      "Model saved!\n",
      "step: 2399 time 0:00:05.363709 train NMSE 0.08774 test NMSE 0.08628\n",
      "step: 2499 time 0:00:05.369751 train NMSE 0.08725 test NMSE 0.08452\n",
      "Model saved!\n",
      "step: 2599 time 0:00:05.366626 train NMSE 0.08654 test NMSE 0.08588\n",
      "step: 2699 time 0:00:05.377616 train NMSE 0.08485 test NMSE 0.09315\n",
      "step: 2799 time 0:00:05.352151 train NMSE 0.08552 test NMSE 0.08486\n",
      "step: 2899 time 0:00:05.384037 train NMSE 0.08416 test NMSE 0.07992\n",
      "Model saved!\n",
      "step: 2999 time 0:00:05.347425 train NMSE 0.08344 test NMSE 0.08446\n",
      "step: 99 time 0:00:06.138122 train NMSE 0.15381 test NMSE 0.12816\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.142695 train NMSE 0.11910 test NMSE 0.11206\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.139653 train NMSE 0.11359 test NMSE 0.09970\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.176651 train NMSE 0.09755 test NMSE 0.09534\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.180867 train NMSE 0.11379 test NMSE 0.08718\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.157560 train NMSE 0.09929 test NMSE 0.08940\n",
      "step: 699 time 0:00:06.174382 train NMSE 0.10885 test NMSE 0.08678\n",
      "Model saved!\n",
      "step: 799 time 0:00:06.162418 train NMSE 0.14230 test NMSE 0.07962\n",
      "Model saved!\n",
      "step: 899 time 0:00:06.167978 train NMSE 0.08465 test NMSE 0.08763\n",
      "step: 999 time 0:00:06.173926 train NMSE 0.10636 test NMSE 0.08633\n",
      "step: 1099 time 0:00:06.148613 train NMSE 0.09471 test NMSE 0.07809\n",
      "Model saved!\n",
      "step: 1199 time 0:00:06.150335 train NMSE 0.08696 test NMSE 0.07665\n",
      "Model saved!\n",
      "step: 1299 time 0:00:06.164935 train NMSE 0.09467 test NMSE 0.07388\n",
      "Model saved!\n",
      "step: 1399 time 0:00:06.161271 train NMSE 0.08735 test NMSE 0.07824\n",
      "step: 1499 time 0:00:06.153589 train NMSE 0.09398 test NMSE 0.08243\n",
      "step: 1599 time 0:00:06.153361 train NMSE 0.08348 test NMSE 0.08130\n",
      "step: 1699 time 0:00:06.161514 train NMSE 0.07977 test NMSE 0.07757\n",
      "step: 1799 time 0:00:06.148234 train NMSE 0.09957 test NMSE 0.08219\n",
      "step: 1899 time 0:00:06.154604 train NMSE 0.08892 test NMSE 0.08796\n",
      "step: 1999 time 0:00:06.158131 train NMSE 0.09957 test NMSE 0.07364\n",
      "Model saved!\n",
      "step: 2099 time 0:00:06.142155 train NMSE 0.08509 test NMSE 0.07283\n",
      "Model saved!\n",
      "step: 2199 time 0:00:06.143599 train NMSE 0.07681 test NMSE 0.07747\n",
      "step: 2299 time 0:00:06.140638 train NMSE 0.08405 test NMSE 0.07153\n",
      "Model saved!\n",
      "step: 2399 time 0:00:06.146141 train NMSE 0.08748 test NMSE 0.07393\n",
      "step: 2499 time 0:00:06.144661 train NMSE 0.08782 test NMSE 0.07600\n",
      "step: 2599 time 0:00:06.163005 train NMSE 0.08059 test NMSE 0.07449\n",
      "step: 2699 time 0:00:06.166669 train NMSE 0.07459 test NMSE 0.07285\n",
      "step: 2799 time 0:00:06.161261 train NMSE 0.08481 test NMSE 0.08005\n",
      "step: 2899 time 0:00:06.169067 train NMSE 0.07371 test NMSE 0.06672\n",
      "Model saved!\n",
      "step: 2999 time 0:00:06.166062 train NMSE 0.07246 test NMSE 0.07535\n",
      "step: 99 time 0:00:06.938877 train NMSE 0.15440 test NMSE 0.10401\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.947481 train NMSE 0.09597 test NMSE 0.08914\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.946742 train NMSE 0.09156 test NMSE 0.08685\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.947479 train NMSE 0.08531 test NMSE 0.08285\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.976745 train NMSE 0.08521 test NMSE 0.08347\n",
      "step: 599 time 0:00:06.946339 train NMSE 0.07941 test NMSE 0.07674\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.943424 train NMSE 0.13402 test NMSE 0.07352\n",
      "Model saved!\n",
      "step: 799 time 0:00:06.953397 train NMSE 0.07705 test NMSE 0.07187\n",
      "Model saved!\n",
      "step: 899 time 0:00:06.934777 train NMSE 0.07701 test NMSE 0.07891\n",
      "step: 999 time 0:00:06.934998 train NMSE 0.07511 test NMSE 0.07401\n",
      "step: 1099 time 0:00:06.947261 train NMSE 0.11229 test NMSE 0.07183\n",
      "Model saved!\n",
      "step: 1199 time 0:00:06.932491 train NMSE 0.08791 test NMSE 0.08114\n",
      "step: 1299 time 0:00:06.942483 train NMSE 0.07234 test NMSE 0.06715\n",
      "Model saved!\n",
      "step: 1399 time 0:00:06.961060 train NMSE 0.07584 test NMSE 0.07283\n",
      "step: 1499 time 0:00:06.946443 train NMSE 0.08026 test NMSE 0.07069\n",
      "step: 1599 time 0:00:06.952783 train NMSE 0.07443 test NMSE 0.07079\n",
      "step: 1699 time 0:00:06.944882 train NMSE 0.07449 test NMSE 0.07138\n",
      "step: 1799 time 0:00:06.937454 train NMSE 0.07263 test NMSE 0.07692\n",
      "step: 1899 time 0:00:06.931569 train NMSE 0.08179 test NMSE 0.07252\n",
      "step: 1999 time 0:00:06.935531 train NMSE 0.08067 test NMSE 0.06910\n",
      "step: 2099 time 0:00:06.941491 train NMSE 0.08231 test NMSE 0.07055\n",
      "step: 2199 time 0:00:06.951493 train NMSE 0.07114 test NMSE 0.06570\n",
      "Model saved!\n",
      "step: 2299 time 0:00:06.931511 train NMSE 0.07529 test NMSE 0.06618\n",
      "step: 2399 time 0:00:06.948479 train NMSE 0.08765 test NMSE 0.06995\n",
      "step: 2499 time 0:00:06.954946 train NMSE 0.06860 test NMSE 0.06807\n",
      "step: 2599 time 0:00:06.948782 train NMSE 0.08302 test NMSE 0.06492\n",
      "Model saved!\n",
      "step: 2699 time 0:00:06.946890 train NMSE 0.06881 test NMSE 0.06754\n",
      "step: 2799 time 0:00:06.935289 train NMSE 0.06867 test NMSE 0.07040\n",
      "step: 2899 time 0:00:06.942586 train NMSE 0.08092 test NMSE 0.09203\n",
      "step: 2999 time 0:00:06.940701 train NMSE 0.08032 test NMSE 0.07082\n",
      "step: 99 time 0:00:07.721866 train NMSE 0.17998 test NMSE 0.08706\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.722918 train NMSE 0.18159 test NMSE 0.08953\n",
      "step: 299 time 0:00:07.731494 train NMSE 0.12129 test NMSE 0.08284\n",
      "Model saved!\n",
      "step: 399 time 0:00:07.718091 train NMSE 0.13437 test NMSE 0.07875\n",
      "Model saved!\n",
      "step: 499 time 0:00:07.722160 train NMSE 0.08465 test NMSE 0.07435\n",
      "Model saved!\n",
      "step: 599 time 0:00:07.730405 train NMSE 0.10682 test NMSE 0.07146\n",
      "Model saved!\n",
      "step: 699 time 0:00:07.736574 train NMSE 0.11826 test NMSE 0.07124\n",
      "Model saved!\n",
      "step: 799 time 0:00:07.729714 train NMSE 0.16012 test NMSE 0.06638\n",
      "Model saved!\n",
      "step: 899 time 0:00:07.715288 train NMSE 0.09507 test NMSE 0.07063\n",
      "step: 999 time 0:00:07.716510 train NMSE 0.13101 test NMSE 0.06823\n",
      "step: 1099 time 0:00:07.731779 train NMSE 0.08848 test NMSE 0.06921\n",
      "step: 1199 time 0:00:07.724876 train NMSE 0.09469 test NMSE 0.06507\n",
      "Model saved!\n",
      "step: 1299 time 0:00:07.720868 train NMSE 0.09812 test NMSE 0.07076\n",
      "step: 1399 time 0:00:07.740376 train NMSE 0.06776 test NMSE 0.06783\n",
      "step: 1499 time 0:00:07.735745 train NMSE 0.10954 test NMSE 0.06836\n",
      "step: 1599 time 0:00:07.736409 train NMSE 0.06640 test NMSE 0.06641\n",
      "step: 1699 time 0:00:07.720938 train NMSE 0.10438 test NMSE 0.06182\n",
      "Model saved!\n",
      "step: 1799 time 0:00:07.723540 train NMSE 0.07106 test NMSE 1.43218\n",
      "step: 1899 time 0:00:07.716314 train NMSE 0.09847 test NMSE 0.06387\n",
      "step: 1999 time 0:00:07.729465 train NMSE 0.07865 test NMSE 0.06181\n",
      "Model saved!\n",
      "step: 2099 time 0:00:07.700199 train NMSE 0.06553 test NMSE 0.06122\n",
      "Model saved!\n",
      "step: 2199 time 0:00:07.746034 train NMSE 0.07577 test NMSE 0.05900\n",
      "Model saved!\n",
      "step: 2299 time 0:00:07.736489 train NMSE 0.10049 test NMSE 0.06461\n",
      "step: 2399 time 0:00:07.724397 train NMSE 0.06701 test NMSE 0.06596\n",
      "step: 2499 time 0:00:07.727183 train NMSE 0.06429 test NMSE 0.06515\n",
      "step: 2599 time 0:00:07.715257 train NMSE 0.08400 test NMSE 0.06339\n",
      "step: 2699 time 0:00:07.720252 train NMSE 0.07515 test NMSE 0.06599\n",
      "step: 2799 time 0:00:07.721437 train NMSE 0.08658 test NMSE 0.05988\n",
      "step: 2899 time 0:00:07.730419 train NMSE 0.06311 test NMSE 0.06743\n",
      "step: 2999 time 0:00:07.728267 train NMSE 0.07480 test NMSE 0.06085\n"
     ]
    }
   ],
   "source": [
    "L  = 16  #观测时隙数\n",
    "sigma = 1e1\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:04.576362 train NMSE 0.83401 test NMSE 0.80441\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.570296 train NMSE 0.74328 test NMSE 0.70490\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.565856 train NMSE 0.64980 test NMSE 0.60753\n",
      "Model saved!\n",
      "step: 399 time 0:00:04.569905 train NMSE 0.59878 test NMSE 0.59611\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.553249 train NMSE 0.55729 test NMSE 0.53690\n",
      "Model saved!\n",
      "step: 599 time 0:00:04.555369 train NMSE 0.53029 test NMSE 0.57681\n",
      "step: 699 time 0:00:04.539899 train NMSE 0.51223 test NMSE 0.47126\n",
      "Model saved!\n",
      "step: 799 time 0:00:04.548173 train NMSE 0.49248 test NMSE 0.50415\n",
      "step: 899 time 0:00:04.545377 train NMSE 0.47818 test NMSE 0.49104\n",
      "step: 999 time 0:00:04.560126 train NMSE 0.46301 test NMSE 0.44418\n",
      "Model saved!\n",
      "step: 1099 time 0:00:04.560339 train NMSE 0.45366 test NMSE 0.42230\n",
      "Model saved!\n",
      "step: 1199 time 0:00:04.561838 train NMSE 0.43613 test NMSE 0.41819\n",
      "Model saved!\n",
      "step: 1299 time 0:00:04.613711 train NMSE 0.41775 test NMSE 0.42166\n",
      "step: 1399 time 0:00:04.559865 train NMSE 0.40433 test NMSE 0.42118\n",
      "step: 1499 time 0:00:04.561175 train NMSE 0.39435 test NMSE 0.39851\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.570657 train NMSE 0.38423 test NMSE 0.38270\n",
      "Model saved!\n",
      "step: 1699 time 0:00:04.567317 train NMSE 0.36637 test NMSE 0.36745\n",
      "Model saved!\n",
      "step: 1799 time 0:00:04.583011 train NMSE 0.34157 test NMSE 0.33614\n",
      "Model saved!\n",
      "step: 1899 time 0:00:04.578306 train NMSE 0.32647 test NMSE 0.32951\n",
      "Model saved!\n",
      "step: 1999 time 0:00:04.570329 train NMSE 0.30632 test NMSE 0.28985\n",
      "Model saved!\n",
      "step: 2099 time 0:00:04.555598 train NMSE 0.29278 test NMSE 0.28459\n",
      "Model saved!\n",
      "step: 2199 time 0:00:04.563306 train NMSE 0.28425 test NMSE 0.26503\n",
      "Model saved!\n",
      "step: 2299 time 0:00:04.580868 train NMSE 0.28156 test NMSE 0.26682\n",
      "step: 2399 time 0:00:04.561331 train NMSE 0.27750 test NMSE 0.28047\n",
      "step: 2499 time 0:00:04.548578 train NMSE 0.27324 test NMSE 0.29465\n",
      "step: 2599 time 0:00:04.559324 train NMSE 0.26647 test NMSE 0.27310\n",
      "step: 2699 time 0:00:04.574888 train NMSE 0.26316 test NMSE 0.26626\n",
      "step: 2799 time 0:00:04.559626 train NMSE 0.26128 test NMSE 0.25455\n",
      "Model saved!\n",
      "step: 2899 time 0:00:04.547898 train NMSE 0.25969 test NMSE 0.27177\n",
      "step: 2999 time 0:00:04.543362 train NMSE 0.25732 test NMSE 0.24900\n",
      "Model saved!\n",
      "step: 99 time 0:00:05.348500 train NMSE 0.30806 test NMSE 0.28619\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.333755 train NMSE 0.27207 test NMSE 0.26128\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.359701 train NMSE 0.25388 test NMSE 0.23914\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.364192 train NMSE 0.24429 test NMSE 0.23980\n",
      "step: 499 time 0:00:05.362694 train NMSE 0.23776 test NMSE 0.24212\n",
      "step: 599 time 0:00:05.367090 train NMSE 0.23463 test NMSE 0.22966\n",
      "Model saved!\n",
      "step: 699 time 0:00:05.347720 train NMSE 0.23189 test NMSE 0.21655\n",
      "Model saved!\n",
      "step: 799 time 0:00:05.369190 train NMSE 0.22904 test NMSE 0.23023\n",
      "step: 899 time 0:00:05.359637 train NMSE 0.22821 test NMSE 0.23335\n",
      "step: 999 time 0:00:05.379622 train NMSE 0.22535 test NMSE 0.23685\n",
      "step: 1099 time 0:00:05.375672 train NMSE 0.22181 test NMSE 0.22487\n",
      "step: 1199 time 0:00:05.369152 train NMSE 0.22325 test NMSE 0.22421\n",
      "step: 1299 time 0:00:05.368647 train NMSE 0.22276 test NMSE 0.23546\n",
      "step: 1399 time 0:00:05.375659 train NMSE 0.22052 test NMSE 0.21141\n",
      "Model saved!\n",
      "step: 1499 time 0:00:05.347240 train NMSE 0.21985 test NMSE 0.22800\n",
      "step: 1599 time 0:00:05.336263 train NMSE 0.21776 test NMSE 0.22191\n",
      "step: 1699 time 0:00:05.362692 train NMSE 0.21767 test NMSE 0.22440\n",
      "step: 1799 time 0:00:05.355688 train NMSE 0.21761 test NMSE 0.22345\n",
      "step: 1899 time 0:00:05.359180 train NMSE 0.21697 test NMSE 0.21573\n",
      "step: 1999 time 0:00:05.365671 train NMSE 0.21249 test NMSE 0.19456\n",
      "Model saved!\n",
      "step: 2099 time 0:00:05.363675 train NMSE 0.20714 test NMSE 0.20507\n",
      "step: 2199 time 0:00:05.361679 train NMSE 0.20492 test NMSE 0.20570\n",
      "step: 2299 time 0:00:05.349226 train NMSE 0.20102 test NMSE 0.19451\n",
      "Model saved!\n",
      "step: 2399 time 0:00:05.358882 train NMSE 0.19558 test NMSE 0.19324\n",
      "Model saved!\n",
      "step: 2499 time 0:00:05.358691 train NMSE 0.19436 test NMSE 0.19226\n",
      "Model saved!\n",
      "step: 2599 time 0:00:05.365870 train NMSE 0.19202 test NMSE 0.19324\n",
      "step: 2699 time 0:00:05.356679 train NMSE 0.18453 test NMSE 0.17709\n",
      "Model saved!\n",
      "step: 2799 time 0:00:05.362784 train NMSE 0.18185 test NMSE 0.17484\n",
      "Model saved!\n",
      "step: 2899 time 0:00:05.366680 train NMSE 0.18034 test NMSE 0.18270\n",
      "step: 2999 time 0:00:05.365522 train NMSE 0.17803 test NMSE 0.18070\n",
      "step: 99 time 0:00:06.146467 train NMSE 0.21276 test NMSE 0.19820\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.183624 train NMSE 0.19903 test NMSE 0.19671\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.160192 train NMSE 0.19181 test NMSE 0.18425\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.163753 train NMSE 0.18646 test NMSE 0.18774\n",
      "step: 499 time 0:00:06.148217 train NMSE 0.18512 test NMSE 0.18856\n",
      "step: 599 time 0:00:06.142613 train NMSE 0.18296 test NMSE 0.18238\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.132588 train NMSE 0.17926 test NMSE 0.17133\n",
      "Model saved!\n",
      "step: 799 time 0:00:06.183456 train NMSE 0.17846 test NMSE 0.17690\n",
      "step: 899 time 0:00:06.157528 train NMSE 0.17665 test NMSE 0.18109\n",
      "step: 999 time 0:00:06.162527 train NMSE 0.17440 test NMSE 0.17127\n",
      "Model saved!\n",
      "step: 1099 time 0:00:06.137917 train NMSE 0.17450 test NMSE 0.17917\n",
      "step: 1199 time 0:00:06.165613 train NMSE 0.17384 test NMSE 0.18968\n",
      "step: 1299 time 0:00:06.162047 train NMSE 0.17228 test NMSE 0.17485\n",
      "step: 1399 time 0:00:06.151611 train NMSE 0.17024 test NMSE 0.16577\n",
      "Model saved!\n",
      "step: 1499 time 0:00:06.162929 train NMSE 0.17127 test NMSE 0.16645\n",
      "step: 1599 time 0:00:06.144109 train NMSE 0.17039 test NMSE 0.16648\n",
      "step: 1699 time 0:00:06.161077 train NMSE 0.17058 test NMSE 0.17529\n",
      "step: 1799 time 0:00:06.160482 train NMSE 0.16998 test NMSE 0.16789\n",
      "step: 1899 time 0:00:06.174026 train NMSE 0.16925 test NMSE 0.16225\n",
      "Model saved!\n",
      "step: 1999 time 0:00:06.159058 train NMSE 0.16999 test NMSE 0.16764\n",
      "step: 2099 time 0:00:06.166562 train NMSE 0.16942 test NMSE 0.16473\n",
      "step: 2199 time 0:00:06.160079 train NMSE 0.16917 test NMSE 0.15943\n",
      "Model saved!\n",
      "step: 2299 time 0:00:06.156588 train NMSE 0.17002 test NMSE 0.16798\n",
      "step: 2399 time 0:00:06.163561 train NMSE 0.16789 test NMSE 0.16951\n",
      "step: 2499 time 0:00:06.156585 train NMSE 0.16875 test NMSE 0.17234\n",
      "step: 2599 time 0:00:06.154613 train NMSE 0.16877 test NMSE 0.16344\n",
      "step: 2699 time 0:00:06.180552 train NMSE 0.16849 test NMSE 0.17234\n",
      "step: 2799 time 0:00:06.164905 train NMSE 0.16818 test NMSE 0.16173\n",
      "step: 2899 time 0:00:06.166628 train NMSE 0.16735 test NMSE 0.16759\n",
      "step: 2999 time 0:00:06.154577 train NMSE 0.16643 test NMSE 0.16421\n",
      "step: 99 time 0:00:06.944315 train NMSE 0.21209 test NMSE 0.20426\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.949450 train NMSE 0.20685 test NMSE 0.20300\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.936844 train NMSE 0.20144 test NMSE 0.20673\n",
      "step: 399 time 0:00:06.933579 train NMSE 0.20063 test NMSE 0.21141\n",
      "step: 499 time 0:00:06.917574 train NMSE 0.19839 test NMSE 0.18213\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.918881 train NMSE 0.19668 test NMSE 0.19075\n",
      "step: 699 time 0:00:06.936547 train NMSE 0.19686 test NMSE 0.20103\n",
      "step: 799 time 0:00:06.939994 train NMSE 0.19486 test NMSE 0.18630\n",
      "step: 899 time 0:00:06.946925 train NMSE 0.19527 test NMSE 0.19352\n",
      "step: 999 time 0:00:06.937491 train NMSE 0.19386 test NMSE 0.19699\n",
      "step: 1099 time 0:00:06.937043 train NMSE 0.19361 test NMSE 0.19036\n",
      "step: 1199 time 0:00:06.952005 train NMSE 0.19416 test NMSE 0.19862\n",
      "step: 1299 time 0:00:06.925526 train NMSE 0.19221 test NMSE 0.18848\n",
      "step: 1399 time 0:00:06.950872 train NMSE 0.19231 test NMSE 0.19024\n",
      "step: 1499 time 0:00:06.931266 train NMSE 0.19200 test NMSE 0.18007\n",
      "Model saved!\n",
      "step: 1599 time 0:00:06.951282 train NMSE 0.19079 test NMSE 0.19647\n",
      "step: 1699 time 0:00:06.937292 train NMSE 0.18917 test NMSE 0.18831\n",
      "step: 1799 time 0:00:06.942891 train NMSE 0.18782 test NMSE 0.19038\n",
      "step: 1899 time 0:00:06.950101 train NMSE 0.18892 test NMSE 0.18222\n",
      "step: 1999 time 0:00:06.955393 train NMSE 0.18881 test NMSE 0.18279\n",
      "step: 2099 time 0:00:06.939434 train NMSE 0.18782 test NMSE 0.17891\n",
      "Model saved!\n",
      "step: 2199 time 0:00:06.944377 train NMSE 0.18659 test NMSE 0.19993\n",
      "step: 2299 time 0:00:06.937560 train NMSE 0.18566 test NMSE 0.18334\n",
      "step: 2399 time 0:00:06.922478 train NMSE 0.18363 test NMSE 0.18334\n",
      "step: 2499 time 0:00:06.924847 train NMSE 0.18394 test NMSE 0.17774\n",
      "Model saved!\n",
      "step: 2599 time 0:00:06.928600 train NMSE 0.18384 test NMSE 0.18905\n",
      "step: 2699 time 0:00:06.958675 train NMSE 0.18187 test NMSE 0.18018\n",
      "step: 2799 time 0:00:06.944498 train NMSE 0.17952 test NMSE 0.17367\n",
      "Model saved!\n",
      "step: 2899 time 0:00:06.947492 train NMSE 0.17832 test NMSE 0.17897\n",
      "step: 2999 time 0:00:06.959256 train NMSE 0.17663 test NMSE 0.18122\n",
      "step: 99 time 0:00:07.719072 train NMSE 0.19758 test NMSE 0.20026\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.724848 train NMSE 0.19342 test NMSE 0.17813\n",
      "Model saved!\n",
      "step: 299 time 0:00:07.732595 train NMSE 0.19145 test NMSE 0.19351\n",
      "step: 399 time 0:00:07.724108 train NMSE 0.19089 test NMSE 0.19503\n",
      "step: 499 time 0:00:07.731047 train NMSE 0.18930 test NMSE 0.19287\n",
      "step: 599 time 0:00:07.718282 train NMSE 0.19019 test NMSE 0.19745\n",
      "step: 699 time 0:00:07.725365 train NMSE 0.18930 test NMSE 0.18653\n",
      "step: 799 time 0:00:07.716444 train NMSE 0.18839 test NMSE 0.18278\n",
      "step: 899 time 0:00:07.726158 train NMSE 0.18803 test NMSE 0.18648\n",
      "step: 999 time 0:00:07.710253 train NMSE 0.18855 test NMSE 0.20003\n",
      "step: 1099 time 0:00:07.740308 train NMSE 0.18805 test NMSE 0.18399\n",
      "step: 1199 time 0:00:07.730372 train NMSE 0.18845 test NMSE 0.18078\n",
      "step: 1299 time 0:00:07.721034 train NMSE 0.18842 test NMSE 0.18812\n",
      "step: 1399 time 0:00:07.718803 train NMSE 0.18667 test NMSE 0.19176\n",
      "step: 1499 time 0:00:07.717623 train NMSE 0.18636 test NMSE 0.18025\n",
      "step: 1599 time 0:00:07.706650 train NMSE 0.18672 test NMSE 0.19853\n",
      "step: 1699 time 0:00:07.710001 train NMSE 0.18603 test NMSE 0.18118\n",
      "step: 1799 time 0:00:07.731905 train NMSE 0.18544 test NMSE 0.17702\n",
      "Model saved!\n",
      "step: 1899 time 0:00:07.710947 train NMSE 0.18397 test NMSE 0.18431\n",
      "step: 1999 time 0:00:07.730394 train NMSE 0.18397 test NMSE 0.18550\n",
      "step: 2099 time 0:00:07.730090 train NMSE 0.18310 test NMSE 0.17641\n",
      "Model saved!\n",
      "step: 2199 time 0:00:07.737391 train NMSE 0.18284 test NMSE 0.17821\n",
      "step: 2299 time 0:00:07.731956 train NMSE 0.18077 test NMSE 0.18085\n",
      "step: 2399 time 0:00:07.771266 train NMSE 0.18037 test NMSE 0.17853\n",
      "step: 2499 time 0:00:07.723389 train NMSE 0.18014 test NMSE 0.17283\n",
      "Model saved!\n",
      "step: 2599 time 0:00:07.730601 train NMSE 0.17855 test NMSE 0.18105\n",
      "step: 2699 time 0:00:07.720429 train NMSE 0.17923 test NMSE 0.16792\n",
      "Model saved!\n",
      "step: 2799 time 0:00:07.733764 train NMSE 0.17677 test NMSE 0.17222\n",
      "step: 2899 time 0:00:07.735285 train NMSE 0.17655 test NMSE 0.17521\n",
      "step: 2999 time 0:00:07.727401 train NMSE 0.17634 test NMSE 0.17896\n"
     ]
    }
   ],
   "source": [
    "L  = 16  #观测时隙数\n",
    "sigma = 1e3\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:04.577830 train NMSE 0.98112 test NMSE 0.96433\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.556846 train NMSE 0.94311 test NMSE 0.91255\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.565869 train NMSE 0.86525 test NMSE 0.83669\n",
      "Model saved!\n",
      "step: 399 time 0:00:04.554872 train NMSE 0.81526 test NMSE 0.81008\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.546963 train NMSE 0.78163 test NMSE 0.75527\n",
      "Model saved!\n",
      "step: 599 time 0:00:04.563331 train NMSE 0.75527 test NMSE 0.73259\n",
      "Model saved!\n",
      "step: 699 time 0:00:04.568834 train NMSE 0.72982 test NMSE 0.71661\n",
      "Model saved!\n",
      "step: 799 time 0:00:04.550359 train NMSE 0.71234 test NMSE 0.69036\n",
      "Model saved!\n",
      "step: 899 time 0:00:04.563563 train NMSE 0.70155 test NMSE 0.72540\n",
      "step: 999 time 0:00:04.550943 train NMSE 0.69179 test NMSE 0.70176\n",
      "step: 1099 time 0:00:04.548527 train NMSE 0.68119 test NMSE 0.67474\n",
      "Model saved!\n",
      "step: 1199 time 0:00:04.564854 train NMSE 0.67095 test NMSE 0.68982\n",
      "step: 1299 time 0:00:04.565321 train NMSE 0.66181 test NMSE 0.66565\n",
      "Model saved!\n",
      "step: 1399 time 0:00:04.567197 train NMSE 0.64037 test NMSE 0.64048\n",
      "Model saved!\n",
      "step: 1499 time 0:00:04.568944 train NMSE 0.62810 test NMSE 0.62528\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.561116 train NMSE 0.60576 test NMSE 0.59798\n",
      "Model saved!\n",
      "step: 1699 time 0:00:04.561839 train NMSE 0.58169 test NMSE 0.57703\n",
      "Model saved!\n",
      "step: 1799 time 0:00:04.562668 train NMSE 0.56693 test NMSE 0.57090\n",
      "Model saved!\n",
      "step: 1899 time 0:00:04.559967 train NMSE 0.54739 test NMSE 0.54320\n",
      "Model saved!\n",
      "step: 1999 time 0:00:04.557167 train NMSE 0.52998 test NMSE 0.51179\n",
      "Model saved!\n",
      "step: 2099 time 0:00:04.565255 train NMSE 0.52266 test NMSE 0.50147\n",
      "Model saved!\n",
      "step: 2199 time 0:00:04.562327 train NMSE 0.51502 test NMSE 0.50372\n",
      "step: 2299 time 0:00:04.550703 train NMSE 0.50910 test NMSE 0.50345\n",
      "step: 2399 time 0:00:04.553092 train NMSE 0.50265 test NMSE 0.51068\n",
      "step: 2499 time 0:00:04.553250 train NMSE 0.50167 test NMSE 0.49509\n",
      "Model saved!\n",
      "step: 2599 time 0:00:04.547778 train NMSE 0.49994 test NMSE 0.50639\n",
      "step: 2699 time 0:00:04.546379 train NMSE 0.49882 test NMSE 0.48868\n",
      "Model saved!\n",
      "step: 2799 time 0:00:04.562959 train NMSE 0.49329 test NMSE 0.48496\n",
      "Model saved!\n",
      "step: 2899 time 0:00:04.573562 train NMSE 0.49191 test NMSE 0.47161\n",
      "Model saved!\n",
      "step: 2999 time 0:00:04.558857 train NMSE 0.48892 test NMSE 0.48223\n",
      "step: 99 time 0:00:05.348257 train NMSE 0.78166 test NMSE 0.71259\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.366206 train NMSE 0.63732 test NMSE 0.58832\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.417579 train NMSE 0.55568 test NMSE 0.53833\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.378071 train NMSE 0.52523 test NMSE 0.52651\n",
      "Model saved!\n",
      "step: 499 time 0:00:05.381602 train NMSE 0.51599 test NMSE 0.50085\n",
      "Model saved!\n",
      "step: 599 time 0:00:05.369635 train NMSE 0.51031 test NMSE 0.48856\n",
      "Model saved!\n",
      "step: 699 time 0:00:05.374622 train NMSE 0.50376 test NMSE 0.49177\n",
      "step: 799 time 0:00:05.376630 train NMSE 0.50096 test NMSE 0.51582\n",
      "step: 899 time 0:00:05.397840 train NMSE 0.49572 test NMSE 0.49333\n",
      "step: 999 time 0:00:05.367150 train NMSE 0.49379 test NMSE 0.49147\n",
      "step: 1099 time 0:00:05.357667 train NMSE 0.48978 test NMSE 0.46973\n",
      "Model saved!\n",
      "step: 1199 time 0:00:05.362668 train NMSE 0.48641 test NMSE 0.50014\n",
      "step: 1299 time 0:00:05.365179 train NMSE 0.48313 test NMSE 0.46016\n",
      "Model saved!\n",
      "step: 1399 time 0:00:05.367652 train NMSE 0.48407 test NMSE 0.47196\n",
      "step: 1499 time 0:00:05.353761 train NMSE 0.47862 test NMSE 0.48135\n",
      "step: 1599 time 0:00:05.343769 train NMSE 0.47770 test NMSE 0.47992\n",
      "step: 1699 time 0:00:05.368667 train NMSE 0.47492 test NMSE 0.46547\n",
      "step: 1799 time 0:00:05.368178 train NMSE 0.47281 test NMSE 0.45061\n",
      "Model saved!\n",
      "step: 1899 time 0:00:05.353125 train NMSE 0.47030 test NMSE 0.44909\n",
      "Model saved!\n",
      "step: 1999 time 0:00:05.374632 train NMSE 0.46779 test NMSE 0.48363\n",
      "step: 2099 time 0:00:05.370140 train NMSE 0.46633 test NMSE 0.47964\n",
      "step: 2199 time 0:00:05.367716 train NMSE 0.46377 test NMSE 0.47335\n",
      "step: 2299 time 0:00:05.368652 train NMSE 0.46438 test NMSE 0.45616\n",
      "step: 2399 time 0:00:05.363660 train NMSE 0.46228 test NMSE 0.46933\n",
      "step: 2499 time 0:00:05.351726 train NMSE 0.46049 test NMSE 0.45730\n",
      "step: 2599 time 0:00:05.340816 train NMSE 0.45827 test NMSE 0.43064\n",
      "Model saved!\n",
      "step: 2699 time 0:00:05.376021 train NMSE 0.45861 test NMSE 0.46319\n",
      "step: 2799 time 0:00:05.359017 train NMSE 0.45904 test NMSE 0.46265\n",
      "step: 2899 time 0:00:05.359197 train NMSE 0.45888 test NMSE 0.45841\n",
      "step: 2999 time 0:00:05.371320 train NMSE 0.45776 test NMSE 0.46300\n",
      "step: 99 time 0:00:06.143043 train NMSE 0.66478 test NMSE 0.60948\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.164529 train NMSE 0.58290 test NMSE 0.56538\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.140625 train NMSE 0.53317 test NMSE 0.52030\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.173565 train NMSE 0.50897 test NMSE 0.51514\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.173512 train NMSE 0.50201 test NMSE 0.50736\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.169032 train NMSE 0.49792 test NMSE 0.49676\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.171725 train NMSE 0.49570 test NMSE 0.49230\n",
      "Model saved!\n",
      "step: 799 time 0:00:06.177033 train NMSE 0.49539 test NMSE 0.46989\n",
      "Model saved!\n",
      "step: 899 time 0:00:06.157541 train NMSE 0.49147 test NMSE 0.51517\n",
      "step: 999 time 0:00:06.174347 train NMSE 0.48930 test NMSE 0.48662\n",
      "step: 1099 time 0:00:06.168343 train NMSE 0.48685 test NMSE 0.49228\n",
      "step: 1199 time 0:00:06.167565 train NMSE 0.48723 test NMSE 0.49016\n",
      "step: 1299 time 0:00:06.155111 train NMSE 0.48327 test NMSE 0.48135\n",
      "step: 1399 time 0:00:06.174035 train NMSE 0.48215 test NMSE 0.46976\n",
      "Model saved!\n",
      "step: 1499 time 0:00:06.159630 train NMSE 0.48217 test NMSE 0.47834\n",
      "step: 1599 time 0:00:06.158077 train NMSE 0.47996 test NMSE 0.51538\n",
      "step: 1699 time 0:00:06.157935 train NMSE 0.47865 test NMSE 0.46335\n",
      "Model saved!\n",
      "step: 1799 time 0:00:06.158868 train NMSE 0.47744 test NMSE 0.48006\n",
      "step: 1899 time 0:00:06.168524 train NMSE 0.47576 test NMSE 0.47320\n",
      "step: 1999 time 0:00:06.155160 train NMSE 0.47570 test NMSE 0.48170\n",
      "step: 2099 time 0:00:06.154150 train NMSE 0.47314 test NMSE 0.46386\n",
      "step: 2199 time 0:00:06.150620 train NMSE 0.47174 test NMSE 0.46454\n",
      "step: 2299 time 0:00:06.160130 train NMSE 0.46763 test NMSE 0.46384\n",
      "step: 2399 time 0:00:06.164103 train NMSE 0.46932 test NMSE 0.45240\n",
      "Model saved!\n",
      "step: 2499 time 0:00:06.167272 train NMSE 0.46744 test NMSE 0.45584\n",
      "step: 2599 time 0:00:06.169016 train NMSE 0.46602 test NMSE 0.45068\n",
      "Model saved!\n",
      "step: 2699 time 0:00:06.147092 train NMSE 0.46373 test NMSE 0.48138\n",
      "step: 2799 time 0:00:06.140614 train NMSE 0.46439 test NMSE 0.46548\n",
      "step: 2899 time 0:00:06.150302 train NMSE 0.46321 test NMSE 0.46815\n",
      "step: 2999 time 0:00:06.168054 train NMSE 0.46144 test NMSE 0.46790\n",
      "step: 99 time 0:00:06.902148 train NMSE 0.58153 test NMSE 0.55875\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.925551 train NMSE 0.53760 test NMSE 0.50867\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.922212 train NMSE 0.50996 test NMSE 0.51546\n",
      "step: 399 time 0:00:06.940501 train NMSE 0.49961 test NMSE 0.49036\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.945645 train NMSE 0.49495 test NMSE 0.49262\n",
      "step: 599 time 0:00:06.936508 train NMSE 0.49020 test NMSE 0.47734\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.938017 train NMSE 0.48840 test NMSE 0.50869\n",
      "step: 799 time 0:00:06.937680 train NMSE 0.48842 test NMSE 0.50172\n",
      "step: 899 time 0:00:06.931456 train NMSE 0.48713 test NMSE 0.47499\n",
      "Model saved!\n",
      "step: 999 time 0:00:06.940505 train NMSE 0.48575 test NMSE 0.50831\n",
      "step: 1099 time 0:00:06.941506 train NMSE 0.48535 test NMSE 0.50316\n",
      "step: 1199 time 0:00:06.943494 train NMSE 0.48286 test NMSE 0.49627\n",
      "step: 1299 time 0:00:06.947954 train NMSE 0.48089 test NMSE 0.46747\n",
      "Model saved!\n",
      "step: 1399 time 0:00:06.955956 train NMSE 0.48047 test NMSE 0.48948\n",
      "step: 1499 time 0:00:06.944680 train NMSE 0.47797 test NMSE 0.48177\n",
      "step: 1599 time 0:00:06.933820 train NMSE 0.47523 test NMSE 0.48074\n",
      "step: 1699 time 0:00:06.969404 train NMSE 0.47720 test NMSE 0.45892\n",
      "Model saved!\n",
      "step: 1799 time 0:00:06.918026 train NMSE 0.47339 test NMSE 0.46920\n",
      "step: 1899 time 0:00:06.944507 train NMSE 0.47508 test NMSE 0.47444\n",
      "step: 1999 time 0:00:06.952961 train NMSE 0.47061 test NMSE 0.47296\n",
      "step: 2099 time 0:00:06.957343 train NMSE 0.47064 test NMSE 0.49802\n",
      "step: 2199 time 0:00:06.948008 train NMSE 0.47104 test NMSE 0.48552\n",
      "step: 2299 time 0:00:06.949498 train NMSE 0.46717 test NMSE 0.47266\n",
      "step: 2399 time 0:00:06.934575 train NMSE 0.46822 test NMSE 0.48235\n",
      "step: 2499 time 0:00:06.962488 train NMSE 0.46445 test NMSE 0.47154\n",
      "step: 2599 time 0:00:06.950468 train NMSE 0.46683 test NMSE 0.47427\n",
      "step: 2699 time 0:00:06.944467 train NMSE 0.46496 test NMSE 0.48204\n",
      "step: 2799 time 0:00:06.927034 train NMSE 0.46479 test NMSE 0.47777\n",
      "step: 2899 time 0:00:06.949996 train NMSE 0.46396 test NMSE 0.46628\n",
      "step: 2999 time 0:00:06.942033 train NMSE 0.46140 test NMSE 0.47196\n",
      "step: 99 time 0:00:07.727932 train NMSE 0.50948 test NMSE 0.49984\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.729391 train NMSE 0.49455 test NMSE 0.48776\n",
      "Model saved!\n",
      "step: 299 time 0:00:07.732397 train NMSE 0.48858 test NMSE 0.49541\n",
      "step: 399 time 0:00:07.735337 train NMSE 0.48662 test NMSE 0.48804\n",
      "step: 499 time 0:00:07.716970 train NMSE 0.48587 test NMSE 0.49458\n",
      "step: 599 time 0:00:07.731432 train NMSE 0.48028 test NMSE 0.46451\n",
      "Model saved!\n",
      "step: 699 time 0:00:07.738376 train NMSE 0.48193 test NMSE 0.47432\n",
      "step: 799 time 0:00:07.730391 train NMSE 0.48277 test NMSE 0.49201\n",
      "step: 899 time 0:00:07.740340 train NMSE 0.48261 test NMSE 0.48898\n",
      "step: 999 time 0:00:07.718546 train NMSE 0.48044 test NMSE 0.47414\n",
      "step: 1099 time 0:00:07.726386 train NMSE 0.47884 test NMSE 0.48594\n",
      "step: 1199 time 0:00:07.726394 train NMSE 0.47795 test NMSE 0.47354\n",
      "step: 1299 time 0:00:07.718504 train NMSE 0.47797 test NMSE 0.47242\n",
      "step: 1399 time 0:00:07.734421 train NMSE 0.47655 test NMSE 0.45850\n",
      "Model saved!\n",
      "step: 1499 time 0:00:07.722878 train NMSE 0.47605 test NMSE 0.50466\n",
      "step: 1599 time 0:00:07.730402 train NMSE 0.47467 test NMSE 0.46623\n",
      "step: 1699 time 0:00:07.719441 train NMSE 0.47309 test NMSE 0.46259\n",
      "step: 1799 time 0:00:07.726243 train NMSE 0.47052 test NMSE 0.47388\n",
      "step: 1899 time 0:00:07.725938 train NMSE 0.47036 test NMSE 0.48608\n",
      "step: 1999 time 0:00:07.720498 train NMSE 0.47104 test NMSE 0.46284\n",
      "step: 2099 time 0:00:07.733878 train NMSE 0.46891 test NMSE 0.49039\n",
      "step: 2199 time 0:00:07.723902 train NMSE 0.46783 test NMSE 0.44934\n",
      "Model saved!\n",
      "step: 2299 time 0:00:07.733918 train NMSE 0.46787 test NMSE 0.49433\n",
      "step: 2399 time 0:00:07.723947 train NMSE 0.46848 test NMSE 0.45726\n",
      "step: 2499 time 0:00:07.718400 train NMSE 0.46754 test NMSE 0.48411\n",
      "step: 2599 time 0:00:07.736414 train NMSE 0.46597 test NMSE 0.47152\n",
      "step: 2699 time 0:00:07.719805 train NMSE 0.46718 test NMSE 0.47338\n",
      "step: 2799 time 0:00:07.730368 train NMSE 0.46548 test NMSE 0.47305\n",
      "step: 2899 time 0:00:07.721937 train NMSE 0.46749 test NMSE 0.47031\n",
      "step: 2999 time 0:00:07.722379 train NMSE 0.46285 test NMSE 0.46522\n"
     ]
    }
   ],
   "source": [
    "L  = 16  #观测时隙数\n",
    "sigma = 1e4\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:04.600764 train NMSE 0.99639 test NMSE 0.99236\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.580816 train NMSE 0.94639 test NMSE 0.91903\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.554826 train NMSE 0.90981 test NMSE 0.91951\n",
      "step: 399 time 0:00:04.551852 train NMSE 0.90571 test NMSE 0.90035\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.560811 train NMSE 0.90058 test NMSE 0.90788\n",
      "step: 599 time 0:00:04.567811 train NMSE 0.90023 test NMSE 0.88766\n",
      "Model saved!\n",
      "step: 699 time 0:00:04.572788 train NMSE 0.89711 test NMSE 0.88844\n",
      "step: 799 time 0:00:04.559849 train NMSE 0.89422 test NMSE 0.89237\n",
      "step: 899 time 0:00:04.543393 train NMSE 0.89349 test NMSE 0.89104\n",
      "step: 999 time 0:00:04.547884 train NMSE 0.89168 test NMSE 0.89503\n",
      "step: 1099 time 0:00:04.567818 train NMSE 0.88880 test NMSE 0.90385\n",
      "step: 1199 time 0:00:04.553846 train NMSE 0.88786 test NMSE 0.90355\n",
      "step: 1299 time 0:00:04.574312 train NMSE 0.88872 test NMSE 0.88799\n",
      "step: 1399 time 0:00:04.606758 train NMSE 0.88678 test NMSE 0.88596\n",
      "Model saved!\n",
      "step: 1499 time 0:00:04.588803 train NMSE 0.88692 test NMSE 0.88010\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.562837 train NMSE 0.88362 test NMSE 0.88358\n",
      "step: 1699 time 0:00:04.554842 train NMSE 0.88302 test NMSE 0.87887\n",
      "Model saved!\n",
      "step: 1799 time 0:00:04.557875 train NMSE 0.88095 test NMSE 0.87489\n",
      "Model saved!\n",
      "step: 1899 time 0:00:04.561795 train NMSE 0.88210 test NMSE 0.87630\n",
      "step: 1999 time 0:00:04.562313 train NMSE 0.88010 test NMSE 0.87565\n",
      "step: 2099 time 0:00:04.533884 train NMSE 0.87992 test NMSE 0.86414\n",
      "Model saved!\n",
      "step: 2199 time 0:00:04.540899 train NMSE 0.87879 test NMSE 0.87696\n",
      "step: 2299 time 0:00:04.549373 train NMSE 0.87789 test NMSE 0.88851\n",
      "step: 2399 time 0:00:04.553227 train NMSE 0.87738 test NMSE 0.87000\n",
      "step: 2499 time 0:00:04.536925 train NMSE 0.87628 test NMSE 0.88189\n",
      "step: 2599 time 0:00:04.549857 train NMSE 0.87641 test NMSE 0.89295\n",
      "step: 2699 time 0:00:04.564530 train NMSE 0.87463 test NMSE 0.86465\n",
      "step: 2799 time 0:00:04.568866 train NMSE 0.87383 test NMSE 0.87792\n",
      "step: 2899 time 0:00:04.561811 train NMSE 0.87058 test NMSE 0.86923\n",
      "step: 2999 time 0:00:04.563995 train NMSE 0.87117 test NMSE 0.86038\n",
      "Model saved!\n",
      "step: 99 time 0:00:05.356721 train NMSE 0.99875 test NMSE 0.99577\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.360261 train NMSE 0.97437 test NMSE 0.93823\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.355247 train NMSE 0.91681 test NMSE 0.89199\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.359204 train NMSE 0.90533 test NMSE 0.90635\n",
      "step: 499 time 0:00:05.357754 train NMSE 0.90211 test NMSE 0.90691\n",
      "step: 599 time 0:00:05.366749 train NMSE 0.90188 test NMSE 0.89738\n",
      "step: 699 time 0:00:05.354747 train NMSE 0.89898 test NMSE 0.89741\n",
      "step: 799 time 0:00:05.379870 train NMSE 0.89703 test NMSE 0.88900\n",
      "Model saved!\n",
      "step: 899 time 0:00:05.338767 train NMSE 0.89604 test NMSE 0.90153\n",
      "step: 999 time 0:00:05.358730 train NMSE 0.89564 test NMSE 0.87656\n",
      "Model saved!\n",
      "step: 1099 time 0:00:05.359724 train NMSE 0.89087 test NMSE 0.88422\n",
      "step: 1199 time 0:00:05.350824 train NMSE 0.89093 test NMSE 0.89903\n",
      "step: 1299 time 0:00:05.364217 train NMSE 0.88870 test NMSE 0.88219\n",
      "step: 1399 time 0:00:05.359734 train NMSE 0.89029 test NMSE 0.90339\n",
      "step: 1499 time 0:00:05.361163 train NMSE 0.88726 test NMSE 0.89403\n",
      "step: 1599 time 0:00:05.371699 train NMSE 0.88586 test NMSE 0.88867\n",
      "step: 1699 time 0:00:05.361712 train NMSE 0.88466 test NMSE 0.88029\n",
      "step: 1799 time 0:00:05.364744 train NMSE 0.88474 test NMSE 0.87909\n",
      "step: 1899 time 0:00:05.375677 train NMSE 0.88462 test NMSE 0.87754\n",
      "step: 1999 time 0:00:05.363753 train NMSE 0.88297 test NMSE 0.87079\n",
      "Model saved!\n",
      "step: 2099 time 0:00:05.376351 train NMSE 0.87928 test NMSE 0.88054\n",
      "step: 2199 time 0:00:05.372691 train NMSE 0.88311 test NMSE 0.88193\n",
      "step: 2299 time 0:00:05.376682 train NMSE 0.88217 test NMSE 0.86710\n",
      "Model saved!\n",
      "step: 2399 time 0:00:05.372677 train NMSE 0.88031 test NMSE 0.87946\n",
      "step: 2499 time 0:00:05.369210 train NMSE 0.88041 test NMSE 0.90398\n",
      "step: 2599 time 0:00:05.360220 train NMSE 0.88032 test NMSE 0.88198\n",
      "step: 2699 time 0:00:05.353264 train NMSE 0.87988 test NMSE 0.88525\n",
      "step: 2799 time 0:00:05.370723 train NMSE 0.87894 test NMSE 0.87494\n",
      "step: 2899 time 0:00:05.345259 train NMSE 0.88004 test NMSE 0.86812\n",
      "step: 2999 time 0:00:05.345792 train NMSE 0.87947 test NMSE 0.87850\n",
      "step: 99 time 0:00:06.149118 train NMSE 0.99742 test NMSE 0.99402\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.164080 train NMSE 0.98576 test NMSE 0.96242\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.157893 train NMSE 0.94865 test NMSE 0.92467\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.163598 train NMSE 0.91348 test NMSE 0.88859\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.165578 train NMSE 0.90016 test NMSE 0.90739\n",
      "step: 599 time 0:00:06.164557 train NMSE 0.89422 test NMSE 0.89515\n",
      "step: 699 time 0:00:06.159213 train NMSE 0.89096 test NMSE 0.89529\n",
      "step: 799 time 0:00:06.151344 train NMSE 0.88805 test NMSE 0.91520\n",
      "step: 899 time 0:00:06.132842 train NMSE 0.88638 test NMSE 0.87425\n",
      "Model saved!\n",
      "step: 999 time 0:00:06.125643 train NMSE 0.88436 test NMSE 0.88131\n",
      "step: 1099 time 0:00:06.147062 train NMSE 0.88483 test NMSE 0.88164\n",
      "step: 1199 time 0:00:06.144585 train NMSE 0.88300 test NMSE 0.87880\n",
      "step: 1299 time 0:00:06.175523 train NMSE 0.88120 test NMSE 0.87625\n",
      "step: 1399 time 0:00:06.161589 train NMSE 0.88034 test NMSE 0.86902\n",
      "Model saved!\n",
      "step: 1499 time 0:00:06.170563 train NMSE 0.87997 test NMSE 0.88225\n",
      "step: 1599 time 0:00:06.157564 train NMSE 0.87902 test NMSE 0.87430\n",
      "step: 1699 time 0:00:06.158845 train NMSE 0.87722 test NMSE 0.89056\n",
      "step: 1799 time 0:00:06.138133 train NMSE 0.87684 test NMSE 0.87911\n",
      "step: 1899 time 0:00:06.151031 train NMSE 0.87678 test NMSE 0.88532\n",
      "step: 1999 time 0:00:06.151630 train NMSE 0.87641 test NMSE 0.88676\n",
      "step: 2099 time 0:00:06.155138 train NMSE 0.87630 test NMSE 0.88385\n",
      "step: 2199 time 0:00:06.159615 train NMSE 0.87497 test NMSE 0.86746\n",
      "Model saved!\n",
      "step: 2299 time 0:00:06.157590 train NMSE 0.87418 test NMSE 0.88214\n",
      "step: 2399 time 0:00:06.140140 train NMSE 0.87409 test NMSE 0.87412\n",
      "step: 2499 time 0:00:06.152092 train NMSE 0.87457 test NMSE 0.86907\n",
      "step: 2599 time 0:00:06.151576 train NMSE 0.87213 test NMSE 0.86825\n",
      "step: 2699 time 0:00:06.150555 train NMSE 0.87360 test NMSE 0.88621\n",
      "step: 2799 time 0:00:06.126159 train NMSE 0.87232 test NMSE 0.88883\n",
      "step: 2899 time 0:00:06.152358 train NMSE 0.87310 test NMSE 0.86357\n",
      "Model saved!\n",
      "step: 2999 time 0:00:06.141649 train NMSE 0.87271 test NMSE 0.86666\n",
      "step: 99 time 0:00:06.918652 train NMSE 0.99780 test NMSE 0.99593\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.955457 train NMSE 0.98616 test NMSE 0.97056\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.947488 train NMSE 0.94151 test NMSE 0.89635\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.940131 train NMSE 0.90733 test NMSE 0.90253\n",
      "step: 499 time 0:00:06.949957 train NMSE 0.89818 test NMSE 0.88451\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.930938 train NMSE 0.89550 test NMSE 0.88810\n",
      "step: 699 time 0:00:06.924536 train NMSE 0.89258 test NMSE 0.91137\n",
      "step: 799 time 0:00:06.966701 train NMSE 0.89147 test NMSE 0.89506\n",
      "step: 899 time 0:00:06.923119 train NMSE 0.88821 test NMSE 0.87882\n",
      "Model saved!\n",
      "step: 999 time 0:00:06.940817 train NMSE 0.88600 test NMSE 0.89666\n",
      "step: 1099 time 0:00:06.928027 train NMSE 0.88502 test NMSE 0.88770\n",
      "step: 1199 time 0:00:06.938516 train NMSE 0.88324 test NMSE 0.88490\n",
      "step: 1299 time 0:00:06.960684 train NMSE 0.88180 test NMSE 0.87206\n",
      "Model saved!\n",
      "step: 1399 time 0:00:06.945759 train NMSE 0.88021 test NMSE 0.87998\n",
      "step: 1499 time 0:00:06.954713 train NMSE 0.88218 test NMSE 0.89331\n",
      "step: 1599 time 0:00:06.958456 train NMSE 0.87869 test NMSE 0.88454\n",
      "step: 1699 time 0:00:06.954031 train NMSE 0.87991 test NMSE 0.88149\n",
      "step: 1799 time 0:00:06.942997 train NMSE 0.87834 test NMSE 0.87379\n",
      "step: 1899 time 0:00:06.915578 train NMSE 0.87781 test NMSE 0.86298\n",
      "Model saved!\n",
      "step: 1999 time 0:00:06.938655 train NMSE 0.87803 test NMSE 0.88708\n",
      "step: 2099 time 0:00:06.927539 train NMSE 0.87852 test NMSE 0.87210\n",
      "step: 2199 time 0:00:06.948950 train NMSE 0.87734 test NMSE 0.87811\n",
      "step: 2299 time 0:00:06.944936 train NMSE 0.87616 test NMSE 0.86855\n",
      "step: 2399 time 0:00:06.941033 train NMSE 0.87563 test NMSE 0.87275\n",
      "step: 2499 time 0:00:06.940904 train NMSE 0.87608 test NMSE 0.86477\n",
      "step: 2599 time 0:00:06.944540 train NMSE 0.87571 test NMSE 0.87657\n",
      "step: 2699 time 0:00:06.942514 train NMSE 0.87509 test NMSE 0.88302\n",
      "step: 2799 time 0:00:06.918594 train NMSE 0.87573 test NMSE 0.85165\n",
      "Model saved!\n",
      "step: 2899 time 0:00:06.942012 train NMSE 0.87458 test NMSE 0.87270\n",
      "step: 2999 time 0:00:06.935809 train NMSE 0.87511 test NMSE 0.88076\n",
      "step: 99 time 0:00:07.729443 train NMSE 0.99707 test NMSE 0.99233\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.734358 train NMSE 0.98387 test NMSE 0.95894\n",
      "Model saved!\n",
      "step: 299 time 0:00:07.731334 train NMSE 0.93781 test NMSE 0.92080\n",
      "Model saved!\n",
      "step: 399 time 0:00:07.730380 train NMSE 0.90458 test NMSE 0.90096\n",
      "Model saved!\n",
      "step: 499 time 0:00:07.723856 train NMSE 0.89639 test NMSE 0.90900\n",
      "step: 599 time 0:00:07.729163 train NMSE 0.89259 test NMSE 0.87698\n",
      "Model saved!\n",
      "step: 699 time 0:00:07.726351 train NMSE 0.88913 test NMSE 0.89583\n",
      "step: 799 time 0:00:07.730414 train NMSE 0.89016 test NMSE 0.87515\n",
      "Model saved!\n",
      "step: 899 time 0:00:07.721403 train NMSE 0.88728 test NMSE 0.87259\n",
      "Model saved!\n",
      "step: 999 time 0:00:07.707938 train NMSE 0.88428 test NMSE 0.87351\n",
      "step: 1099 time 0:00:07.698813 train NMSE 0.88385 test NMSE 0.89964\n",
      "step: 1199 time 0:00:07.724409 train NMSE 0.88000 test NMSE 0.89066\n",
      "step: 1299 time 0:00:07.728398 train NMSE 0.87868 test NMSE 0.87466\n",
      "step: 1399 time 0:00:07.716397 train NMSE 0.88165 test NMSE 0.87908\n",
      "step: 1499 time 0:00:07.742347 train NMSE 0.87967 test NMSE 0.87479\n",
      "step: 1599 time 0:00:07.721412 train NMSE 0.87734 test NMSE 0.86052\n",
      "Model saved!\n",
      "step: 1699 time 0:00:07.720499 train NMSE 0.87747 test NMSE 0.88058\n",
      "step: 1799 time 0:00:07.723406 train NMSE 0.87654 test NMSE 0.88705\n",
      "step: 1899 time 0:00:07.739412 train NMSE 0.87656 test NMSE 0.88979\n",
      "step: 1999 time 0:00:07.731392 train NMSE 0.87666 test NMSE 0.89387\n",
      "step: 2099 time 0:00:07.726475 train NMSE 0.87503 test NMSE 0.87497\n",
      "step: 2199 time 0:00:07.729774 train NMSE 0.87480 test NMSE 0.86692\n",
      "step: 2299 time 0:00:07.730415 train NMSE 0.87482 test NMSE 0.88773\n",
      "step: 2399 time 0:00:07.727388 train NMSE 0.87460 test NMSE 0.87598\n",
      "step: 2499 time 0:00:07.727344 train NMSE 0.87466 test NMSE 0.85569\n",
      "Model saved!\n",
      "step: 2599 time 0:00:07.713963 train NMSE 0.87247 test NMSE 0.88116\n",
      "step: 2699 time 0:00:07.726345 train NMSE 0.87367 test NMSE 0.87723\n",
      "step: 2799 time 0:00:07.744109 train NMSE 0.87244 test NMSE 0.87327\n",
      "step: 2899 time 0:00:07.746460 train NMSE 0.87329 test NMSE 0.87588\n",
      "step: 2999 time 0:00:07.757259 train NMSE 0.87208 test NMSE 0.87338\n"
     ]
    }
   ],
   "source": [
    "L  = 16  #观测时隙数\n",
    "sigma = 1e5\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step: 99 time 0:00:04.591717 train NMSE 1.00459 test NMSE 1.00118\n",
      "Model saved!\n",
      "step: 199 time 0:00:04.584746 train NMSE 1.00024 test NMSE 0.99951\n",
      "Model saved!\n",
      "step: 299 time 0:00:04.564863 train NMSE 0.99878 test NMSE 0.99866\n",
      "Model saved!\n",
      "step: 399 time 0:00:04.559880 train NMSE 0.99806 test NMSE 0.99760\n",
      "Model saved!\n",
      "step: 499 time 0:00:04.567338 train NMSE 0.99758 test NMSE 0.99818\n",
      "step: 599 time 0:00:04.568813 train NMSE 0.99728 test NMSE 0.99694\n",
      "Model saved!\n",
      "step: 699 time 0:00:04.562345 train NMSE 0.99709 test NMSE 0.99732\n",
      "step: 799 time 0:00:04.574841 train NMSE 0.99705 test NMSE 0.99699\n",
      "step: 899 time 0:00:04.559384 train NMSE 0.99701 test NMSE 0.99680\n",
      "Model saved!\n",
      "step: 999 time 0:00:04.555860 train NMSE 0.99695 test NMSE 0.99697\n",
      "step: 1099 time 0:00:04.561350 train NMSE 0.99689 test NMSE 0.99706\n",
      "step: 1199 time 0:00:04.557892 train NMSE 0.99681 test NMSE 0.99691\n",
      "step: 1299 time 0:00:04.548922 train NMSE 0.99685 test NMSE 0.99659\n",
      "Model saved!\n",
      "step: 1399 time 0:00:04.559834 train NMSE 0.99680 test NMSE 0.99720\n",
      "step: 1499 time 0:00:04.567860 train NMSE 0.99682 test NMSE 0.99603\n",
      "Model saved!\n",
      "step: 1599 time 0:00:04.550496 train NMSE 0.99684 test NMSE 0.99722\n",
      "step: 1699 time 0:00:04.545882 train NMSE 0.99670 test NMSE 0.99658\n",
      "step: 1799 time 0:00:04.543415 train NMSE 0.99687 test NMSE 0.99683\n",
      "step: 1899 time 0:00:04.560771 train NMSE 0.99681 test NMSE 0.99712\n",
      "step: 1999 time 0:00:04.566816 train NMSE 0.99680 test NMSE 0.99656\n",
      "step: 2099 time 0:00:04.549839 train NMSE 0.99678 test NMSE 0.99713\n",
      "step: 2199 time 0:00:04.562848 train NMSE 0.99679 test NMSE 0.99754\n",
      "step: 2299 time 0:00:04.552855 train NMSE 0.99671 test NMSE 0.99729\n",
      "step: 2399 time 0:00:04.540907 train NMSE 0.99676 test NMSE 0.99684\n",
      "step: 2499 time 0:00:04.549977 train NMSE 0.99673 test NMSE 0.99641\n",
      "step: 2599 time 0:00:04.551885 train NMSE 0.99677 test NMSE 0.99685\n",
      "step: 2699 time 0:00:04.550902 train NMSE 0.99676 test NMSE 0.99679\n",
      "step: 2799 time 0:00:04.544389 train NMSE 0.99676 test NMSE 0.99708\n",
      "step: 2899 time 0:00:04.553979 train NMSE 0.99676 test NMSE 0.99687\n",
      "step: 2999 time 0:00:04.553315 train NMSE 0.99668 test NMSE 0.99738\n",
      "step: 99 time 0:00:05.354354 train NMSE 1.00555 test NMSE 1.00222\n",
      "Model saved!\n",
      "step: 199 time 0:00:05.335793 train NMSE 1.00068 test NMSE 0.99969\n",
      "Model saved!\n",
      "step: 299 time 0:00:05.353729 train NMSE 0.99902 test NMSE 0.99853\n",
      "Model saved!\n",
      "step: 399 time 0:00:05.367197 train NMSE 0.99817 test NMSE 0.99777\n",
      "Model saved!\n",
      "step: 499 time 0:00:05.345768 train NMSE 0.99763 test NMSE 0.99715\n",
      "Model saved!\n",
      "step: 599 time 0:00:05.366713 train NMSE 0.99737 test NMSE 0.99764\n",
      "step: 699 time 0:00:05.350837 train NMSE 0.99722 test NMSE 0.99817\n",
      "step: 799 time 0:00:05.350736 train NMSE 0.99710 test NMSE 0.99651\n",
      "Model saved!\n",
      "step: 899 time 0:00:05.352743 train NMSE 0.99699 test NMSE 0.99677\n",
      "step: 999 time 0:00:05.349750 train NMSE 0.99691 test NMSE 0.99721\n",
      "step: 1099 time 0:00:05.361605 train NMSE 0.99694 test NMSE 0.99704\n",
      "step: 1199 time 0:00:05.354757 train NMSE 0.99693 test NMSE 0.99784\n",
      "step: 1299 time 0:00:05.392337 train NMSE 0.99681 test NMSE 0.99644\n",
      "Model saved!\n",
      "step: 1399 time 0:00:05.356583 train NMSE 0.99683 test NMSE 0.99690\n",
      "step: 1499 time 0:00:05.366261 train NMSE 0.99686 test NMSE 0.99566\n",
      "Model saved!\n",
      "step: 1599 time 0:00:05.374421 train NMSE 0.99677 test NMSE 0.99656\n",
      "step: 1699 time 0:00:05.364001 train NMSE 0.99680 test NMSE 0.99611\n",
      "step: 1799 time 0:00:05.366724 train NMSE 0.99677 test NMSE 0.99686\n",
      "step: 1899 time 0:00:05.360743 train NMSE 0.99678 test NMSE 0.99679\n",
      "step: 1999 time 0:00:05.363757 train NMSE 0.99684 test NMSE 0.99712\n",
      "step: 2099 time 0:00:05.364584 train NMSE 0.99672 test NMSE 0.99699\n",
      "step: 2199 time 0:00:05.375524 train NMSE 0.99675 test NMSE 0.99726\n",
      "step: 2299 time 0:00:05.368711 train NMSE 0.99685 test NMSE 0.99640\n",
      "step: 2399 time 0:00:05.358331 train NMSE 0.99670 test NMSE 0.99617\n",
      "step: 2499 time 0:00:05.368794 train NMSE 0.99673 test NMSE 0.99704\n",
      "step: 2599 time 0:00:05.372742 train NMSE 0.99676 test NMSE 0.99717\n",
      "step: 2699 time 0:00:05.379669 train NMSE 0.99678 test NMSE 0.99702\n",
      "step: 2799 time 0:00:05.370128 train NMSE 0.99673 test NMSE 0.99623\n",
      "step: 2899 time 0:00:05.353778 train NMSE 0.99673 test NMSE 0.99587\n",
      "step: 2999 time 0:00:05.361246 train NMSE 0.99677 test NMSE 0.99658\n",
      "step: 99 time 0:00:06.135632 train NMSE 1.00561 test NMSE 1.00153\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.146666 train NMSE 1.00068 test NMSE 0.99976\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.130875 train NMSE 0.99897 test NMSE 0.99860\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.170568 train NMSE 0.99814 test NMSE 0.99810\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.149634 train NMSE 0.99757 test NMSE 0.99788\n",
      "Model saved!\n",
      "step: 599 time 0:00:06.159533 train NMSE 0.99739 test NMSE 0.99725\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.137468 train NMSE 0.99713 test NMSE 0.99739\n",
      "step: 799 time 0:00:06.146597 train NMSE 0.99714 test NMSE 0.99678\n",
      "Model saved!\n",
      "step: 899 time 0:00:06.137658 train NMSE 0.99699 test NMSE 0.99677\n",
      "Model saved!\n",
      "step: 999 time 0:00:06.146272 train NMSE 0.99694 test NMSE 0.99678\n",
      "step: 1099 time 0:00:06.149949 train NMSE 0.99693 test NMSE 0.99724\n",
      "step: 1199 time 0:00:06.164591 train NMSE 0.99688 test NMSE 0.99642\n",
      "Model saved!\n",
      "step: 1299 time 0:00:06.154598 train NMSE 0.99685 test NMSE 0.99649\n",
      "step: 1399 time 0:00:06.155554 train NMSE 0.99684 test NMSE 0.99711\n",
      "step: 1499 time 0:00:06.166058 train NMSE 0.99686 test NMSE 0.99659\n",
      "step: 1599 time 0:00:06.153558 train NMSE 0.99684 test NMSE 0.99652\n",
      "step: 1699 time 0:00:06.154570 train NMSE 0.99673 test NMSE 0.99610\n",
      "Model saved!\n",
      "step: 1799 time 0:00:06.160580 train NMSE 0.99679 test NMSE 0.99702\n",
      "step: 1899 time 0:00:06.149610 train NMSE 0.99672 test NMSE 0.99541\n",
      "Model saved!\n",
      "step: 1999 time 0:00:06.160596 train NMSE 0.99680 test NMSE 0.99670\n",
      "step: 2099 time 0:00:06.158093 train NMSE 0.99678 test NMSE 0.99704\n",
      "step: 2199 time 0:00:06.155601 train NMSE 0.99681 test NMSE 0.99647\n",
      "step: 2299 time 0:00:06.138236 train NMSE 0.99674 test NMSE 0.99630\n",
      "step: 2399 time 0:00:06.158553 train NMSE 0.99674 test NMSE 0.99650\n",
      "step: 2499 time 0:00:06.147151 train NMSE 0.99673 test NMSE 0.99686\n",
      "step: 2599 time 0:00:06.154620 train NMSE 0.99670 test NMSE 0.99673\n",
      "step: 2699 time 0:00:06.176070 train NMSE 0.99676 test NMSE 0.99683\n",
      "step: 2799 time 0:00:06.159604 train NMSE 0.99667 test NMSE 0.99677\n",
      "step: 2899 time 0:00:06.152606 train NMSE 0.99663 test NMSE 0.99692\n",
      "step: 2999 time 0:00:06.168572 train NMSE 0.99678 test NMSE 0.99698\n",
      "step: 99 time 0:00:06.942574 train NMSE 1.00559 test NMSE 1.00214\n",
      "Model saved!\n",
      "step: 199 time 0:00:06.937522 train NMSE 1.00067 test NMSE 0.99954\n",
      "Model saved!\n",
      "step: 299 time 0:00:06.944729 train NMSE 0.99894 test NMSE 0.99856\n",
      "Model saved!\n",
      "step: 399 time 0:00:06.952960 train NMSE 0.99815 test NMSE 0.99744\n",
      "Model saved!\n",
      "step: 499 time 0:00:06.939484 train NMSE 0.99772 test NMSE 0.99758\n",
      "step: 599 time 0:00:06.940456 train NMSE 0.99726 test NMSE 0.99726\n",
      "Model saved!\n",
      "step: 699 time 0:00:06.946970 train NMSE 0.99715 test NMSE 0.99709\n",
      "Model saved!\n",
      "step: 799 time 0:00:06.960402 train NMSE 0.99705 test NMSE 0.99792\n",
      "step: 899 time 0:00:06.939952 train NMSE 0.99699 test NMSE 0.99680\n",
      "Model saved!\n",
      "step: 999 time 0:00:06.955926 train NMSE 0.99686 test NMSE 0.99673\n",
      "Model saved!\n",
      "step: 1099 time 0:00:06.947442 train NMSE 0.99685 test NMSE 0.99665\n",
      "Model saved!\n",
      "step: 1199 time 0:00:06.944461 train NMSE 0.99685 test NMSE 0.99705\n",
      "step: 1299 time 0:00:06.947445 train NMSE 0.99689 test NMSE 0.99684\n",
      "step: 1399 time 0:00:06.956478 train NMSE 0.99681 test NMSE 0.99704\n",
      "step: 1499 time 0:00:06.928500 train NMSE 0.99678 test NMSE 0.99732\n",
      "step: 1599 time 0:00:06.943502 train NMSE 0.99684 test NMSE 0.99694\n",
      "step: 1699 time 0:00:06.940782 train NMSE 0.99679 test NMSE 0.99663\n",
      "Model saved!\n",
      "step: 1799 time 0:00:06.939494 train NMSE 0.99688 test NMSE 0.99661\n",
      "Model saved!\n",
      "step: 1899 time 0:00:06.942989 train NMSE 0.99682 test NMSE 0.99659\n",
      "Model saved!\n",
      "step: 1999 time 0:00:06.951470 train NMSE 0.99686 test NMSE 0.99702\n",
      "step: 2099 time 0:00:06.947486 train NMSE 0.99677 test NMSE 0.99695\n",
      "step: 2199 time 0:00:06.947253 train NMSE 0.99682 test NMSE 0.99684\n",
      "step: 2299 time 0:00:06.909577 train NMSE 0.99683 test NMSE 0.99649\n",
      "Model saved!\n",
      "step: 2399 time 0:00:06.939459 train NMSE 0.99676 test NMSE 0.99683\n",
      "step: 2499 time 0:00:06.951971 train NMSE 0.99679 test NMSE 0.99654\n",
      "step: 2599 time 0:00:06.927522 train NMSE 0.99676 test NMSE 0.99722\n",
      "step: 2699 time 0:00:06.932049 train NMSE 0.99673 test NMSE 0.99620\n",
      "Model saved!\n",
      "step: 2799 time 0:00:06.938987 train NMSE 0.99671 test NMSE 0.99662\n",
      "step: 2899 time 0:00:06.936515 train NMSE 0.99669 test NMSE 0.99659\n",
      "step: 2999 time 0:00:06.978414 train NMSE 0.99669 test NMSE 0.99661\n",
      "step: 99 time 0:00:07.727428 train NMSE 1.00544 test NMSE 1.00254\n",
      "Model saved!\n",
      "step: 199 time 0:00:07.742359 train NMSE 1.00061 test NMSE 0.99950\n",
      "Model saved!\n",
      "step: 299 time 0:00:07.731387 train NMSE 0.99891 test NMSE 0.99880\n",
      "Model saved!\n",
      "step: 399 time 0:00:07.742893 train NMSE 0.99810 test NMSE 0.99726\n",
      "Model saved!\n",
      "step: 499 time 0:00:07.755814 train NMSE 0.99767 test NMSE 0.99777\n",
      "step: 599 time 0:00:07.745845 train NMSE 0.99733 test NMSE 0.99731\n",
      "step: 699 time 0:00:07.734927 train NMSE 0.99718 test NMSE 0.99758\n",
      "step: 799 time 0:00:07.721935 train NMSE 0.99706 test NMSE 0.99693\n",
      "Model saved!\n",
      "step: 899 time 0:00:07.718152 train NMSE 0.99701 test NMSE 0.99707\n",
      "step: 999 time 0:00:07.720929 train NMSE 0.99692 test NMSE 0.99706\n",
      "step: 1099 time 0:00:07.720079 train NMSE 0.99690 test NMSE 0.99709\n",
      "step: 1199 time 0:00:07.732908 train NMSE 0.99689 test NMSE 0.99740\n",
      "step: 1299 time 0:00:07.728926 train NMSE 0.99685 test NMSE 0.99683\n",
      "Model saved!\n",
      "step: 1399 time 0:00:07.725405 train NMSE 0.99680 test NMSE 0.99729\n",
      "step: 1499 time 0:00:07.719898 train NMSE 0.99684 test NMSE 0.99628\n",
      "Model saved!\n",
      "step: 1599 time 0:00:07.731397 train NMSE 0.99681 test NMSE 0.99703\n",
      "step: 1699 time 0:00:07.740981 train NMSE 0.99685 test NMSE 0.99683\n",
      "step: 1799 time 0:00:07.730899 train NMSE 0.99677 test NMSE 0.99740\n",
      "step: 1899 time 0:00:07.735869 train NMSE 0.99682 test NMSE 0.99658\n",
      "step: 1999 time 0:00:07.726970 train NMSE 0.99681 test NMSE 0.99718\n",
      "step: 2099 time 0:00:07.692511 train NMSE 0.99682 test NMSE 0.99676\n",
      "step: 2199 time 0:00:07.701520 train NMSE 0.99679 test NMSE 0.99706\n",
      "step: 2299 time 0:00:07.726958 train NMSE 0.99682 test NMSE 0.99653\n",
      "step: 2399 time 0:00:07.725956 train NMSE 0.99673 test NMSE 0.99690\n",
      "step: 2499 time 0:00:07.723965 train NMSE 0.99681 test NMSE 0.99655\n",
      "step: 2599 time 0:00:07.716957 train NMSE 0.99680 test NMSE 0.99630\n",
      "step: 2699 time 0:00:07.727067 train NMSE 0.99672 test NMSE 0.99695\n",
      "step: 2799 time 0:00:07.728549 train NMSE 0.99677 test NMSE 0.99711\n",
      "step: 2899 time 0:00:07.728231 train NMSE 0.99674 test NMSE 0.99694\n",
      "step: 2999 time 0:00:07.732919 train NMSE 0.99666 test NMSE 0.99695\n"
     ]
    }
   ],
   "source": [
    "L  = 16  #观测时隙数\n",
    "sigma = 1e6\n",
    "K = 4 #用户数\n",
    "param_list = [fc,BW,Nc,M,N,D_sub,D_ant,R,M_ant,h1,h2,L,sigma,B,K,Lp]\n",
    "CE_LAMP_GMMV(param_list,L,sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "6752431b5c56664d9e32492cd42b0f7944f43cec6b3b38a08056858cf10fc2ca"
  },
  "kernelspec": {
   "display_name": "Python 3.7.4 ('base')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}