lab_7.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dIrdWn2Xm3rq",
        "outputId": "d61e8aa4-97dd-472c-bbf1-a380142af347"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11490434/11490434 [==============================] - 0s 0us/step\n"
          ]
        }
      ],
      "source": [
        "from keras.datasets import mnist\n",
        "import numpy as np\n",
        "\n",
        "#podział na zbiór testowy i uczący\n",
        "train, test = mnist.load_data()\n",
        "X_train, y_train = train[0], train[1]\n",
        "X_test, y_test = test[0], test[1]\n",
        "X_train = np.expand_dims(X_train, axis=-1)\n",
        "X_test = np.expand_dims(X_test, axis=-1)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "RC63op-lml5v"
      },
      "outputs": [],
      "source": [
        "\"\"\"\n",
        "Zadanie 7.2. Znalezienie optymalnej liczby warstw\n",
        "Stosując Listing 7.1 napisz skrypt, który pozwoli na znalezienie optymalnej liczby\n",
        "warstw. Jako optymalną rozumiemy liczbę, przy której jakość działania modelu będzie\n",
        "najlepsza, a liczba warstw nie będzie za duża.\n",
        "\"\"\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wh4M77qipjkH"
      },
      "outputs": [],
      "source": [
        "#funkcja budująca sieć konwolusyjną\n",
        "from keras.layers import Conv2D, Flatten, Dense,MaxPooling2D\n",
        "from keras.models import Sequential\n",
        "from keras.optimizers import Adam\n",
        "\n",
        "\n",
        "def buildConvo(layer_num=1):\n",
        "    #papametry sieci konwolucyjnej\n",
        "    class_cnt = np.unique(y_train).shape[0]\n",
        "    filter_cnt = 32\n",
        "    neuron_cnt = 32\n",
        "    learning_rate = 0.0001\n",
        "    act_func = 'relu'\n",
        "    kernel_size = (3,3)\n",
        "    model = Sequential()\n",
        "    conv_rule = 'same'\n",
        "    pooling_size=(2,2)\n",
        "    \n",
        "    #dodanie warstw do sieci konwolucyjnej\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    for a in range(layer_num):\n",
        "      model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    \n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(class_cnt, activation='softmax'))\n",
        "    model.compile(optimizer=Adam(learning_rate),loss='SparseCategoricalCrossentropy',metrics='accuracy')\n",
        "    return model\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qlsST7hcqi4q",
        "outputId": "f59daa95-343b-4e73-e37f-b2f868dc59c4"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "<ipython-input-3-00fc6b5fe9cb>:5: DeprecationWarning: KerasClassifier is deprecated, use Sci-Keras (https://github.com/adriangb/scikeras) instead. See https://www.adriangb.com/scikeras/stable/migration.html for help migrating.\n",
            "  keras_classifier = KerasClassifier(buildConvo)\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 11s 3ms/step - loss: 1.1206 - accuracy: 0.8974\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.2525 - accuracy: 0.9597\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.1308 - accuracy: 0.9746\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0770 - accuracy: 0.9822\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0577 - accuracy: 0.9859\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0395 - accuracy: 0.9896\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0305 - accuracy: 0.9919\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0252 - accuracy: 0.9927\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0190 - accuracy: 0.9939\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0144 - accuracy: 0.9951\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.2151 - accuracy: 0.9711\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 1.2709 - accuracy: 0.8930\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.3011 - accuracy: 0.9566\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.1531 - accuracy: 0.9725\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0924 - accuracy: 0.9804\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0618 - accuracy: 0.9847\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0419 - accuracy: 0.9890\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0323 - accuracy: 0.9911\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0242 - accuracy: 0.9929\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0163 - accuracy: 0.9946\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0125 - accuracy: 0.9962\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.1905 - accuracy: 0.9711\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 1.1880 - accuracy: 0.8953\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.2764 - accuracy: 0.9580\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.1452 - accuracy: 0.9729\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0940 - accuracy: 0.9804\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0635 - accuracy: 0.9847\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0438 - accuracy: 0.9886\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0348 - accuracy: 0.9905\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0248 - accuracy: 0.9928\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0179 - accuracy: 0.9947\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0198 - accuracy: 0.9941\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.2261 - accuracy: 0.9728\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 1.5106 - accuracy: 0.8906\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.3610 - accuracy: 0.9552\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.1826 - accuracy: 0.9702\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.1173 - accuracy: 0.9784\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0792 - accuracy: 0.9836\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0556 - accuracy: 0.9870\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0392 - accuracy: 0.9905\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0293 - accuracy: 0.9918\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0259 - accuracy: 0.9929\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0157 - accuracy: 0.9959\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.2385 - accuracy: 0.9690\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 1.3504 - accuracy: 0.8840\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.3101 - accuracy: 0.9555\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.1611 - accuracy: 0.9702\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0990 - accuracy: 0.9793\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0761 - accuracy: 0.9830\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0497 - accuracy: 0.9872\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0354 - accuracy: 0.9899\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0282 - accuracy: 0.9920\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0222 - accuracy: 0.9932\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0177 - accuracy: 0.9949\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.2086 - accuracy: 0.9761\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.4877 - accuracy: 0.9187\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.0730 - accuracy: 0.9784\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.0362 - accuracy: 0.9887\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.0238 - accuracy: 0.9922\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.0167 - accuracy: 0.9947\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.0125 - accuracy: 0.9958\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0088 - accuracy: 0.9970\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0073 - accuracy: 0.9976\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0040 - accuracy: 0.9988\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0056 - accuracy: 0.9980\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.1059 - accuracy: 0.9803\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.5394 - accuracy: 0.9210\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0879 - accuracy: 0.9746\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0460 - accuracy: 0.9851\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0275 - accuracy: 0.9909\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0219 - accuracy: 0.9924\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0143 - accuracy: 0.9948\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0103 - accuracy: 0.9962\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0084 - accuracy: 0.9972\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0067 - accuracy: 0.9976\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0054 - accuracy: 0.9982\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.0969 - accuracy: 0.9801\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.5938 - accuracy: 0.9168\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.0875 - accuracy: 0.9748\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0447 - accuracy: 0.9857\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0271 - accuracy: 0.9910\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0214 - accuracy: 0.9926\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0128 - accuracy: 0.9955\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0105 - accuracy: 0.9963\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0095 - accuracy: 0.9965\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0061 - accuracy: 0.9980\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0059 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.1146 - accuracy: 0.9788\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.4923 - accuracy: 0.9212\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0691 - accuracy: 0.9796\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0343 - accuracy: 0.9893\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0212 - accuracy: 0.9925\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0142 - accuracy: 0.9950\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0120 - accuracy: 0.9957\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0079 - accuracy: 0.9973\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0066 - accuracy: 0.9977\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0067 - accuracy: 0.9976\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0040 - accuracy: 0.9986\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.1475 - accuracy: 0.9752\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 5s 3ms/step - loss: 0.5120 - accuracy: 0.9193\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0768 - accuracy: 0.9780\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0391 - accuracy: 0.9876\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0236 - accuracy: 0.9922\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0167 - accuracy: 0.9945\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0116 - accuracy: 0.9962\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0089 - accuracy: 0.9969\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0076 - accuracy: 0.9972\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0068 - accuracy: 0.9975\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0041 - accuracy: 0.9985\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.1015 - accuracy: 0.9830\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.3963 - accuracy: 0.9219\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0685 - accuracy: 0.9791\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0359 - accuracy: 0.9885\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0222 - accuracy: 0.9925\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0164 - accuracy: 0.9947\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0124 - accuracy: 0.9954\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0101 - accuracy: 0.9966\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0061 - accuracy: 0.9979\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0058 - accuracy: 0.9981\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0071 - accuracy: 0.9975\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.0957 - accuracy: 0.9825\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.3934 - accuracy: 0.9269\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0674 - accuracy: 0.9794\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0400 - accuracy: 0.9880\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0247 - accuracy: 0.9916\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0165 - accuracy: 0.9948\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0137 - accuracy: 0.9956\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0105 - accuracy: 0.9962\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0082 - accuracy: 0.9973\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0048 - accuracy: 0.9983\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0066 - accuracy: 0.9977\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.0813 - accuracy: 0.9840\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.3008 - accuracy: 0.9330\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0593 - accuracy: 0.9816\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0318 - accuracy: 0.9897\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0203 - accuracy: 0.9933\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0136 - accuracy: 0.9954\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0097 - accuracy: 0.9967\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0072 - accuracy: 0.9976\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0070 - accuracy: 0.9975\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0040 - accuracy: 0.9987\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0049 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.1058 - accuracy: 0.9806\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.3765 - accuracy: 0.9264\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0631 - accuracy: 0.9811\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0337 - accuracy: 0.9893\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0229 - accuracy: 0.9924\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0151 - accuracy: 0.9951\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0108 - accuracy: 0.9964\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0081 - accuracy: 0.9972\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0080 - accuracy: 0.9974\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0039 - accuracy: 0.9989\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0051 - accuracy: 0.9984\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0940 - accuracy: 0.9812\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.5117 - accuracy: 0.9152\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0766 - accuracy: 0.9768\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0399 - accuracy: 0.9876\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 5s 4ms/step - loss: 0.0255 - accuracy: 0.9919\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0171 - accuracy: 0.9943\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0145 - accuracy: 0.9952\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0103 - accuracy: 0.9964\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0081 - accuracy: 0.9972\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0060 - accuracy: 0.9981\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 6s 4ms/step - loss: 0.0053 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.0830 - accuracy: 0.9853\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.2341 - accuracy: 0.9364\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 7s 4ms/step - loss: 0.0554 - accuracy: 0.9828\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 7s 4ms/step - loss: 0.0318 - accuracy: 0.9898\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 7s 4ms/step - loss: 0.0203 - accuracy: 0.9933\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 7s 4ms/step - loss: 0.0134 - accuracy: 0.9954\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0106 - accuracy: 0.9964\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 7s 4ms/step - loss: 0.0076 - accuracy: 0.9975\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0074 - accuracy: 0.9975\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0069 - accuracy: 0.9976\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0037 - accuracy: 0.9987\n",
            "375/375 [==============================] - 1s 2ms/step - loss: 0.0679 - accuracy: 0.9858\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.2152 - accuracy: 0.9402\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0521 - accuracy: 0.9838\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0281 - accuracy: 0.9916\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0187 - accuracy: 0.9939\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0129 - accuracy: 0.9957\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0086 - accuracy: 0.9973\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0087 - accuracy: 0.9973\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0048 - accuracy: 0.9984\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0050 - accuracy: 0.9983\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0051 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0685 - accuracy: 0.9841\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.2265 - accuracy: 0.9360\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0522 - accuracy: 0.9837\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0292 - accuracy: 0.9911\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0176 - accuracy: 0.9948\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0128 - accuracy: 0.9959\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0094 - accuracy: 0.9971\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0066 - accuracy: 0.9979\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0061 - accuracy: 0.9980\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0038 - accuracy: 0.9989\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0057 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0639 - accuracy: 0.9862\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.2255 - accuracy: 0.9416\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0491 - accuracy: 0.9846\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0299 - accuracy: 0.9902\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0187 - accuracy: 0.9938\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0140 - accuracy: 0.9954\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0115 - accuracy: 0.9963\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0085 - accuracy: 0.9972\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0062 - accuracy: 0.9978\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0052 - accuracy: 0.9984\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0042 - accuracy: 0.9985\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.1022 - accuracy: 0.9810\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.2321 - accuracy: 0.9345\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0558 - accuracy: 0.9829\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0314 - accuracy: 0.9897\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0197 - accuracy: 0.9934\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0144 - accuracy: 0.9951\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0098 - accuracy: 0.9966\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0082 - accuracy: 0.9972\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0056 - accuracy: 0.9983\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0055 - accuracy: 0.9983\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 7s 5ms/step - loss: 0.0038 - accuracy: 0.9988\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0720 - accuracy: 0.9844\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 9s 5ms/step - loss: 0.1862 - accuracy: 0.9436\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0514 - accuracy: 0.9840\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0303 - accuracy: 0.9902\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0207 - accuracy: 0.9934\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0153 - accuracy: 0.9949\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0112 - accuracy: 0.9959\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0087 - accuracy: 0.9968\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0058 - accuracy: 0.9980\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0057 - accuracy: 0.9979\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0037 - accuracy: 0.9989\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0713 - accuracy: 0.9858\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 9s 5ms/step - loss: 0.2130 - accuracy: 0.9407\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0551 - accuracy: 0.9832\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0309 - accuracy: 0.9906\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0196 - accuracy: 0.9934\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0149 - accuracy: 0.9950\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0090 - accuracy: 0.9971\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0096 - accuracy: 0.9970\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0067 - accuracy: 0.9977\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0045 - accuracy: 0.9986\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0058 - accuracy: 0.9981\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0608 - accuracy: 0.9850\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 9s 5ms/step - loss: 0.2073 - accuracy: 0.9395\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 8s 6ms/step - loss: 0.0502 - accuracy: 0.9845\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0286 - accuracy: 0.9913\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0186 - accuracy: 0.9942\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0139 - accuracy: 0.9954\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0110 - accuracy: 0.9965\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0080 - accuracy: 0.9974\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0061 - accuracy: 0.9979\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0070 - accuracy: 0.9977\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0042 - accuracy: 0.9984\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0780 - accuracy: 0.9842\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.1811 - accuracy: 0.9450\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0487 - accuracy: 0.9848\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0290 - accuracy: 0.9910\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0202 - accuracy: 0.9936\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0145 - accuracy: 0.9952\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0092 - accuracy: 0.9969\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0072 - accuracy: 0.9976\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0059 - accuracy: 0.9981\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0051 - accuracy: 0.9985\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0036 - accuracy: 0.9988\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0916 - accuracy: 0.9821\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 9s 5ms/step - loss: 0.2590 - accuracy: 0.9358\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0590 - accuracy: 0.9817\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0367 - accuracy: 0.9885\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0240 - accuracy: 0.9923\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0170 - accuracy: 0.9945\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0139 - accuracy: 0.9952\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0100 - accuracy: 0.9969\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 8s 5ms/step - loss: 0.0090 - accuracy: 0.9969\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 8s 6ms/step - loss: 0.0073 - accuracy: 0.9975\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 8s 6ms/step - loss: 0.0047 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0747 - accuracy: 0.9858\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.1835 - accuracy: 0.9442\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0509 - accuracy: 0.9845\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0303 - accuracy: 0.9898\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0210 - accuracy: 0.9938\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0157 - accuracy: 0.9947\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0112 - accuracy: 0.9964\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0086 - accuracy: 0.9969\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0071 - accuracy: 0.9976\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0068 - accuracy: 0.9976\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0045 - accuracy: 0.9984\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0775 - accuracy: 0.9850\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.1702 - accuracy: 0.9482\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0508 - accuracy: 0.9840\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0317 - accuracy: 0.9906\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0211 - accuracy: 0.9930\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0167 - accuracy: 0.9946\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0103 - accuracy: 0.9968\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0093 - accuracy: 0.9968\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0087 - accuracy: 0.9972\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0070 - accuracy: 0.9975\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0053 - accuracy: 0.9981\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0614 - accuracy: 0.9851\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.1601 - accuracy: 0.9507\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0454 - accuracy: 0.9860\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0274 - accuracy: 0.9912\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0197 - accuracy: 0.9937\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0126 - accuracy: 0.9960\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0108 - accuracy: 0.9961\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0069 - accuracy: 0.9977\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0066 - accuracy: 0.9976\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0055 - accuracy: 0.9981\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0046 - accuracy: 0.9984\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0621 - accuracy: 0.9857\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.1638 - accuracy: 0.9498\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0487 - accuracy: 0.9853\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0307 - accuracy: 0.9903\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0192 - accuracy: 0.9938\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0148 - accuracy: 0.9949\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0098 - accuracy: 0.9967\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0074 - accuracy: 0.9975\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0065 - accuracy: 0.9980\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0061 - accuracy: 0.9981\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0056 - accuracy: 0.9980\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0647 - accuracy: 0.9845\n",
            "Epoch 1/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.1706 - accuracy: 0.9462\n",
            "Epoch 2/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0485 - accuracy: 0.9848\n",
            "Epoch 3/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0303 - accuracy: 0.9901\n",
            "Epoch 4/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0199 - accuracy: 0.9937\n",
            "Epoch 5/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0141 - accuracy: 0.9953\n",
            "Epoch 6/10\n",
            "1500/1500 [==============================] - 10s 6ms/step - loss: 0.0105 - accuracy: 0.9968\n",
            "Epoch 7/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0084 - accuracy: 0.9973\n",
            "Epoch 8/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0073 - accuracy: 0.9976\n",
            "Epoch 9/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0050 - accuracy: 0.9985\n",
            "Epoch 10/10\n",
            "1500/1500 [==============================] - 9s 6ms/step - loss: 0.0053 - accuracy: 0.9983\n",
            "375/375 [==============================] - 1s 3ms/step - loss: 0.0571 - accuracy: 0.9875\n",
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.1559 - accuracy: 0.9547\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0462 - accuracy: 0.9859\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0304 - accuracy: 0.9901\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0202 - accuracy: 0.9933\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0138 - accuracy: 0.9955\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0108 - accuracy: 0.9964\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0073 - accuracy: 0.9975\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0061 - accuracy: 0.9978\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0060 - accuracy: 0.9982\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0038 - accuracy: 0.9987\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "{'layer_num': 5}"
            ]
          },
          "execution_count": 3,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "#sprawdzenie najlepszej ilości warstw w sieci konwolucyjnej\n",
        "from keras.wrappers.scikit_learn import KerasClassifier #depricated\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "keras_classifier = KerasClassifier(buildConvo)\n",
        "param_distribs = {\n",
        "    'layer_num':[0,1,2,3,4,5]\n",
        "}\n",
        "\n",
        "grid_search_cv = GridSearchCV(keras_classifier,param_distribs)\n",
        "grid_result = grid_search_cv.fit(X_train,y_train,epochs=10)\n",
        "best_params_from_grid = grid_search_cv.best_params_\n",
        "best_model_from_grid = grid_search_cv.best_estimator_\n",
        "\n",
        "#najlepsze parametry znalezione\n",
        "best_params_from_grid\n",
        "\n",
        "\"\"\"\n",
        "Epoch 8/10\n",
        "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0061 - accuracy: 0.9978\n",
        "Epoch 9/10\n",
        "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0060 - accuracy: 0.9982\n",
        "Epoch 10/10\n",
        "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0038 - accuracy: 0.9987\n",
        "{'layer_num': 5}\n",
        "\"\"\""
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U2xUxcbvx7tz"
      },
      "source": [
        "### **Wnioski zad. 7.2**\n",
        "Optymalna liczba warstw dla modelu wyznaczona przy pomocy GridSearch wyniosła 5 warstw ukrytych."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "t9dTj2g1sL_m"
      },
      "outputs": [],
      "source": [
        "\"\"\"\n",
        "Zadanie 7.3. Wpływ warstw głosujących\n",
        "Zbadaj wpływ warstw głosujących. Po każdej warstwie konwolucyjnej dodaj warstwę\n",
        "głosującą. Porównaj krzywe uczenia uzyskane z zastosowaniem warstw głosujących oraz\n",
        "bez nich. Przetestuj dwa rodzaje warstw: MaxPooling oraz AveragePooling.\n",
        "\"\"\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "t4PtjXoXtbNM"
      },
      "outputs": [],
      "source": [
        "from keras.layers import Conv2D, Flatten, Dense,MaxPooling2D\n",
        "from keras.models import Sequential\n",
        "from keras.optimizers import Adam\n",
        "from keras.layers import Conv2D, Flatten,Dense, AveragePooling2D, MaxPooling2D\n",
        "\n",
        "# bez warstw głowujących\n",
        "def buildConvoNoPool():\n",
        "    class_cnt = np.unique(y_train).shape[0]\n",
        "    filter_cnt = 32\n",
        "    neuron_cnt = 32\n",
        "    learning_rate = 0.0001\n",
        "    act_func = 'relu'\n",
        "    kernel_size = (3,3)\n",
        "    model = Sequential()\n",
        "    conv_rule = 'same'\n",
        "    pooling_size=(2,2)\n",
        "\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    \n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    \n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(class_cnt, activation='softmax'))\n",
        "    model.compile(optimizer=Adam(learning_rate),loss='SparseCategoricalCrossentropy',metrics='accuracy')\n",
        "    return model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "h6HnMfJ5tebs"
      },
      "outputs": [],
      "source": [
        "from keras.layers import Conv2D, Flatten, Dense,MaxPooling2D\n",
        "from keras.models import Sequential\n",
        "from keras.optimizers import Adam\n",
        "from keras.layers import Conv2D, Flatten,Dense, AveragePooling2D, MaxPooling2D\n",
        "\n",
        "# z warstwą głosującą MaxPooling\n",
        "def buildConvoMaxPool():\n",
        "    class_cnt = np.unique(y_train).shape[0]\n",
        "    filter_cnt = 32\n",
        "    neuron_cnt = 32\n",
        "    learning_rate = 0.0001\n",
        "    act_func = 'relu'\n",
        "    kernel_size = (3,3)\n",
        "    model = Sequential()\n",
        "    conv_rule = 'same'\n",
        "    pooling_size=(2,2)\n",
        "\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    \n",
        "    model.add(MaxPooling2D(pooling_size))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(MaxPooling2D(pooling_size))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(MaxPooling2D(pooling_size))\n",
        "\n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(class_cnt, activation='softmax'))\n",
        "    model.compile(optimizer=Adam(learning_rate),loss='SparseCategoricalCrossentropy',metrics='accuracy')\n",
        "    return model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "_vCAS2ITtiCO"
      },
      "outputs": [],
      "source": [
        "from keras.layers import Conv2D, Flatten, Dense,MaxPooling2D\n",
        "from keras.models import Sequential\n",
        "from keras.optimizers import Adam\n",
        "from keras.layers import Conv2D, Flatten,Dense, AveragePooling2D, MaxPooling2D\n",
        "\n",
        "# z warstwą głowującą AveragePooling\n",
        "def buildConvoAvgPool():\n",
        "    class_cnt = np.unique(y_train).shape[0]\n",
        "    filter_cnt = 32\n",
        "    neuron_cnt = 32\n",
        "    learning_rate = 0.0001\n",
        "    act_func = 'relu'\n",
        "    kernel_size = (3,3)\n",
        "    model = Sequential()\n",
        "    conv_rule = 'same'\n",
        "    pooling_size=(2,2)\n",
        "\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    \n",
        "    model.add(AveragePooling2D(pooling_size))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(AveragePooling2D(pooling_size))\n",
        "    model.add(Conv2D(input_shape = X_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    model.add(AveragePooling2D(pooling_size))\n",
        "\n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(class_cnt, activation='softmax'))\n",
        "    model.compile(optimizer=Adam(learning_rate),loss='SparseCategoricalCrossentropy',metrics='accuracy')\n",
        "    return model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "-cRwC3Jpj3mY"
      },
      "outputs": [],
      "source": [
        "#funkcja rysująca wykres uczenia dla danego modelu\n",
        "def generateNetworkLearningCurve(model,epchs=10):\n",
        "  from matplotlib import pyplot as plt\n",
        "  historia = model.history.history\n",
        "  floss_train = historia['loss']\n",
        "  floss_test = historia['val_loss']\n",
        "  acc_train = historia['accuracy']\n",
        "  acc_test = historia['val_accuracy']\n",
        "  fig,ax = plt.subplots(1,2, figsize=(20,10))\n",
        "  epochs = np.arange(0, epchs)\n",
        "  ax[0].plot(epochs, floss_train, label='floss_train')\n",
        "  ax[0].plot(epochs, floss_test, label='floss_test')\n",
        "  ax[0].set_title('Funkcje strat')\n",
        "  ax[0].legend()\n",
        "  ax[1].set_title('Dokladnosci')\n",
        "  ax[1].plot(epochs, acc_train, label='acc_train')\n",
        "  ax[1].plot(epochs, acc_test, label='acc_test')\n",
        "  ax[1].legend()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "YxeHq0T3wT2T",
        "outputId": "0e6e3601-177d-4615-a182-e088cd78f415"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 13s 6ms/step - loss: 0.1684 - accuracy: 0.9513 - val_loss: 0.0665 - val_accuracy: 0.9790\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0483 - accuracy: 0.9852 - val_loss: 0.0423 - val_accuracy: 0.9865\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0322 - accuracy: 0.9902 - val_loss: 0.0385 - val_accuracy: 0.9872\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 12s 7ms/step - loss: 0.0222 - accuracy: 0.9927 - val_loss: 0.0364 - val_accuracy: 0.9888\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0156 - accuracy: 0.9948 - val_loss: 0.0440 - val_accuracy: 0.9867\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 12s 7ms/step - loss: 0.0108 - accuracy: 0.9967 - val_loss: 0.0393 - val_accuracy: 0.9888\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0087 - accuracy: 0.9970 - val_loss: 0.0490 - val_accuracy: 0.9875\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0066 - accuracy: 0.9978 - val_loss: 0.0592 - val_accuracy: 0.9857\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0056 - accuracy: 0.9983 - val_loss: 0.0422 - val_accuracy: 0.9904\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0054 - accuracy: 0.9984 - val_loss: 0.0559 - val_accuracy: 0.9857\n",
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 8s 4ms/step - loss: 1.4510 - accuracy: 0.7772 - val_loss: 0.3171 - val_accuracy: 0.9145\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2563 - accuracy: 0.9312 - val_loss: 0.1841 - val_accuracy: 0.9459\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1528 - accuracy: 0.9560 - val_loss: 0.1410 - val_accuracy: 0.9621\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1082 - accuracy: 0.9690 - val_loss: 0.0970 - val_accuracy: 0.9727\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0821 - accuracy: 0.9749 - val_loss: 0.1030 - val_accuracy: 0.9699\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0628 - accuracy: 0.9802 - val_loss: 0.0740 - val_accuracy: 0.9791\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0506 - accuracy: 0.9844 - val_loss: 0.0715 - val_accuracy: 0.9799\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0409 - accuracy: 0.9867 - val_loss: 0.0625 - val_accuracy: 0.9819\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0341 - accuracy: 0.9891 - val_loss: 0.0590 - val_accuracy: 0.9830\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0278 - accuracy: 0.9910 - val_loss: 0.0638 - val_accuracy: 0.9818\n",
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.4062 - accuracy: 0.8810 - val_loss: 0.1293 - val_accuracy: 0.9606\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1069 - accuracy: 0.9676 - val_loss: 0.0793 - val_accuracy: 0.9751\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0759 - accuracy: 0.9765 - val_loss: 0.0585 - val_accuracy: 0.9806\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0615 - accuracy: 0.9807 - val_loss: 0.0472 - val_accuracy: 0.9836\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0510 - accuracy: 0.9837 - val_loss: 0.0428 - val_accuracy: 0.9856\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0449 - accuracy: 0.9860 - val_loss: 0.0371 - val_accuracy: 0.9882\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0401 - accuracy: 0.9882 - val_loss: 0.0402 - val_accuracy: 0.9865\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0367 - accuracy: 0.9889 - val_loss: 0.0371 - val_accuracy: 0.9879\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0330 - accuracy: 0.9904 - val_loss: 0.0315 - val_accuracy: 0.9900\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0306 - accuracy: 0.9905 - val_loss: 0.0324 - val_accuracy: 0.9888\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1440x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1440x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABIcAAAJOCAYAAAAtRmfdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd3yV5f3/8deVnUOSk5wkjCwIQwTCRkFwVb8OHFXBauuoo5Za56/DOqq1arVarXWX0n61/TrqorbWLYjiQJQgK4BsyWBlJ2TnXL8/7jvJIYQhJDkH8n4+Hnkk59zrkyPH3Hnnuj6XsdYiIiIiIiIiIiI9U1iwCxARERERERERkeBROCQiIiIiIiIi0oMpHBIRERERERER6cEUDomIiIiIiIiI9GAKh0REREREREREejCFQyIiIiIiIiIiPZjCIRE5KMaY3xpjnjuA46qNMQO7oiYRERGRw40x5nJjzCffdtt+nPdEY0zBwVV3YIwxWe49YXgwri8ibRQOifQgxphNxpha94dwy0daMGqx1sZZazd05jmNMdYYM/ggz/GhMeaqzqpJREREpEXAvViVMabcGPOZMeZqY0yP/L3MWrvZvSdsDnYtIj1dj/yfkEgPd7b7Q7jloyjYBXUXY0xEsGsQERGRHu9sa2080B+4H7gZ+N/gliQiPZ3CIRFp+SvW/wQ8bp0qZowZ4I7IucwYs9kYU2yM+fUezhNpjPmnMWa2MSbKGBNujLnNGLPe/QtZrjEm0923dZSPMSbaGPOQe/5txpiZxpjYPVxjsDHmI2NMhVvLS+7z891dlrojoi5sGSZtjLnZGLMVeMYYk2SMecMYs8MYU+Z+neGe417gOOAJ9xxPdM4rLCIiIrIra22FtfZ14ELgMmNMjjHGa4z5P/c+5RtjzO17GlVkjHnQGPOJMcbbwbZHjTH5xphK9/7ruIBtscaYv7v3QSuBo9odu8kY80tjzDL3fuslY0xMwPYfG2PWGWNKjTGvt4xCN44/GWO2u9ddbozJCbjmH93vqcKtOzbgPlN/wBMJMoVDIrK/jgWGAicDvzHGDAvc6IY5/wbqgQustQ3Az4EfAGcACcCVQE0H574fOAIYAwwG0oHf7KGOe4D3gCQgA3gcwFp7vLt9tDsi6iX3cV/Ah/PXuRk4/997xn2cBdQCT7jn+DXwMXCde47r9ueFERERETlQ1tovgAKcP1A9DniBgcAJwA+BKwL3N8aEGWP+CowCTrXWVnRw2i9x7qt8wAvAKwEBz53AIPfjNOCyDo6/ADgdyHavc7l77ZOA37vb+wHfAC+6x5wKHI9zT+d19ylxtz0EjAcmuzX9CvDv46URkW6kcEik5/m3O8e93Bjz729x3F3W2lpr7VJgKTA6YFsC8A6wHrgiYN74VcDt1tqvrWOptbYk8KTGGIMT2vzMWltqra0C7gO+v4c6GnGCnTRrbZ21dl/NF/3Andbaerf+EmvtbGttjXute3FuvkRERESCpQgnNPk+cKu1tspauwn4I3BpwH6RwD/dfc+21nb0Rzestc+59zxN1to/AtE4f+QDJ7S5173vygce6+AUj1lri6y1pcB/cYImgIuBp621i6219cCtwDHGmAE492jxwJGAsdaustZucUc+XQncaK0ttNY2W2s/c48XkRChcEik5znXWpvofpz7LY7bGvB1DRAX8HgSzl+V7rfW2oDnM3ECo71JBTxAbktohRM0pe5h/18BBvjCGJNnjLlyH+ffYa2ta3lgjPEYY/7iDmuuBOYDiUarZIiIiEjwpAMROOHPNwHPf+NuazEYOAfnj3YNezqZOy1slTuFqxxnJE+KuzkNyG93jfb2dN+XFri/tbYaZ3RQurX2A5zR2E8C240xs4wxCe51Y9j3PaGIBJHCIREB2IkT0LTo+y2Pfw9niPFcY0yfgOfzcYYs700xztSuEQGhlddaG9fRztbardbaH1tr04CfAE+Zva9QZts9/gXOX84mWmsTcIY/gxM4dbS/iIiISJcxxhyFEwD9m7YR0i2ygMKAx6twppm9bYwZSgfc/kK/whkhlGStTQQqaLvX2YLzB7zAa+yvosD6jDG9gOSWGq21j1lrxwPDcaaX3YRzr1fHvu8JRSSIFA6JCMAS4PtuQ+kJwPnf9gTW2j/gzGmfa4xp+cvU34B7jDFD3CaFo4wxye2O8wN/Bf5kjOkNYIxJN8ac1tF1jDHfa2kgDZThhDktc9a34czR35t4nDCq3Bjjw5l3H2h/ziEiIiJyUIwxCcaYs3B69jznTt1/GbjXGBNvjOmP07/xucDjrLX/BG4D5hhjOgpc4oEmYAcQYYz5DU4LgBYvA7e6i3RkANd/i7L/CVxhjBljjInGaQWw0Fq7yRhzlDFmojEmEucPj3WA373Xexp42BiTZpwFS45xjxeREKFwSEQA7sD5a04ZcBdOyPOtWWvvwfmr1xw3eHkY5wbkPaASZ5nWjlYhuxlYB3zuTvWaQ9u8+PaOAhYaY6qB13Hmr29wt/0W+Ic7Pe2CPRz/iFtDMfA5zhS2QI8C57sreHQ0B19ERETkYPzXGFOFM8L61zj3Sy1Np6/HCVY2AJ/g3JM93f4E1tp/AHcDH7j9fgK9i3N/swZnClgdu04ju8t9fiPOPdqz+1u4tXYOzn3jbJwRSINo6xOZgPMHvzL3/CXAg+62XwLLcRpllwIPoN9FRUKK2bU9iIhI13MbEzYD/a21m4Ndj4iIiIiISE+mtFZEgiEH569YW/e1o4iIiIiIiHQthUMi0q2MMdOBecDNe1tlQ0RERERERLqHppWJiIiIiIiIiPRgGjkkIiIiIiIiItKDRQS7gPZSUlLsgAEDgl2GiIiIdKHc3Nxia21qsOuQNroHExERObzt7f4r5MKhAQMGsGjRomCXISIiIl3IGPNNsGuQXekeTERE5PC2t/svTSsTEREREREREenBFA6JiIiIiIiIiPRgCodERERERERERHqwkOs5JCIi0t0aGxspKCigrq4u2KUcdmJiYsjIyCAyMjLYpcgB0Hvj4Ok9ICIihwKFQyIi0uMVFBQQHx/PgAEDMMYEu5zDhrWWkpISCgoKyM7ODnY5cgD03jg4eg+IiMihQtPKRESkx6urqyM5OVm//HYyYwzJyckadXII03vj4Og9ICIihwqFQyIiIqBffruIXtdDn/4bHhy9fiIicihQOCQiIiIiIiIi0oMpHBIRERERERER6cEUDomIiISIxx57jGHDhpGens51113XZdcpLy/nqaeeOqBjzzjjDMrLyzu5IpHud9999x3QcVdddRUrV67s5GpERESCS+GQiIhIiHjqqad4//33uffee7v0OnsLh5qamvZ67FtvvUViYmJXlCXSrfYUDllr8fv9ezzub3/7G8OHD++qskRERIJCS9mLiIgEuOu/eawsquzUcw5PS+DOs0fsdZ+rr76aDRs2MHXqVK688srW5zdt2sSVV15JcXExqampPPPMM2RlZfHKK69w1113ER4ejtfrZf78+eTl5XHFFVfQ0NCA3+9n9uzZDBkyZLdr3XLLLaxfv54xY8ZwyimncOaZZ3LHHXeQlJTE6tWrWbNmDeeeey75+fnU1dVx4403MmPGDAAGDBjAokWLqK6uZurUqRx77LF89tlnpKen85///IfY2NhOfe0kdATrvQF0+O/xnXfe4bbbbqO5uZmUlBTmzp1LdXU1119/PYsWLcIYw5133sn06dN3O98tt9xCbW0tY8aMYcSIEdx7772cdtppTJw4kdzcXN566y3uv/9+vvzyS2prazn//PO56667ADjxxBN56KGHmDBhAnFxcdx444288cYbxMbG8p///Ic+ffp06mskIiLSHTRySEREJATMnDmTtLQ05s2bR1JSUuvz119/PZdddhnLli3j4osv5oYbbgDg7rvv5t1332Xp0qW8/vrrree48cYbWbJkCYsWLSIjI6PDa91///0MGjSIJUuW8OCDDwKwePFiHn30UdasWQPA008/TW5uLosWLeKxxx6jpKRkt/OsXbuWa6+9lry8PBITE5k9e3anviYiLdr/e9y2bRs//vGPmT17NkuXLuWVV14B4J577sHr9bJ8+XKWLVvGSSed1OH57r//fmJjY1myZAnPP/884Px7vuaaa8jLy6N///7ce++9LFq0iGXLlvHRRx+xbNmy3c6zc+dOJk2axNKlSzn++OP561//2nUvgoiISBfSyCEREZEA+zOKoTstWLCAf/3rXwBceuml/OpXvwJgypQpXH755VxwwQVMmzYNgGOOOYZ7772XgoICpk2b1uGooT05+uijyc7Obn382GOP8dprrwGQn5/P2rVrSU5O3uWY7OxsxowZA8D48ePZtGnTAX+fEvqC+d5o/+9x1qxZHH/88a3/Zn0+HwBz5szhxRdfbD0uMGjdl/79+zNp0qTWxy+//DKzZs2iqamJLVu2sHLlSkaNGrXLMVFRUZx11lmA8x54//33D+wbFBERCTKNHBIRETkEzZw5k9/97nfk5+czfvx4SkpKuOiii3j99deJjY3ljDPO4IMPPtjv8/Xq1av16w8//JA5c+awYMECli5dytixY6mrq9vtmOjo6Navw8PD99mvSORAdPTvsSWU7EyB74GNGzfy0EMPMXfuXJYtW8aZZ57Z4XsgMjISYwyg94CIiBzaFA6JiIiEsMmTJ7eOhHj++ec57rjjAFi/fj0TJ07k7rvvJjU1lfz8fDZs2MDAgQO54YYbOOecczqcBgMQHx9PVVXVHq9ZUVFBUlISHo+H1atX8/nnn3f+Nyaynzr691hXV8f8+fPZuHEjAKWlpQCccsopPPnkk63HlpWV7fG8kZGRNDY2dritsrKSXr164fV62bZtG2+//XYnfkciIiKhR+GQiIhICHv88cd55plnGDVqFM8++yyPPvooADfddBMjR44kJyeHyZMnM3r0aF5++WVycnIYM2YMK1as4Ic//GGH50xOTmbKlCnk5ORw00037bb99NNPp6mpiWHDhnHLLbfsMtVGpLt19O8xNTWVWbNmMW3aNEaPHs2FF14IwO23305ZWRk5OTmMHj2aefPm7fG8M2bMYNSoUVx88cW7bRs9ejRjx47lyCOP5KKLLmLKlCld9v2JiIiEAmOtDXYNu5gwYYJdtGhRsMsQEZEeZNWqVQwbNizYZRy2Onp9jTG51toJQSpJOtDRPZjeG51Dr6OIiISCvd1/aeSQiIiIiIiIiEgPptXKREREDlMlJSWcfPLJuz0/d+7c3VYeEzlcTZw4kfr6+l2ee/bZZxk5cmSQKhIREQk9CodEREQOU8nJySxZsiTYZYgE1cKFC4NdgoiISMjTtDIRERERERERkR6sx4wcWry5jF++vJRHvz+WkRneYJcjIiIiIiIiIgehtqGZgrIaSnY24Pdbmq2l2W/xW0uzn4Cv2z7vst1a57jA7a3PQbPf756TDo4PPCcdHB94Tlqfa90e8LnZzy7PPXvVRNITY7v1tewx4ZAnKpwNxTvZVLJT4ZCIiIiIiIhIiGv2W7ZU1JJfWkt+aQ35ZTVsLq1xv65lR1X9vk9ykMIMhIcZwowhPCzgwxjC3M/hYYawMHZ/rnWbITzgPFERYbtuN4bwsLbtkWGmy7+v9npMOJSZ5AFgc2lNkCsRERERkWC77777uO222w7o2L///e+ceuqppKWldXJVIiI9i7WW8ppGJ/ApqyG/tJbNpTUUuCFQUXktjc22df8wA/28sWT6YjnpiGRGxJZypPkGn6mmqVdfmuP60tSrH9aTTHhYuBPYdBjktDzHbs/tGtqAMd0f1ARDjwmHekVHkNwrioIyhUMiIhKaHnvsMf785z9TWVnJeeedxxNPPNEl1ykvL+eFF17gmmuuOaDjH3nkEWbMmIHH4+nkykS6z8GGQzk5OQqHRET2Q12jM/WrJfjJL3VH/5Q5o4Gq65t22T/JE0mWz0NOupczRvYjM8nDgATI9m+i9841hG/Pg60rYE0eNO7s+KJhkRDfDxL6uZ/T2n3uB/FpEBnTDa/AoaHHhEMAGT4P+aW1wS5DRESkQ0899RRz5sxhzpw5LFq0qMuuU15ezlNPPXVQ4dAll1yicEi6zbnnnkt+fj51dXXceOONzJgxg3feeYfbbruN5uZmUlJSmDt3LtXV1Vx//fUsWrQIYwx33nkn06dP3+18t9xyC7W1tYwZM4YRI0bw/PPP89xzz/HYY4/R0NDAxIkTeeqppwD40Y9+1Hq+K6+8kszMTBYtWsTFF19MbGwsCxYsIDa2e/tCiIiEkma/ZVtlXdt0Lzf4aXm8vd3Ur+iIMDJ9HrJ8Ho4ekESmz+N8JHnITIohvmEHbF0O2xa4IdByKN0AuCOIor3QNwfGXuJ87jsSeqVC1VaoLIKqLbt+3rYC1r7fcZAUm+SERK0hUnpbcNTy2eODHjB6qEeFQ1k+D0vzy4NdhoiIhLK3b3FuSDpT35Ew9f697nL11VezYcMGpk6dypVXXtn6/KZNm7jyyispLi4mNTWVZ555hqysLF555RXuuusuwsPD8Xq9zJ8/n7y8PK644goaGhrw+/3Mnj2bIUOG7HatW265hfXr1zNmzBhOOeUUHnzwQR588EFefvll6uvrOe+887jrrrvYuXMnF1xwAQUFBTQ3N3PHHXewbds2ioqK+M53vkNKSgrz5s3r3NdKQleQ3hsATz/9ND6fj9raWo466ijOOeccfvzjHzN//nyys7MpLS0F4J577sHr9bJ8uVNnWVlZh+e7//77eeKJJ1iyZAkAq1at4qWXXuLTTz8lMjKSa665hueff54RI0ZQWFjIihUrACdYTUxM5IknnuChhx5iwoQJnfEqiIiENGstFbWNbSN/Avv+lNZQ2G7qlzGQ5o0lIymWE45IdcOfWLLcACg1PtqZqtXUADtWw7bPIX8FfLnMCXJqA/7fnTQA+uTAqAucnxl9ciAxq+Owxpuxt28C6iuhcgtUFbX77IZIW5dD9XZaQ6gW4dEQ33fPo49agqWI6IN6nYNtv8IhY8zpwKNAOPA3a22HP8WNMdOBV4GjrLWL3OduBX4ENAM3WGvf7YzCD0RmUixvL99Cs98SHoQGTyIiInsyc+ZM3nnnHebNm8cbb7zR+vz111/PZZddxmWXXcbTTz/NDTfcwL///W/uvvtu3n33XdLT0ykvL289x4033sjFF19MQ0MDzc3NHV7r/vvvZ8WKFa2/GL/33nusXbuWL774Amst3/3ud5k/fz47duwgLS2NN998E4CKigq8Xi8PP/ww8+bNIyUlpYtfFRHHY489xmuvvQZAfn4+s2bN4vjjjyc7OxsAn88HwJw5c3jxxRdbj0tKStqv88+dO5fc3FyOOuooAGpra+nduzdnn302GzZs4Prrr+fMM8/k1FNP7cxvS0QkZDhTv2rdvj8BU79Kneeq6nad+pXoTv0akebltJy+rcFPls9DWmIsURFhu15gZwlsWwYrVrijglbAjq/B3+hsj4iB3sNh2NnQd5QTAvUZATEJnfMNGgMxXuej95F73q+5Eaq3OWHRbqOQtsCWJfD129DUwYwkT7IbHKXtPvqoJUCKTQrZUUj7DIeMMeHAk8ApQAHwpTHmdWvtynb7xQM3AgsDnhsOfB8YAaQBc4wxR1hrO75b7WKZPg9NbrfzjCQNhRcRkQ7sxyiG7rRgwQL+9a9/AXDppZfyq1/9CoApU6Zw+eWXc8EFFzBt2jQAjjnmGO69914KCgqYNm1ah6OGOvLee+/x3nvvMXbsWACqq6tZu3Ytxx13HL/4xS+4+eabOeusszjuuOO64DuUQ0aQ3hsffvghc+bMYcGCBXg8Hk488UTGjBnD6tWrO+0a1louu+wyfv/73++2benSpbz77rvMnDmTl19+maeffrrTrisi0lUam/1U1jZSWdfkfm6koraRytqm1q+3VdS1jgLaVrn71K+MJGe0z4QBSc6UL3cEUKbPQ0JMZMcX9jdD6fq2AGjrcmdqWFVR2z5xfZ3pYENOcUKgviPBNwjCQ2BiU3ikMwJpX6OQ6so7Hn3U8rloMezcsfuxEbHtRiG1D5HSnO3he3h9u9D+vPpHA+ustRsAjDEvAucAK9vtdw/wAHBTwHPnAC9aa+uBjcaYde75Fhxs4Qciy9e2YpnCIREROZTNnDmThQsX8uabbzJ+/Hhyc3O56KKLmDhxIm+++SZnnHEGf/nLXzjppJP2eS5rLbfeeis/+clPdtu2ePFi3nrrLW6//XZOPvlkfvOb33TFtyOyRxUVFSQlJeHxeFi9ejWff/45dXV1zJ8/n40bN7ZOK/P5fJxyyik8+eSTPPLII4AzrWxPo4ciIyNpbGwkMjKSk08+mXPOOYef/exn9O7dm9LSUqqqqujVqxdRUVFMnz6doUOHcskllwAQHx9PVVVVt70GItLzNPstVXVOmFPhhjuVrZ+bAsKeXQOglv1rG/c+HiM8zNA7PppMn4djB6c6I39apn75PKTGRRO2r9k29VWwbSVsW94WAm1fCY3uIlAmHFKHQvZxbgiUA31GQlxqJ71KQWKMMwIoNgn6DN/zfk0NUL111xCpsrBtFFLhIli1BZrr2x1o4NqFzmvXjfYnHEoH8gMeFwATA3cwxowDMq21bxpjbmp37Oftjk1vfwFjzAxgBkBWVtb+VX4AWpazLyithUFddhkREZFOM3nyZF588UUuvfRSnn/++dbRO+vXr2fixIlMnDiRt99+m/z8fCoqKhg4cCA33HADmzdvZtmyZR2GQ+1/sT3ttNO44447uPjii4mLi6OwsJDIyEiamprw+XxccsklJCYm8re//W2X4zWtTLrD6aefzsyZMxk2bBhDhw5l0qRJpKamMmvWLKZNm4bf76d37968//773H777Vx77bXk5OQQHh7OnXfe2Tqyrr0ZM2YwatQoxo0bx/PPP8/vfvc7Tj31VPx+P5GRkTz55JPExsZyxRVX4Pf7AVpHFl1++eVcffXVakgtInvk91uq6pt2C3Qqa91Qp67jbS1hT/sVvNoLMxAfE4k3NpKE2AgSYiIZmBLX+nVC7K7bEmIj3c8ReGMjiY0M3/8l2q2FioKAkUDuqKDSDW37xHid6WDjLmtrEp165CHfh+egREQ5/ZES95JxWOv0WGo/hS2h+1fDPOhxW8aYMOBh4PIDPYe1dhYwC2DChAl2H7sfsH6JMYQZyNdy9iIicoh4/PHHueKKK3jwwQdbG1ID3HTTTaxduxZrLSeffDKjR4/mgQce4NlnnyUyMpK+ffvucZnu5ORkpkyZQk5ODlOnTuXBBx9k1apVHHPMMQDExcXx3HPPsW7dOm666SbCwsKIjIzkz3/+M+D8Un366aeTlpamhtTS5aKjo3n77bc73DZ16tRdHsfFxfGPf/xjv877wAMP8MADD7Q+vvDCC7nwwgt322/x4sW7PTd9+vQOV0ETkQDWQkU+FK8Bb6bTWPgQDQqstWyvqmfVlkpKqhv2EPbs+lxVfRN2H7/ZxscEBjkRZPk8u4Q4u4Q8MRHONvfrXlER+x7ZcyCa6p0m0S0jgVoCobqAhZ2Ssp3wZ/QP2qaFeTNCtpdOSDPGWQ3N43NCtWCWYvfxL9YYcwzwW2vtae7jWwGstb93H3uB9UC1e0hfoBT4Lk6fosB933XPtcdpZRMmTLBduXzvsQ98wPj+STz6/bFddg0RETm0rFq1imHDhgW7jMNWR6+vMSbXWqulnkJIR/dgem90Dr2O0iM0N8LWZZD/BWz+3Pkc2GfGhEFif0ge7HykDG77Oj4NwsL2fO5u5PdbNpfWkFdUyYqiCvKKKllZVEFxdcNu+8ZFR7SFNu0CnYTAQCdgm9fdFhcdEfxFknYWO//NAkOg4jXgd0ctRcQ6TaH75rgh0ChnGlV0fHDrlgO2t/uv/Rk59CUwxBiTDRTiNJi+qGWjtbYCaB1Xboz5EPiltXaRMaYWeMEY8zBOQ+ohwBcH+o10hswkD/mlGjkkIiIi0hNMnDiR+vpd+zk8++yzjBw5MkgViRwmassg/0vId4Ogwty2XjPeTOh/DGROgt7DnKkyJevcj7Xwzadt+wJEepyGxMmDIGVIW2iUPBhiE7vsW2hs9rNuezV5RZXkFVWQV1jJyi2VrVO6IsIMQ/rEc+LQ3uSkJTA8zUufhGgSYiKJj4kgIjw0Aq09qqt0Rm+Vb4byfCj/JuDxZqgpads3Ps0JgY443RkJ1Hck+AZCWHjw6pdutc9wyFrbZIy5DngXZyn7p621ecaYu4FF1trX93JsnjHmZZzm1U3AtcFaqaxFpi+WeV930DVcRETkMFNSUsLJJ5+82/Nz584lOTk5CBWJdL+FCxfueycR2Ttrnf4y+QvdUUELnalH4DQd7jsSxv0QMo92AiHvbm1mdz9f1RYoXusGRuud0GjrMlj1Xwj8ldGTsvtIo+Qh4Mv+VtPUahuaWbW1snUkUF5RJau3VtHQ5PQUi40MZ1i/eM4bm05OegIj0rwM6RNHdESIhiMtK2aVu2FPYOjT8hE4FQwgPNrtgZMJ/UY7r2XfkU6T6F66L+jp9qvnkLX2LeCtds91uFyJtfbEdo/vBe49wPo6XWaShx1V9dQ2NBMbFaJvdBER6XbW2v1vzHiISE5OZsmSJUGtYV/T1yX0HY7vje6k94AckprqoWiJMypo80InDKopdrZFe50QKOd8yJoI6eMhqte3O78xTsPdhDQYeEK7azdA2aZdRxqVrIe178NXzwWcI8wZodR+pFHyYCoie5O3paptRFBRJet3VON3346JnkhGpCVw+eQBjEhzgqDslF7Bn+YVyFqoKW032icgAKrIh/rKXY+J9LQ1QM44qi0ISuzvvFa9UkNm+p6EnoNuSH2oyUp2Vywrq2FIH82VFBERiImJoaSkhOTkZP0S3ImstZSUlBATExPsUuQA6b1xcPQekENG9Q4nAGr5KPoKmt0eO76BMOQUyJzofKQe2bUBQ0QUpB7hfLRXV+GOMnJGGtmSdTRtX4vZtICIpp2tu0XbSJJsX9JsPyKjMhmVNIjYo46kT/YIhmb3J80bE/z/p1kL1dvd4OebjkcANbZrhxIVD0n9ndBnwLFu8JPlBD+J/Z2mxo7Hh4UAACAASURBVMH+vuSQ1ePCoQx3Oft8hUMiIuLKyMigoKCAHTs07bizxcTEkJGREewy5ADpvXHw9B6QkOP3Q/HX7hSxhc7ooJYlycOjoN8YmPiTtjAorndw63X5/ZbN1RHklfQjr8jDiqL+rCw6ym0UbelNOVMSyzg6oYRhUTvI9BdyRO03hJfnQnEzFAPLgFhfwGijQc4UteTBTggW2YlBrt8P1VsDRvu0GwFUkQ9NdbseE5vkBD3Jg2HQSQHBjzsCKCZR4Y90mR4XDmX6YgHIL60NciUiIhIqIiMjyc7ODnYZIiFH7w2Rw0BDjdMsuqVxdP4Xbb1oPMlOADTuMsia5ARDnRmQHKDdGkUXVbKyaPdG0d8Z2tuZFpbuZVi/BOKiO/j1trkRyr5xp6e5U9WK18G6ubDk+YAdjRPABPY1Sh7kfO3N3H20VHOT0zdpl9E+ASOAKgvbRl+18KQ4QU+f4TD0dPBmtQU/3kyISejcF1LkW+hx4VBqXDQxkWFs1oplIiIiIiJyuKksaltKPv9zZ3nylqXJU4bC8O86TaMzJzrhR5BHonR5o+jwSKeZdcrg3bfVVwU0xF7X1iA7/wVoqA44R7TzWiVlO1PbKjZDReGujbMB4vo4YU/6OBh+Tlv/n8Qs8GZ8+95MIt2ox4VDxhgtZy8iIiIiIoc+fzNsywtYRewLJ7gAiIh1mkVPvsEZFZRxlNOTJogqahrJ2+IsGR8SjaKj4yFtrPMRyFqo3hYw0shtil26AWK8Trg2Miug548b/oTAqCuRA9XjwiGATJ+H/DJNKxMRERERkUNIXSUUfNnWOLpgUdsIl7i+zuphk37qfO47yhk1EyTF1fUsL6xgRUEFK9wgqCDgd7B+3hhGpCUwdWQ/RqQlkJPuDY1G0eCMporv63wMODbY1Yh0ix4ZDmX5PHyxsVRLs4qIiIiISGiy1ulhk/+FOypooTNKCOss4957BIz+flvj6MSsoE0RK2kJggorWFbgfC6qaGu2PDClF2MyE7l4Yn93RFACyXHRQalVRDrWI8OhjKRYquubKK9pJKlXVLDLERERERGRnsxaqCiA7SudAGjLEmclseqtzvaoeMiYACfc7IwKSp8QtObFpTsbAoKgclYUVlJY3jYiaGBKL47K9jEy3cvIdC8j0r0dN4oWkZDSI9+lmb625ewVDomIiIiISLepLW8LgbavhG0rYfsqqK9o2ycxC7KPaxsV1GcEhO1nA+ZOVF7jBEEto4GWFVTsEgRlp/RiXP8kLp88gJx0LyPSE0iICd5UNhE5cD0yHMpyw6HNpTWMykgMcjUiIiIiInLYaWqA4jVtQVBLGFRZ2LZPtNdZ1nzk+c7n3iOg9zCI7f7fUSpqGlleWOF+lLO8sIL80rYgqH+yh7FZifzwmP6MzPAyIs2LN1ZBkMjhokeGQ60jh0rVlFpERERERA6CtVC+effRQCVr25aQD4uElCOg/2ToPdwZCdR7uLPCVRD6BFXUNpJXWMGyljCooILNAas5Z/k8jMpwegSNTPeSk+bF61EQJHI465HhUFx0BEmeSPLLtJy9iIiIiIjsp9oydxpYuyCooaptH2+WMwpo6NS2ECh5MEQEp51FZV0jK9o1i95U0vZ7UKYvlpHpXn5wdJYTBKUnkOhR6w2RnqZHhkPgpOH5pQqHRERERESknaZ62PH17iFQVVHbPjFed8WwCwNGAw1zng+SqrpGVhRWsqJ1elgFG4t3tm5PT4xlVIaX703IZFSGMyJIPVhFBHpwOJTh85BXWLHvHUVERERE5PDk90PFZnc0UF7bqKDitWCbnX3CoyBlqNMgOnBKWEJa0JaOB6iubyIvIARaXlDBhnZBUE56AuePzyDHXTnMpyBIRPagx4ZDmUke3svbSrPfEh4WvP+pi4iIiIhIN6gpDRgF5H7evgoaqtv2ScxyRgMdeWZbEJQ8GMKD229nZ30TeUWVbgjkNIveULwTa53t/bwxjEz3ct7YdEZmOEFQclx0UGsWkUNLzw2HfLE0Nlu2VtaRnhgb7HJERERERKQzNNZB8de7jwaq2tK2T0yiE/yMuagtBEo9EmISgle3q66xeZf+QMsKK1i/o7o1COqbEENOupdzxqS7PYK8pMYrCBKRg9Njw6Gs1hXLahQOiYiIiIgcivx+2LoMNnwIRV85IVDJ+l2nhKUOhewT2paK7zMc4vsFdUpYoGa/Ja+ogk/WFfPpumIWbSqjvskPQO/4aEZleDlrVD+nR1C6l97xMUGuWEQORz02HMpMaguHJg1MDnI1IiIiIiKyT9ZC6QbY+JETCG2c76wgBpDY3xkBNPycttFAvkEQHlq/8lhr2Vi8k0/XFfPpuhI+W19MZZ2z5P2RfeO5ZFJ/Jg1MZlSGlz4JCoJEpHuE1v8pu1FaYizGoBXLRERERERCWfV2JwTa8CFs+MhpIA2QkA5HTIWBJ0L28ZDQL4hF7t32qjo+W1fCJ+uK+WxdMUUVdYDTNPr0nL5MGZzC5EEpmh4mIkHTY8OhqIgw0ryx5JfVBrsUERERERFpUV8N33zmjgz6CLatcJ6P9jorhk25wQmEkgeHzNSw9qrrm1i4oaR1qtiabU7T60RPJJMHJXPNoBSOHZxC/2QPJkS/BxHpWXpsOASQkRSrkUMiIiIiIsHU3AiFue7IoA+h4EvwN0F4NGRNhJN/44RB/cZAWHhwa92DhiY/X20u49P1JXy6rpgl+eU0+y3REWEcne1j2rgMjh2cwvB+CYRppWQRCUE9OhzK9Hn4eO2OYJchIiIiItJzWOs0jt7g9g365lN3OXkDaWNg8vVOA+msSRAZmgvH+P2W1VurnL5B64tZuKGU2sZmwgyMykjk6hMGMmVwCuOykoiJDM1AS0QkUI8Oh7J8HrZV1lPX2Kz/aYuIiIiIdJXy/LZpYhs+gp3bned9g2DUhc7IoAHHgscXxCL3Lr+0hk/XFfPJumIWrC+hZGcDAINSe3HBhAymDE5h4sBkvLGRQa5UROTb69HhUKbP+UtEQVktg3vHBbkaEREREZHDRE0pbPq4rYl06Xrn+V69YeAJbhPpEyAxM4hF7l3pzgYWrG/rG7TZbUfRJyGaE45IZcrgFKYMTqGvVyuKicihr2eHQy3L2ZfVKBwSERERETlQjbWw+fO2vkFblgIWouKcEUFHXeUEQr2HhWwT6dqGZr7YVOouMV9MXlElAPHREUwalMyVUwZw7JAUBqXGqYm0iBx2enQ4lOVzwyE1pRYRERER2X/+ZihaAhs/dMKgzQuhuR7CIiHjKDjxVmeEUPp4CA/NaVZNzX6WFVbw6VpnqthXm8tpaPYTFR7GuP6J/PLUI5gyOIWR6V4iwsOCXa6ISJfq0eFQanw00RFhCodERERERPbGWihZ1zYyaNPHUFfhbOszEo7+sTMyKOsYiA7NEfnWWtZtr3b7BpWwcEMJVfVNGAMj0hK4YsoApgxO4agBPmKj1I9URHqWHh0OGWPc5exrg12KiIiIiEhoqdratqLYhg+hqsh53psFw89xegZlnwBxqcGscq+2VNTy6boSPnMbSW+vqgegf7KHs8ekcezgFI4ZmExSr6ggVyoiElw9OhwCZ2rZZo0cEhEREZGerq4CNn3qrij2IexY7Twf64Ps452RQQNPgKTskO0bVFHbyOcbSlr7Bq3fsROA5F5RTB6cwrGDk5k8KIVMt72EiIg4enw4lOnzsOibsmCXISIiIiLSvZrqoeDLttFBhblgmyEiFvpPhjEXOYFQn5EQFro9dzaX1PDaV4V88PV2lheU47fgiQpnYraPHxydxZTBKQztE09YWGgGWiIioUDhUJKHqromKmoa8XpCs1meiIiI9EzGmNOBR4Fw4G/W2vvbbe8PPA2kAqXAJdbaAnfbH4AzgTDgfeBGa63txvIlVO0sgYV/hoWzoL4CTJjTOPq4nzvTxDKPhojoYFe5V9X1Tby1fAuv5hbwxcZSjIFxWUlcf9IQpgxOYUxmIlERoRtoiYiEGoVD7pDSzaU1jPR4g1yNiIiIiMMYEw48CZwCFABfGmNet9auDNjtIeD/rLX/MMacBPweuNQYMxmYAoxy9/sEOAH4sLvqlxBUtQ0WPA5fPg2NO2HY2TD6IhgwBWJC/z7Y77cs2FDC7NwC3l6xldrGZgam9OKm04Zy3th00hJjg12iiMghS+GQz/khkl9Ww8iM0P+hKCIiIj3G0cA6a+0GAGPMi8A5QGA4NBz4ufv1PODf7tcWiAGiAANEAtu6oWYJRRUF8OmjsPj/oLkBcs53Rgn1HhbsyvbLxuKdzM4t4LWvCiksryU+JoLzxqUzfVwG47ISMSHa/0hE5FCicMgdOaTl7EVERCTEpAP5AY8LgInt9lkKTMOZenYeEG+MSbbWLjDGzAO24IRDT1hrV7W/gDFmBjADICsrq/O/Awmu0g3wyZ9gyT8BC6N/AMf+DJIHBbuyfaqobeTNZVuYvbiA3G/KCDNw3JBUbpl6JKcM70NMpJaaFxHpTD0+HEqIicQbG6kVy0RERORQ9EvgCWPM5cB8oBBoNsYMBoYBGe5+7xtjjrPWfhx4sLV2FjALYMKECepHdLjY8TV8/DAsfwXCImD85TDlBkgM7QCw2W/5ZF0xr+YW8F7eVuqb/AzpHcctU4/kvLHp9EmICXaJIiKHrR4fDoGznH1+WW2wyxAREREJVAhkBjzOcJ9rZa0twhk5hDEmDphurS03xvwY+NxaW+1uexs4BtglHJLDzJZl8PEfYeV/IDIWJv0UJl8P8X2DXdlerdtexau5hbz2VQHbKuvxxkZy4VGZTB+XwagMr6aNiYh0A4VDOH2HVm+pCnYZIiIiIoG+BIYYY7JxQqHvAxcF7mCMSQFKrbV+4FaclcsANgM/Nsb8Hmda2QnAI91VuHSzgkUw/yFY8zZEJ8Bxv4BJ10Cv5GBXtkflNQ38d2kRr+YWsLSggvAww3eGpvLbszM4aVhvoiM0bUxEpDspHMJZzn7Oyu34/ZawMP1lQkRERILPWttkjLkOeBdnKfunrbV5xpi7gUXW2teBE4HfG2MszrSya93DXwVOApbjNKd+x1r73+7+HqSLbfoU5j8IG+ZBbBJ859dw9AyITQx2ZR1qavbz0ZodzF5cwJyV22lo9nNk33huP3MY54xJJzU+Otglioj0WAqHcJpSNzT72VZVRz+vlsAUERGR0GCtfQt4q91zvwn4+lWcIKj9cc3AT7q8QOl+1sL6D5yRQps/g16pcMrdMOFKiI4PdnUdWr21klcXFfDvJUUUV9fj6xXFxZOyOH98BiPStFqwiEgoUDhE4IpltQqHRERERCT0WAtfv+2MFCpaDAnpMPUPMO6HTn+hEFNSXc/r7rSxvKJKIsIMJw/rzfRxGZw4tDdREWHBLlFERAIoHAIyk5wfqPmlNRyd7QtyNSIiIiIiLn+z02D64z/CthWQ2B/OftRZlj4itKZhNTT5mff1dl7NLWDe6u00+S056Qn89uzhfHdMOr5eUcEuUURE9kDhEJCeFIsxaDl7EREREQkNzY2w/FUnFCpZCylHwHl/gZzzITx0buGtteQVVfJqbgGvLy2idGcDKXHRXDFlANPHZ3Bk34RglygiIvshdH6yBFF0RDh9E2LIL1M4JCIiIiJB1FQPS16AT/4E5d9Anxz43t9h2HchLHRW8NpeVcd/vipi9uICVm+tIio8jFOG92H6+HSOH5JKRLimjYmIHEoUDrkykzwUlNYGuwwRERER6YkaamDx/8Gnj0JVEaSPh6kPwBGngwmN1XTrm5qZu8qZNvbRmh00+y2jMxO559wczh7Vj0SPpo2JiByqFA65Mn0ePl1XHOwyRERERKQnqa+CL/8XFjwBO3dA/ylw7pMw8DshEQpZa1laUMFsd9pYRW0jfRKimXH8QKaPS2dw79BcIU1ERL4dhUOuTF8s26rqqG9qJjoidIbsioiIiMhhqLYMFs6Cz5+CunIYdBIc90sYMCXYlQGwtaKO174q5NXcfNbv2El0RBinjejL+eMzmDI4hfCw4AdXIiLSeRQOuTKTPFgLhWW1DEyNC3Y5IiIiInI42lkMC56EL/4KDVUw9AwnFMoYH+zKqGts5t28rcxeXMgna3fgtzChfxK/nzaQM0f1IyEmMtgliohIF1E45MpK9gDOimUKh0RERESkU1Vugc8eh9xnoLEWRpwLx/0C+o4MalnWWhZvLuPV3ALeWLqFqvom0rwxXPudwUwbl0F2Sq+g1iciIt1D4ZArM8kJh/LL1JRaRERERDpJ+Wb45BH46lnwN8OoC+DYn0PqEUEtq6nZz/9+spEXv8xnY/FOYiPDmZrjTBubNDCZME0bExHpURQOuXrHRxMVEUZBqZazFxEREZGDVLIePn4Ylr0IGBh7MUz5f+DLDnZl7Kiq57oXFrNwYylHZ/v46YmDOGNkP+Ki9auBiEhPpZ8ArrAwQ0ZiLJsVDomIiIjIgdq2Ej7+I+T9C8Kj4KirYPIN4E0PdmUALN5cxk+fy6W8ppE/fm8008dnBLskEREJAQqHAmT6POSXKRwSERERkW+p6CuY/xCsfgOi4mDy9XDMdRDXO9iVAU5voec+/4a731hJX28M/7pmMiPSvMEuS0REQoTCoQCZvliW5JcHuwwREREROVRsXgjzH4R170O0F064GSZeDR5fsCtrVdfYzG2vLedfiwv5ztBUHrlwLF6PVh4TEZE2CocCZCZ5qKhtpKK2EW+sfmCKiIiISAeshY3znVBo08fgSYaTf+NMIYsJrdE4m0tquPq5XFZtreT//c8QbjhpiJpNi4jIbhQOBcjyuSuWldbgTQ+tH+wiIiIiEgLWzoH5f4D8hRDXF067D8ZfDlGht+T7vK+38/9eXIK1lqcvO4rvHBkaU9xERCT0KBwKkOmGQwVlNeQoHBIRERGRQJ//Gd65BbyZcOYfYcwlEBkT7Kp24/dbHv9gHY/MXcORfROYeck4+ieHXnglIiKhQ+FQgMyklpFDtUGuRERERERCyjefwXu3w9Az4YJ/QHhotiCoqGnkZy8v4YPV2zlvbDr3nTeS2KjwYJclIiIhLmx/djLGnG6M+doYs84Yc0sH2682xiw3xiwxxnxijBnuPj/AGFPrPr/EGDOzs7+BzuT1RJIQE6Hl7EVERESkTdVWeOVySOwP5/05ZIOhVVsq+e6TnzB/zQ7uPmcED18wWsGQiIjsl32OHDLGhANPAqcABcCXxpjXrbUrA3Z7wVo7093/u8DDwOnutvXW2jGdW3bX0XL2IiIiItKqqQFevgzqq+DSf4dcw+kWr31VwK3/Wk5CTCQv/WQS4/uHzmppIiIS+vZnWtnRwDpr7QYAY8yLwDlAazhkra0M2L8XYDuzyO6UmeRh7faqYJchIiIiIqHg/Tsg/3OY/r/QZ3iwq9lNQ5Ofe99cyT8WfMPR2T6euGgsveNDrw+SiIiEtv0Jh9KB/IDHBcDE9jsZY64Ffg5EAScFbMo2xnwFVAK3W2s/7uDYGcAMgKysrP0uvitkJXv44Ovt+P1Wy3yKiIiI9GTLXoGFM2HSNTDy/GBXs5ttlXVc8/xicr8p46pjs7l56pFEhu9X1wgREZFddNpPD2vtk9baQcDNwO3u01uALGvtWJzg6AVjTEIHx86y1k6w1k5ITU3trJIOSGZSLA1NfnZU1we1DhEREREJoq0r4PXrIWsynHJ3sKvZzcINJZz52Ces2lLJ4z8Yy+1nDVcwJCIiB2x/foIUApkBjzPc5/bkReBcAGttvbW2xP06F1gPHHFgpXaPDF/LimXqOyQiIiLSI9WWw8uXOv2Fvvf3kGpAba3lfz/ZyEV/W0h8TAT/vnYKZ49OC3ZZIiJyiNufcOhLYIgxJtsYEwV8H3g9cAdjzJCAh2cCa93nU92G1hhjBgJDgA2dUXhXyXLDIa1YJiIiItID+f3w2tVQvtlZsj6+T7AralXT0MQNLy7hnjdWctKRvfnPdVM4ok98sMsSEZHDwD57Dllrm4wx1wHvAuHA09baPGPM3cAia+3rwHXGmP8BGoEy4DL38OOBu40xjYAfuNpaW9oV30hnSU+MBSC/tDbIlYiIiIhIt/vkj7DmbZj6B8iaFOxqWm0s3snVz+aydnsVN502lJ+eMEj9MUVEpNPsT0NqrLVvAW+1e+43AV/fuIfjZgOzD6bA7hYTGU6fhGgtZy8iIiLS06ybAx/cCyO/B0fPCHY1rd5fuY2fv7SEiHDDP648muOGBLdHp4iIHH72KxzqaTKTPJpWJiIiItKTlH0Ds6+C3sPh7EfBBH9UTrPf8qf31/DEvHWMTPfy50vGkZHkCXZZIiJyGFI41IEsn4fPN5QEuwwRERER6Q6NtU4Dar8fLnwWonoFuyLKdjZww4tf8fHaYi6ckMld54wgJjI82GWJiMhhSuFQBzJ8HrYsKaShyU9UhJYEFRERETlsWQtv/hK2LIUfvAjJg4JdEcsLKrj6uVx2VNXz+2kj+cHRWcEuSUREDnNKPjqQmRSLtVBYrqbUIiIiIoe1xf+AJc/B8TfB0KnBroaXv8xn+szPsNbyytXHKBgSEZFuoZFDHWhZzj6/tIbslOAPKxYRERGRLlCQC2/dBINOhhNvDWop9U3N/Pb1lfzzi81MGZzMY98fS3JcdFBrEhGRnkPhUAcyW8IhrVgmIiIicnjaWQwv/xDi+sL0v0FY8Pr5FJXX8tPncllaUMFPTxzEL08dSriWqRcRkW6kcKgDfRJiiAw35JdqWpmIiIjIYae5CV69EnbugB+9Bx5f0Er5bF0x1/3zKxqa/My8ZDyn5/QNWi0iItJzKRzqQHiYISPJQ76WsxcRERE5/Mz7HWz8CM55EtLGBKUEay1/mb+BP7yzmkGpccy8dDyDUuOCUouIiIjCoT3ISIrVtDIRERGRw82q/8Inf4Lxl8PYS4JSQlVdIze9sox38rZy5sh+/OH8UfSK1m25iIgEj34K7UGmz8OK5VuCXYaIiIiIdJbitfDaTyFtHEz9Q1BKWLe9ip88m8umkhpuP3MYPzo2G2PUX0hERIJL4dAeZPk8lNU0UlXXSHxMZLDLEREREZGDUV8NL10CEVFw4bMQ0f0rgb25bAu/enUpsVHhPPejiRwzKLnbaxAREemIwqE9yExqWc6+luFpCodEREREDlnWwuvXQfEauPQ18GZ06+Wbmv384d2vmTV/A2OzEnnq4nH088Z2aw0iIiJ7o3BoDzJ9zg/s/LIahqclBLkaERERETlgnz8Fea/B//wWBp7YrZcurq7nuhcW8/mGUi6d1J87zhpOVERYt9YgIiKyLwqH9iDL1zJySE2pRURERA5Zmz6B9+6AI8+CKf+vWy+9eHMZ1zy3mLKaBv74vdFMH9+9I5ZERET2l8KhPfDGRhIfHaFwSERERORQVVkEr1wOvmw498/QTY2frbU8v3Azd/03j77eGP51zWRGpHm75doiIiIHQuHQHhhjyPB5yC+rDXYpIiIiIvJtNTXAy5dBQw1c9gbEdE+bgLrGZn792gpmLy7gxKGpPHLhGBI9Ud1ybRERkQOlcGgvMpNi2VC8M9hliIiIiMi39d6voeALOP8Z6H1kt1wyv7SGq5/LJa+okhtPHsKNJw8hLEzL1IuISOhTOLQXWT4P89fuwFqL6aZhyCIiIiJykJa+BF/MgmOug5xp3XLJj9bs4IZ/foW1lqcvn8BJR/bpluuKiIh0BoVDe5Hp81DX6GdHdT2942OCXY6IiIiI7MvW5fDfG6H/sfA/d3X55fx+y5Pz1vHwnDUM7RPPXy4dT//kXl1+XRERkc6kcGgvWpezL61ROCQiIiIS6mrL4KVLIDYRvvcMhHftrW5FbSM/f2kJc1dv57yx6dx33khio8K79JoiIiJdQeHQXrQtZ1/L+P5BLkZERERE9szvh9euhopCuOItiOvdpZdbtaWSq5/LpbCslru+O4IfHtNfbQhEROSQpXBoLzKSWsIhLWcvIiIiEtI+fgjWvANnPASZR3fppf6zpJCbZy8jISaSl34yifH9fV16PRERka6mcGgvYiLDSY2PJr9M4ZCIiIhIyFr7Psy7D0Z9H466qssu4/db7n5jJX//bBNHD/DxxMVj1XpAREQOCwqH9iHL52GzRg6JiIiIhKayTTD7KugzAs76E3Th1K6P1xXz98828cNj+nPHWcOJDA/rsmuJiIh0J/1E24fMpFjyS2uDXYaIiIiItNdY6zSgxsKFz0KUp0svt2FHNQA3nDxEwZCIiBxW9FNtHzJ9HrZU1NLY7A92KSIiIiLSwlp44+fO0vXT/gq+gV1+yfzSWmIjw0nuFdXl1xIREelOCof2IdPnwW+hqFyjh0RERERCRu4zsPQFOOFmOOK0brnk5tIasnwerUomIiKHHYVD+5CZ1LacvYiIiIiEgIJF8NavYPApcMIt3XfZshoyfbHddj0REZHuonBoH1puALRimYiIiEgIqN4BL/+Q/8/enYc5VpZ5H//dqTXppTqpbqC7K9UgoNLIpk2zC6OAqMPWyCYoOiAugM4or+LyouIgvriMMyyOjAICKrQICtqKiDDKardsCgg0iElVN1BWUr0lteZ5/6ikKIpeUlVJnpPk+7muvkhOzulz81ef63ee5741e4G07CopVJnHWeecEqmM4rHy9jUCAMAHppVtw/y2sBpDxsQyAAAA30aGpZs/KGV6pTN/I0ViFbt1atOgMoMjY6vKAQCoJYRD29AQMi2MhpUkHAIAAPDrdxdJL/xBOu470vy9KnrrwovCTlYOAQBqENvKihCPRpRM03MIAADAmyd/Lt33n9KSM6W931vx24+FQ+2EQwCA2kM4VIR4LMLKIQAAAF96npF+9jFp4RLpqEu8lNCVf1HYEaUhNQCg9hAOFSEeCyu1aVCbBoZ9lwIAAFBfBjZIN50mNbZKJ10nNbZ4KSPRm9HcmS2KNNOVAQBQewiH/EMk8wAAIABJREFUijA2zp6JZQAAAJXjnPTzc6Te1dKJ10htC72VkmSMPQCghhEOFaEwsjTRSzgEAABQMQ9cPtpr6PAvSTu91WspiVSGZtQAgJpFOFSEwoMATakBAAAq5G9/kO78orTbMdKBH/daytBITmvX9TPGHgBQswiHihCNNGlGcwNNqQEAACphXbf0kw9I7TtLx10pmXktZ21fv0ZyjpVDAICaRThUBDNTPBZRFz2HAAAAymt4QPrJGdJwv3TyDVLLLN8VjY2xjxMOAQBqFOMWihSPRfT33k2+ywAAAKhtd3xO6lopnfgDad4bfFcj6ZWhJDSkBgDUKlYOFSkejSiZyso557sUAACA2vToj6WV35MOPE/a/Tjf1YxJpDJqDJnmtxEOAQBqE+FQkeKxsLJDI+rdNOi7FAAAgNqz9nHpF/8q7XiI9PYv+a7mVZKpjBZGw2oI+e19BABAuRAOFanQgDBBU2oAAIDSyqalm06XwjHpPddIDcHqfJBkjD0AoMYRDhWp0ICQiWUAAAAllMtJt5wtrV8jnXSdNHOe74peI5HKqIMx9gCAGhas1zIB1hEd3WPelc56rgQAAKCG/P5S6dnfSO/+phTf13c1r7Ghf0jpzBArhwAANY2VQ0WKNDdq7swWJXpZOQQAAFASz/xGuudr0l7vlZac6buazUqmRl8MEg4BAGoZ4dAkxGPhsVGmAAAAmIbU36RbzpJ2eJP0z9+SLJjNnhljDwCoB4RDkxCPRgiHAAAApmswI930PkkmnXS91BTc4KXQb5KVQwCAWkY4NAmdsYjW9PVreCTnuxQAAIDq5Jz0i3+TXvqLdML3pNhOvivaqmQqo1ktjWoLN/kuBQCAsiEcmoR4LKyRnNPadf2+SwEAAKhOq74vPX6jdNgF0q5H+K5mmxKpjOKxiCyg294AACgFwqFJiEcZZw8AADBlyT9Kv7pA2vVI6a2f9l1NUUbDoeBuewMAoBQIhyYhnt9rniAcAgAAmJyNL0vL3y+1LZSWXSWFgv8Ymss5daWz9BsCANS8Rt8FVJP5ba1qCBlNqQEAACZjZFi6+V+kbFo667dSOOq7oqL0bBzQwHCOcAgAUPMIhyahsSGkBXNalUxlfZcCAABQPe76kvTCH6TjvyvtsIfvaopWaCXQQTgEAKhxwV/PGzDxaIRtZQAAAMV64lbp/sukfT8k7XWK72omJcEYewBAnSAcmqTOWERdbCsDAAAVYGZHmdnTZrbazC7YzO+LzOwuM3vczO4xs45xv3Wa2W/M7Ckze9LMdqxk7WOGB6QdD5He8VUvt5+OwmrxhXNoSA0AqG2EQ5MUj0X0j42DygwO+y4FAADUMDNrkHSFpHdKWizpVDNbPOG0b0i6zjm3p6SLJF0y7rfrJH3dObebpKWSXi5/1Zux1ynSGbdLjc1ebj8diVRGO8xuVWtTg+9SAAAoq6LCoSLeWn3EzP5sZo+a2b3jH1zM7LP56542s3eUsngfOqKjb4660vQdAgAAZbVU0mrn3PPOuUFJN0o6dsI5iyX9Lv/57sLv+WexRufcnZLknNvonPO39NnM262nI8kYewBAndhmOFTkW6sfOef2cM7tLelSSd/KX7tY0imSdpd0lKQr839f1SrsOU/0srUMAACU1UJJyXHfu/LHxntM0rL85+MlzTKzdkmvl9RnZreY2SNm9vXNPYOZ2dlmtsrMVvX09JThf6G6JdMZxek3BACoA8WsHNrmWyvn3PpxX2dIcvnPx0q60Tk34Jz7m6TV+b+vahUeEBhnDwAAAuB8SYea2SOSDpXULWlEoxNpD8n/vq+k10n6wMSLnXNXOeeWOOeWzJs3r2JFV4OB4RG9uL6fZtQAgLpQTDhUzFsrmdk5ZvacRlcOfXyS11bNW6v2Gc0KNzUwzh4AAJRbt6T4uO8d+WNjnHNrnHPLnHP7SPp8/lifRp+5Hs2/3BuW9DNJb65M2bWhO52Vc6OTagEAqHUla0jtnLvCObezpM9I+sIkr62at1Zmps4Y4+wBAEDZrZS0q5ntZGbNGt2qf9v4E8xsrpkVnuc+K+nqcdfOMbPCg9XbJD1ZgZprxtgY+3bCIQBA7SsmHNrmW6sJbpR03BSvrQrxWJhx9gAAoKzyK37OlXSHpKckLXfOPWFmF5nZMfnTDpP0tJk9I2l7SRfnrx3R6Jayu8zsz5JM0v9U+H+hqiXzw0dYOQQAqAeNRZwz9tZKo8HOKZLeO/4EM9vVOfds/uu7JRU+3ybpR2b2LUkLJO0q6Y+lKNynjmhEDzzXK+ecrEqnbwAAgOBzzq2QtGLCsQvHfb5Z0s1buPZOSXuWtcAalkxl1NwY0nazWnyXAgBA2W0zHHLODZtZ4a1Vg6SrC2+tJK1yzt0m6VwzO1zSkKS0pDPy1z5hZss1uox5WNI5+TdZVa0zFtGmwRGlNg2qfSYPDAAAALUm0ZtRRzSsUIgXgQCA2lfMyqFi3lp9YivXXqz8Euda8crEsizhEAAAQA1KpjNMKgMA1I2SNaSuJ/FYWNLocmMAAADUnkSKcAgAUD8Ih6ag0JiQiWUAAAC1Z11mSBv6h2lGDQCoG4RDUzCjpVHtM5qZWAYAAFCDCi8A46wcAgDUCcKhKeqIRZRMZX2XAQAAgBJLpgvhUNhzJQAAVAbh0BTFo2G2lQEAANQgVg4BAOoN4dAUdcYiWtOX1UjO+S4FAAAAJZRIZRSNNGl2a5PvUgAAqAjCoSmKxyIazjmtXcfWMgAAgFqSTGVYNQQAqCuEQ1NUmF5B3yEAAIDaQjgEAKg3hENT1BkrhEP0HQIAAKgVIzmn7r4sY+wBAHWFcGiK5s9pVchemWYBAACA6vfi+n4NjbixF4EAANQDwqEpamoIaX5bmJVDAAAANSSZYow9AKD+EA5NQ2cswjh7AACAGlJ4tmPlEACgnhAOTUM8FlYyTUNqAACAWpFMZRQyacEcVg4BAOoH4dA0xKMR9WwYUHZwxHcpAAAAKIFkKqP5bWE1NfCYDACoH/yrNw2d7aPLjbtoSg0AAFATEqkMW8oAAHWHcGgaOvIjTplYBgAAUBuS6SzNqAEAdYdwaBoKDw7JFH2HAAAAql12cEQ9GwZYOQQAqDuEQ9Mwb2aLWptCTCwDAACoAYVWAXHCIQBAnSEcmgYzUzwaUZJwCAAAoOoVXvgRDgEA6g3h0DTFYxHG2QMAANSAQjjEtjIAQL0hHJqmeDSsZCoj55zvUgAAADANyVRW4aYGtc9o9l0KAAAVRTg0TfFYRBsHhtWXGfJdCgAAAKahMMbezHyXAgBARREOTVNhTzrj7AEAAKpbVzrDGHsAQF0iHJqmeDQfDjHOHgAAoGo555RIZWhGDQCoS4RD01R4u8Q4ewAAgOrVu2lQmcGRsRd/AADUE8KhaZrV2qRopIltZQAAAFUsyaQyAEAdIxwqgXgsMvZAAQAAgOozNsa+nXAIAFB/CIdKgHAIAACgunWlR/tHdkRpSA0AqD+EQyUQj0bU3ZfVSM75LgUAAABTkOjNaO7MFkWaG32XAgBAxREOlUA8FtbQiNNL6/t9lwIAAIApSDLGHgBQxwiHSqDQuJCJZQAAANUpkcrQjBoAULcIh0qgMPKUvkMAAADVZ2gkpzV9WcbYAwDqFuFQCSyYE5aZlMw3MgQAAED1WNvXr5xjjD0AoH4RDpVAc2NIC9rCrBwCAACoQoXWAHHCIQBAnSIcKpGOKOEQAABANUqmC+EQDakBAPWJcKhE4rHI2IMFAAAAqkcilVFjyDS/jXAIAFCfCIdKJB6N6KX1A+ofGvFdCgAAACYhmcpoYTSshpD5LgUAAC8Ih0qks330TVMXTakBAACqSpIx9gCAOkc4VCJj4+zZWgYAAFBVEqmMOhhjDwCoY4RDJVKYbkFTagAAgOqxoX9I6cwQK4cAAHWNcKhE5s1sUUtjiHAIAACgiiRToy0BCIcAAPWMcKhEQiHLj7On5xAAAEC1YIw9AACEQyXFOHsAAIDqUlj1zcohAEA9Ixwqoc5YRAm2lQEAAFSNZCqjWS2Nags3+S4FAABvCIdKKB6NaEP/sNZlhnyXAgAAgCIkUhnFYxGZme9SAADwhnCohAp71dlaBgAAUB0SqQxbygAAdY9wqIQK4+zZWgYAABB8uZxTVzpLM2oAQN0jHCqhQjjEOHsAAIDg69k4oIHhHCuHAAB1j3CohGa3Nqkt3MS2MgAAgCpQeKHXQTgEAKhzhEMlNjqxLOu7DAAAAGxDgjH2AABIIhwquXgsrC62lQEAAAReMv9Cb+Eceg4BAOob4VCJxaMRdaWzyuWc71IAAACwFYlURjvMblVrU4PvUgAA8IpwqMQ6YhENjuT00oZ+36UAAABgK5KMsQcAQBLhUMl1jk0so+8QAABAkCXTGXUwxh4AAMKhUotHRx8wGGcPAAAQXAPDI3pxfT8rhwAAEOFQyS2MhmX2yvQLAAAABE93OivnRvtFAgBQ74oKh8zsKDN72sxWm9kFm/n9k2b2pJk9bmZ3mdmicb+NmNmj+T+3lbL4IGppbNAOs1uVTBMOAQAABNXYGPt2wiEAABq3dYKZNUi6QtIRkrokrTSz25xzT4477RFJS5xzGTP7qKRLJZ2c/y3rnNu7xHUHWjwaURc9hwAAAAIrmR59VmPlEAAAxa0cWipptXPueefcoKQbJR07/gTn3N3OucJSmQcldZS2zOrSEQuzcggAACDAkqmMmhtD2m5Wi+9SAADwrphwaKGk5LjvXfljW3KmpF+N+95qZqvM7EEzO25zF5jZ2flzVvX09BRRUrB1xiJ6cX2/BoZHfJcCAACAzUj0ZhSPhhUKme9SAADwrqQNqc3sdElLJH193OFFzrklkt4r6dtmtvPE65xzVznnljjnlsybN6+UJXkRj0bk3GijQwAAAARPMp1RnEllAABIKi4c6pYUH/e9I3/sVczscEmfl3SMc26gcNw5153/7/OS7pG0zzTqrQqFB40k4RAAAEAgJVIZxtgDAJBXTDi0UtKuZraTmTVLOkXSq6aOmdk+kr6r0WDo5XHHo2bWkv88V9JBksY3sq5JhQcNxtkDAAAEz7rMkDb0D9OMGgCAvG1OK3PODZvZuZLukNQg6Wrn3BNmdpGkVc652zS6jWympJ+YmSQlnHPHSNpN0nfNLKfRIOprE6ac1aTtZrWouTGkLsIhAACAwCm8wGNbGQAAo7YZDkmSc26FpBUTjl047vPhW7jufkl7TKfAahQKmTrmMLEMAAAgiArPaPFY2HMlAAAEQ0kbUuMV8ViEbWUAAAABxMohAABejXCoTOKxsJIpGlIDAAAETSKVUTTSpNmtTb5LAQAgEAiHyiQejWhddkjrskO+SwEAAMA4yRRj7AEAGI9wqEzGxtmztQwAACBQCIcAAHg1wqEyKYyz76IpNQAAQGCM5Jy6+7KMsQcAYBzCoTIpPHDQdwgAACA4Xlzfr6ERN/YiDwAAEA6VTVukSbNaG5lYBgAAECCJXsbYAwAwEeFQGXXGIkqyrQwAACAwCs9mrBwCAOAVhENlFI9GaEgNAAAQIMlURiGTFsxh5RAAAAWEQ2UUj4XVlc4ql3O+SwEAAIBGw6H5bWE1NfAYDABAAf8qllFnLKKB4Zx6Ng74LgUAAACSEqkMW8oAAJiAcKiMOmKFiWVsLQMAAAiCZDpLM2oAACYgHCqjsXH2NKUGAADwLjs4op4NA6wcAgBgAsKhMuqIjr6VSvRmPVcCAACAwgu7OOEQAACvQjhURq1NDdp+dgsrhwAAwJSY2VFm9rSZrTazCzbz+yIzu8vMHjeze8ysY8Lvs82sy8wur1zVwVXY6k84BADAqxEOlRnj7AEAwFSYWYOkKyS9U9JiSaea2eIJp31D0nXOuT0lXSTpkgm/f0XS78tda7VI5J/J2FYGAMCrEQ6VWWeMcAgAAEzJUkmrnXPPO+cGJd0o6dgJ5yyW9Lv857vH/25mb5G0vaTfVKDWqpBMZRVualD7jGbfpQAAECiEQ2XWEYto7fp+DQ7nfJcCAACqy0JJyXHfu/LHxntM0rL85+MlzTKzdjMLSfqmpPO3dgMzO9vMVpnZqp6enhKVHVyFMfZm5rsUAAAChXCozOLRsJyT1vTRlBoAAJTc+ZIONbNHJB0qqVvSiKSPSVrhnOva2sXOuaucc0ucc0vmzZtX/mo960pnGGMPAMBmNPouoNYVGh4mUhntOHeG52oAAEAV6ZYUH/e9I39sjHNujfIrh8xspqQTnHN9ZnaApEPM7GOSZkpqNrONzrnXNLWuF845JVIZHbBzu+9SAAAIHMKhMis0PGRiGQAAmKSVknY1s500GgqdIum9408ws7mSUs65nKTPSrpakpxzp4075wOSltRzMCRJvZsGlRkcoRk1AACbwbayMtt+dquaGkzJFNvKAABA8Zxzw5LOlXSHpKckLXfOPWFmF5nZMfnTDpP0tJk9o9Hm0xd7KbYKjI2xjxIOAQAwESuHyqwhZFo4J8zEMgAAMGnOuRWSVkw4duG4zzdLunkbf8e1kq4tQ3lVZWyMfTvhEAAAE7FyqALisQjbygAAADzqSo+u4u6I0pAaAICJCIcqIB6LsHIIAADAo0RvRnNntijSzMJ5AAAmIhyqgHg0onRmSBv6h3yXAgAAUJeSjLEHAGCLCIcqYGxiGU2pAQAAvEikMkwqAwBgCwiHKqDwloq+QwAAAJU3NJLTmr4s4RAAAFtAOFQBhZGp9B0CAACovLV9/co5xtgDALAlhEMVMCfSpFktjYRDAAAAHhTG2MdZOQQAwGYRDlWAmakjFlEyTc8hAACASits7achNQAAm0c4VCHxaJiVQwAAAB4kUhk1hkzz2wiHAADYHMKhCumMRZRMZ+Sc810KAABAXUmmMloYDashZL5LAQAgkAiHKiQei6h/KKeejQO+SwEAAKgrScbYAwCwVYRDFTI2zj5F3yEAAIBKSqQyNKMGAGArCIcqpPC2ir5DAAAAlbOhf0jpzBBj7AEA2ArCoQrpiBIOAQAAVFph1TbbygAA2DLCoQppbWrQvFktY6NUAQAAUH6MsQcAYNsIhyooHg0rwcohAACAiims2mblEAAAW0Y4VEGdsQgNqQEAACoomcpoVkuj2sJNvksBACCwCIcqKB6LaO26rIZGcr5LAQAAqAuFSWVm5rsUAAACi3CoguLRiHJOWtPH6iEAAIBKSKQybCkDAGAbCIcqKD42zp5wCAAAoNxyOaeudJZm1AAAbAPhUAUVHkyYWAYAAFB+PRsHNDCcY+UQAADbQDhUQfPbwmoM2djUDAAAAJRP4Zmrg3AIAICtIhyqoIaQaSHj7AEAACoiwRh7AACKQjhUYfFoRMk0PYcAAADKrRAOLZxDzyEAALaGcKjC4rGwulg5BAAAUHbJVFY7zG5Va1OD71IAAAg0wqEKi8ci6t00qE0Dw75LAQAAqGlJxtgDAFAUwqEKi0fz4+yZWAYAAFBWyXRGHYyxBwBgmwiHKiyef3uVTNF3CAAAoFwGhkf04vp+Vg4BAFAEwqEKKzygMLEMAACgfLrTWTn3yqptAACwZYRDFRaNNGlGc4OShEMAAABlMzbGvp1wCACAbSEcqjAzUzwWURc9hwAAAMqm8CKOlUMAAGwb4ZAHHdEI28oAAADKKJnOqrkxpO1mtfguBQCAwCMc8qAzFlEylZVzzncpAAAANSnRm1E8GlYoZL5LAQAg8AiHPIjHwsoOjah306DvUgAAAGpSMp0ZmxILAAC2rqhwyMyOMrOnzWy1mV2wmd8/aWZPmtnjZnaXmS0a99sZZvZs/s8ZpSy+WhX2vrO1DAAAoDwSqQxj7AEAKNI2wyEza5B0haR3Slos6VQzWzzhtEckLXHO7SnpZkmX5q+NSfqipP0kLZX0RTOLlq786lSYmsHEMgAAgNJblxnShv5hmlEDAFCkYlYOLZW02jn3vHNuUNKNko4df4Jz7m7nXCHpeFBSR/7zOyTd6ZxLOefSku6UdFRpSq9eHdGwJKkrnfVcCQAAQO0prM5mWxkAAMUpJhxaKCk57ntX/tiWnCnpV5O51szONrNVZraqp6eniJKqW6S5UXNnNivRy8ohAACAUiuEQ2wrAwCgOCVtSG1mp0taIunrk7nOOXeVc26Jc27JvHnzSllSYMVjESXThEMAAAClVnjGisfCnisBAKA6FBMOdUuKj/vekT/2KmZ2uKTPSzrGOTcwmWvrUTxKOAQAAFAOiVRG0UiTZrU2+S4FAICqUEw4tFLSrma2k5k1SzpF0m3jTzCzfSR9V6PB0MvjfrpD0pFmFs03oj4yf6zuxWNhrenr1/BIzncpAAAANSWZYow9AACTsc1wyDk3LOlcjYY6T0la7px7wswuMrNj8qd9XdJMST8xs0fN7Lb8tSlJX9FowLRS0kX5Y3WvMxbRSM5p7bp+36UAAADUFMIhAAAmp7GYk5xzKyStmHDswnGfD9/KtVdLunqqBdaqwmhVHl4AAABKZyTn1N2X1VFvmu+7FAAAqkZJG1KjeIVAiL5DAAAApfPi+n4NjTgmlQEAMAmEQ57Mb2tVQ8jGRq0CAABg+hK9jLEHAGCyCIc8aWwIacGcViVTWd+lAAAA1AzG2AMAMHmEQx4xzh4AAKC0kqmMQiYtmEM4BABAsQiHPIpHI0qyrQwAAKBkkqmM5reF1dTAYy4AAMXiX02POtsj+sfGQWUGh32XAgAAUBMSqQz9hgAAmCTCIY86oqPLnbvS9B0CAAAohWQ6S78hAAAmiXDIo8I4+8JUDQAAAExddnBEPRsGWDkEAMAkEQ55VHhwoSk1AADA9L0yqYxwCACAySAc8qh9RrPCTQ2MswcAACiBwqAPwiEAACaHcMgjM1M8FlaCiWUAAADTVnimYlsZAACTQzjkWWcsoi62lQEAAExbMpVVuKlB7TOafZcCAEBVIRzyrCMaUTKVkXPOdykAAABVrTDG3sx8lwIAQFUhHPIsHoto0+CI0pkh36UAAABUta50hjH2AABMAeGQZ4U98fQdAgAAmDrnnBKpDM2oAQCYAsIhzwpvt5KEQwAAAFPWu2lQmcERmlEDADAFhEOexaOjDzBJmlIDAABM2dgY+yjhEAAAk0U45NmMlka1z2hm5RAAAMA0jI2xbyccAgBgsgiHAqAjFlEylfVdBgAAQNXqSo8+S3VEaUgNAMBkEQ4FQDwaZlsZAADANCR6M5o7s0WR5kbfpQAAUHUIhwIgHouoO53VSM75LgUAAKAqJRljDwDAlBEOBUBnLKLhnNPadWwtAwAAmIpEKsOkMgAApohwKADGJpbRdwgAAGDShkZyWtOXJRwCAGCKCIcCoLAEmollAAAAk7e2r185xxh7AACminAoABbMCStkoik1AADAFBTG2MdZOQQAwJQQDgVAU0NI89vCrBwCAACYgsILNhpSAwAwNYRDARGPhcfeegEAAKB4iVRGjSHT/DbCIQAApoJwKCA6YxEl0zSkBgAAmKxEKqOF0bAaQua7FAAAqhLhUEDEoxH1bBhQ/9CI71IAAACqShdj7AEAmBbCoYAoNFDsoik1AADApCRSGZpRAwAwDYRDAVF4oKHvEAAAQPE29A8pnRlijD0AANNAOBQQhekayRR9hwAAAIpVeHZiWxkAAFNHOBQQ82a2qLUpxDh7AACASWCMPQAA00c4FBBmpng0wrYyAACASSi8WGPlEAAAU0c4FCBxxtkDAABMSiKV0azWRrWFm3yXAgBA1SIcCpB4NKyuVEbOOd+lAAAAVIVkKqN4NCIz810KAABVi3AoQOKxiDYMDKsvM+S7FAAAgKqQSGXYUgYAwDQRDgVIYZx9obEiAAAAtiyXc+pKZ2lGDQDANBEOBUg8mg+HGGcPAAAkmdlRZva0ma02sws28/siM7vLzB43s3vMrCN/fG8ze8DMnsj/dnLlqy+/no0DGhjOsXIIAIBpIhwKkMJbLyaWAQAAM2uQdIWkd0paLOlUM1s84bRvSLrOObenpIskXZI/npH0fufc7pKOkvRtM5tTmcorpzCprINwCACAaSEcCpBZrU2KRprYVgYAACRpqaTVzrnnnXODkm6UdOyEcxZL+l3+892F351zzzjnns1/XiPpZUnzKlJ1BSUYYw8AQEkQDgVMPBYZewsGAADq2kJJyXHfu/LHxntM0rL85+MlzTKz9vEnmNlSSc2Snpt4AzM728xWmdmqnp6ekhVeKYlURmbSwjn0HAIAYDoIhwImHiUcAgAARTtf0qFm9oikQyV1Sxop/Ghm8yVdL+mDzrncxIudc1c555Y455bMm1d9C4uSqay2n9Wq1qYG36UAAFDVCIcCJh6LqLsvq5Gc810KAADwq1tSfNz3jvyxMc65Nc65Zc65fSR9Pn+sT5LMbLakX0r6vHPuwcqUXFlJxtgDAFAShEMBE4+FNTTi9NL6ft+lAAAAv1ZK2tXMdjKzZkmnSLpt/AlmNtfMCs9zn5V0df54s6RbNdqs+uYK1lxRyXRGHYyxBwBg2giHAqYwzp6JZQAA1Dfn3LCkcyXdIekpScudc0+Y2UVmdkz+tMMkPW1mz0jaXtLF+eMnSXqrpA+Y2aP5P3tX9v+gvAaGR/Ti+n5WDgEAUAKNvgvAqxUecJKpjPZ/Xfs2zgYAALXMObdC0ooJxy4c9/lmSa9ZGeScu0HSDWUv0KPudFbOvfJiDQAATB0rhwJmwZywzKRkOuu7FAAAgMAaG2PfTjgEAMB0EQ4FTHNjSPNnt6qLbWUAAABbVJjuyrYyAACmj3AogOKxCD2HAAAAtiKZzqq5MaR5M1t8lwIAQNUjHAqgeCyiZJpwCAAAYEsSvRnFo2GFQua7FAAAqh7hUADFoxG9tH5A/UMjvksBAAAIpGQ6ozhbygAAKAnCoQDqbA9LkrpoSg0AALBZiVSGfkMAAJQI4VAAFUaysrUMAADgtdZlhrShf5gx9gAAlAjhUAAVlkgzsQyYwKrTAAAgAElEQVQAAOC1CoM72FYGAEBpEA4F0LyZLWpuDDGxDAAAYDMSjLEHAKCkCIcCKBQyxaNhJVP0HAIAAJiosPU+Hgt7rgQAgNpQVDhkZkeZ2dNmttrMLtjM7281s4fNbNjM3jPhtxEzezT/57ZSFV7rGGcPAACweYlURtFIk2a1NvkuBQCAmtC4rRPMrEHSFZKOkNQlaaWZ3eace3LcaQlJH5B0/mb+iqxzbu8S1FpX4tGI/vT3tO8yAAAAAieZYow9AAClVMzKoaWSVjvnnnfODUq6UdKx409wzr3gnHtcUq4MNdalzlhEG/qHtS4z5LsUAACAQCEcAgCgtIoJhxZKSo773pU/VqxWM1tlZg+a2XGbO8HMzs6fs6qnp2cSf3XtKuyhZ2sZAADAK0ZyTt19WcbYAwBQQpVoSL3IObdE0nslfdvMdp54gnPuKufcEufcknnz5lWgpODryD/wMLEMAADgFS+u79fQiGNSGQAAJVRMONQtKT7ue0f+WFGcc935/z4v6R5J+0yivrrV2T76wJMkHAIAABiT6GWMPQAApVZMOLRS0q5mtpOZNUs6RVJRU8fMLGpmLfnPcyUdJOnJrV8FSZrd2qS2cBPbygAAAMZhjD0AAKW3zXDIOTcs6VxJd0h6StJy59wTZnaRmR0jSWa2r5l1STpR0nfN7In85btJWmVmj0m6W9LXJkw5w1bEY2ElU1nfZQAAAARGMpVRyKQFcwiHAAAolW2Ospck59wKSSsmHLtw3OeVGt1uNvG6+yXtMc0a61ZnLKK/rt3guwwAAIDASKYymt8WVlNDJVpnAgBQH/hXNcDi0Yi60lnlcs53KQAAAIGQSGXoNwQAQIkRDgVYRyyiwZGcXt4w4LsUAACAQEiksvQbAgCgxAiHAqzwVoxx9gAAAFJ2cET/2DjAyiEAAEqMcCjA4tHRt2KMswcAABg/qYxwCACAUiIcCrCF0bDMxDh7AAAAvfLCjHAIAIDSIhwKsJbGBm0/q5VtZQAAAHplqz3bygAAKC3CoYDrjEXUlcr6LgMAAMC7ZCqrcFOD2mc0+y4FAICaQjgUcB2xMNvKAAAA9MoYezPzXQoAADWFcCjg4tGIXlzfr4HhEd+lAAAAeJVMZeg3BABAGRAOBVxnLCLnpO40W8sAAED9cs4pmc4oHgv7LgUAgJpDOBRwhbdjScIhAABQx3o3DSozOEIzagAAyoBwKOAKb8eYWAYAAOrZ2Bj7KOEQAAClRjgUcNvPalVzQ0hdhEMAAKCOjY2xbyccAgCg1AiHAi4UMnVEmVgGAADqW1d+i31HlJ5DAACUGuFQFeiIRZRM0XMIAADUr0RvRnNntijS3Oi7FAAAag7hUBXojIXpOQQAAOpaIpVRJ5PKAAAoC8KhKhCPRrQuO6T1/UO+SwEAAPBidIw9/YYAACgHwqEqMDbOntVDAACgDg2N5LSmL8sYewAAyoRwqAp0Eg4BAIA6travXznHGHsAAMqFcKgKFB6EaEoNAADqUaH3ItvKAAAoj/oJhzIp6ZefkgY2+K5k0toiTZrV2sg4ewAAUJcKz0BxGlIDAFAW9RMO9fxVWnWN9PNzJOd8VzNp8WiEiWUAAKAuJVIZNYZM89sIhwAAKIf6CYcWHSgd/kXpyZ9LD1zhu5pJ64xF6DkEAADqUiKVUUc0rIaQ+S4FAICaVD/hkCQd+HFpt6OlOy+UXrjPdzWTEo+F1ZXOKpervlVPAAAA09GVYow9AADlVF/hkJl07JVSbCfpJx+Q1q/1XVHR4rGIBoZz6tk44LsUAACAikoQDgEAUFb1FQ5JUuts6eQbpMGNowHRyJDviooSZ5w9AACoQxv6h5TODDHGHgCAMqq/cEiStttNOuYyKfmg9Jv/67uaooyNs2diGQAAqCPJVFbSaP9FAABQHvUZDknSHu+R9vuo9NB3pD/f7LuabeqIjk7nSPRmPVcCAABQOYyxBwCg/Oo3HJKkI78ixfeXbjtPevkp39VsVWtTg7af3cLKIQAAUFcKW+pZOQQAQPnUdzjU0CSdeK3UPFO66XSpf73virYqHmWcPQAAqC+JVEazWhvVFm7yXQoAADWrvsMhSZo9fzQgSv1N+vnHJBfcUfHxGOEQAACoL8lURvFoRGbmuxQAAGoW4ZAk7XiQdMSXpadul+7/L9/VbFE8FtHa9f0aHM75LgUAAKAiEqkMW8oAACgzwqGCA86VFh8r/fZL0t9+77uazYpHw3JOWtNHU2oAAFD7cjmnrnSWZtQAAJQZ4VCBmXTsFVL7LtLN/yKtX+O7oteIxxhnDwAA6kfPxgENDOdYOQQAQJkRDo3XMks6+QZpKCstP0MaHvRd0asUHowS9B0CAAB1oNBrsYNwCACAsiIcmmjeG6RjL5e6/ij95gu+q3mV7We3qqnBlEyxrQwAANS+BGPsAQCoCMKhzdn9eGn/c6Q/fld6fLnvasY0hEwL54TZVgYAAOpCIpWRmbRwDj2HAAAoJ8KhLTniy1LngdLtn5BeesJ3NWMYZw8AAOpFMpXV9rNa1drU4LsUAABqGuHQljQ0SSdeM9qH6KbTpf51viuSRDgEAADqR5Ix9gAAVATh0NbM2kE68QdSX0L62cck53xXpHg0onRmSBv6h3yXAgAAUFbJdEYdjLEHAKDsCIe2ZdEB0hFfkf76C+m+b/uuRvH8AxJNqQEAQC0bGB7Ri+v7WTkEAEAFEA4VY/+PSrsvk+66SHr+Hq+lFB6QaEoNAABqWXc6K+dGV00DAIDyIhwqhpl0zGVS+67SzWdK67q8lVJ4QKLvEAAAqGVjY+zbCYcAACg3wqFitcyUTr5BGu6Xlp8hDQ94KWNOpEkzWxoJhwAAQE0rPOuwrQwAgPIjHJqMea+XjrtS6l4l3fE5LyWY2ejEsjQ9hwAAQO1KprNqbgxp3swW36UAAFDzCIcma/Gx0oHnSSu/Jz12o5cS4tEwK4cAAEBNS/RmFI+GFQqZ71IAAKh5hENT8fYvSYsOlm7/V+nFv1T89qMrhzJyzlX83gAAAJWQTGcUZ0sZAAAVQTg0FQ2N0onXSOE50k2nS9m+it6+MxZR/1BOPRv99D0CAAAoJ+ecEr0Z+g0BAFAhhENTNXM76cQfSOuS0q0fkXK5it06HgtLkpIp+g4BAIDasy47pA0Dw4RDAABUCOHQdHTuJx15sfTMr6R7v1Wx2xbG2Xel6TsEAABqT+EFWEeUcAgAgEogHJqu/T4svek90t0XS8/9riK37IhG1BAyLV+V1Pr+oYrcEwAAoFISjLEHAKCiCIemy0w65r+kuW+Qbj5T6kuW/Zbh5gZ9+Zjd9dDzKR13+X169qUNZb8nAABApSTzq6MLW+kBAEB5EQ6VQvMM6eQbpJEhafn7peHyN4o+ff9F+uFZ+2l9/5COu+I+/fova8t+TwAAgEpIpDKKRpo0q7XJdykAANQFwqFSmbuLdPx3pDUPS7++oCK33O917br9vIO1y/az9JEbHtalv/6rRnKMtwcAANUtmWKMPQAAlUQ4VEq7HS0d9Alp1dXSoz+qyC3nt4W1/MP76+QlcV15z3P6l2tXal2GPkQAAKB6EQ4BAFBZhEOl9rYLpR0PkX7xb9Laxytyy5bGBn3thD108fFv0v3P/UNHX36vnlq7viL3BgAAKKWRnFN3X5Zm1AAAVBDhUKk1NErvuUYKx6SbTpey6Yrc1sx02n6LdOPZB6h/aETLrrxftz+2piL3BgAAKJUX1/draMQpzhh7AAAqpqhwyMyOMrOnzWy1mb2moY6ZvdXMHjazYTN7z4TfzjCzZ/N/zihV4YE2c5500g+k9WukWz4s5XIVu/VbFkX1i/MO1uIFs3Xejx/RV1c8peGRyt0fAABgOhK9jLEHAKDSthkOmVmDpCskvVPSYkmnmtniCaclJH1A0o8mXBuT9EVJ+0laKumLZhadftlVIL5UOuoS6dk7pD98s6K33m52q378of31vv0X6arfP68zrvmjUpsGK1oDAADAVDDGHgCAyitm5dBSSaudc8875wYl3Sjp2PEnOOdecM49LmniEpV3SLrTOZdyzqUl3SnpqBLUXR32PUva4yTp7oul1b+t6K2bG0P6ynFv0qXv2VMrX0jr6Mvu1V+611W0BgAAgMlKpjIKmbRgDuEQAACVUkw4tFBSctz3rvyxYhR1rZmdbWarzGxVT09PkX91FTCTjv62tN1i6adnSX2Jipdw0pK4fvLhA5RzTid8537d8nBXxWsAAAAoVjKV0fy2sJoaaI0JAEClBOJfXefcVc65Jc65JfPmzfNdTmk1z5BOvl7KjUjL3y8N9Ve8hL3ic3T7eQdr7/gcfXL5Y/rSbU9oiD5EAAAggBKpDP2GAACosGLCoW5J8XHfO/LHijGda2tH+87S8f8trXlE+tWnvZQwd2aLbjhrP33woB117f0v6LTvPaSeDQNeagEAAMUpYijIIjO7y8weN7N7zKxj3G9VORQkkWKMPQAAlVZMOLRS0q5mtpOZNUs6RdJtRf79d0g60syi+UbUR+aP1Z83vls6+JPSwz+QHr7eSwlNDSF98ejd9R8n76XHkn06+rJ79Wiyz0stAABg64ocCvINSdc55/aUdJGkS/LXVuVQkOzgiP6xcYBm1AAAVNg2wyHn3LCkczUa6jwlablz7gkzu8jMjpEkM9vXzLoknSjpu2b2RP7alKSvaDRgWinpovyx+vS2L0g7HSr98lPSmke9lXH8Ph366UcPVEPIdNJ/P6CbVla+FxIAANimbQ4F0Who9Lv857vH/V6VQ0FemVTGyiEAACqpqJ5DzrkVzrnXO+d2ds5dnD92oXPutvznlc65DufcDOdcu3Nu93HXXu2c2yX/55ry/G9UiVCD9J6rpRlzpeXvkzL+crI3LWzT7ecdrKU7xfSZn/5Zn7/1zxocpg8RAAABUsxgj8ckLct/Pl7SLDNrL/LawA0FSaYIhwAA8CEQDanryoy50knXSevXSrecLeX8BTKxGc269oP76sOHvk4/fCihU//nQb20vvINswEAwJSdL+lQM3tE0qEa7e04UuzFQRsKksiHQ/QcAgCgsgiHfOhYIr3za9LqO6XfX+q1lMaGkD77zt10+Xv30ZNr1uufL7tXf/p7/e78AwAgQLY52MM5t8Y5t8w5t4+kz+eP9RVzbRAlU1mFmxrUPqPZdykAANQVwiFflpwp7XWqdM/XpGfv9F2N/nnPBbr1nAMVaW7QKVc9qBse/Lucc77LAgCgnm1zKIiZzTWzwvPcZyVdnf9clUNBCmPszcx3KQAA1BXCIV/MpHd/S9p+d+mnZ0npF3xXpDfuMFu3nXOwDtplrr7ws7/oMz99XP1DRa9MBwAAJVTMUBBJh0l62syekbS9pEJvyKocCpJMZeg3BACAB4RDPjVHpJOvl5yTlr9fGvLf76ct0qTvn7Gvzv2nXbR8VZdO/u4DWtOX9V0WAAB1qYihIDc753bNn3OWc25g3LVVNRTEOadkOsMYewAAPCAc8i32OmnZVdLax6QV5/uuRpLUEDKd/4436L9Pf4tWv7xRR192rx58vtd3WQAAoIb1bhpUZnCEZtQAAHhAOBQEbzhKOuR86ZHrpT/9wHc1Y4560w76+bkHqS3SpNO+95Cuue9v9CECAABlMTbGPko4BABApREOBcU/fU563T9JK/6PtOYR39WM2WW7WfrZOQfpn96wnb58+5P61PLH6EMEAABKbmyMfTvhEAAAlUY4FBShBumE70szt5Nuer+UCU7PyNmtTbrqfW/RJ494vW59tFsnfOf+sbd7AAAApdCVHu1x2BGl5xAAAJVGOBQkM9qlk34gbXxxdIJZLjgrdEIh08ffvqu+f8YSJVIZHXP5vbpv9T98lwUAAGpEojejuTNbFGlu9F0KAAB1h3AoaBa+RXrnpdJzd0n/+/98V/Mab3vj9rrt3IM1d2aL3vf9h3TV75+jDxEAAJi2RCqjTiaVAQDgBeFQEL3lA9Lep42GQ8/c4bua19hp7gzdes5BesfuO+irK/6q8378iDKDw77LAgAAVWx0jD39hgAA8IFwKIjMpHd/U9phD+mWD0mpv/mu6DVmtjTqytPerE8f9Qb98s9rtezK+/X33k2+ywIAAFVoaCSnNX1ZxtgDAOAJ4VBQNYWlk64f/bz8fdJQ1m89m2Fm+thhu+jaDy7V2nX9Ovqye3XP0y/7LgsAAFSZtX39yjnG2AMA4AvhUJDFdpKW/Y/04p+lX35KCmhvn0NfP0+3n3uwFswJ64PXrtQVd6+mDxEAAChaYYw928oAAPCDcCjoXv8O6dDPSI/+UPrTtb6r2aLO9ohu+diBOnrPBfr6HU/rozc8rI0D9CECAADblkwXwiEaUgMA4APhUDU49DPSzm+XfvVpqftPvqvZokhzo/7zlL31hXfvpjufeknHXXGfnuvZ6LssAAAQcIlURo0h0/w2wiEAAHwgHKoGoQbphO9JM3eQbnq/tKnXd0VbZGY665DX6fp/WarUpkEdd/l9+u2TL/kuCwAABFgilVFHNKyGkPkuBQCAukQ4VC0iMenk66RNPdJPz5RyI74r2qoDd5mr2849SIvmRnTWdav0H3c+o1yOPkQAAOC1ulKMsQcAwCfCoWqyYB/pXV+Xnr9buvurvqvZpo5oRDd/5EAte/NC/eddz+pD163S+v4h32UBAICASRAOAQDgFeFQtXnLGdI+75P+8A3pvv+Shgd8V7RVrU0N+uaJe+nLx+yu/32mR8defp+efWmD77IAAEBAbOgfUjozxBh7AAA8IhyqRu/6hrTLEdKd/1e6fIn02I2B3mZmZjrjwB31ow/trw39Qzruivv067+s9V0WAAAIgGQqK0nqZOUQAADeEA5Vo6ZW6bSfSKffIrXOkW79sPTfh0jP3CG54Pb1WbpTTL847xDtuv0sfeSGh3Xpr/+qEfoQAQBQ1xKp0TH2hEMAAPhDOFStzKRd3i6d/b/SCd+XhjLSj06SrnmXlHjId3VbtENbq2768P46dWlcV97znD547Ur1ZQZ9lwUAADzpSo+GQ/EYY+wBAPCFcKjahULSHu+Rzl0pvfubUu9q6eojpR+fKr38lO/qNqulsUGXLNtTXz1+Dz3w3D90zOX36U9/T8kFeNUTAAAoj0Qqo1mtjWoLN/kuBQCAukU4VCsamqR9z5I+8aj0ti9IL9wrXXmAdOtHpb6E7+o26737derGsw9Q/9CITvjOA3rHt3+v7/7vc3ppfb/v0gAAQIUkUxnFoxGZme9SAACoW4RDtaZ5hvTW/yN94jHpgHOkv/xUuuwt0q8/J23q9V3da7xlUVR3fvJQ/ftxb9KMlkZd8qu/6oBL7tL7r/6jfv5ot7KDwW20DQAApi+RytBvCAAAzxp9F4AyicSkd1ws7fcR6Z6vSQ99R3rkeunAj0v7f1Rqmem7wjFt4Sadvv8inb7/Ij3fs1G3PtKtWx7u1idufFQzWxr1rj120LI3d2jpjjGFQrxVBACgVuRyTl3prN72xu18lwIAQF0jHKp1c+LScVdIB54n/e4r0t3/Lv3xKunQT0tvPkNqbPZd4au8bt5MferIN+jfDn+9HvpbSj99uEu/fHytlq/qUkc0rGX7LNSyN3dox7kzfJcKAACmqWfjgAaGc6wcAgDAM7aV1Yvt3iid8kPpzDulubtKK86XrthXevwnUi7nu7rXCIVMB+zcrm+cuJdWfuFw/cfJe2mnuTN02d2rddg37tEJ37lfP3zo71qXHfJdKgAAmKLCGPs44RAAAF4RDtWb+FLpA7+UTrtZap4l3XKWdNVbpWd/KwV0WlikuVHH79Oh68/cT/df8DZ95qg3al12SJ+/9S/a9+Lf6pwfPqzf/fUlDY0EL+QCAABbliQcAgAgENhWVo/MpF2PkHZ+u/SXm6Xf/bv0wxOkHQ+R3v5FKb6v7wq3aH5bWB89bGd95NDX6c/d63TLw936+aPd+uWf12ruzGYdu/dCLXvzQu2+oM13qQAAYBsSqYzMpIVzwr5LAQCgrhEO1bNQSNrzJGnxcdKfrpV+f6n0/cOlN/6z9PYLpXlv8F3hFpmZ9uyYoz075uhz79pN9zz9sm55uFvXPfCCvn/v3/TGHWbphDd36Ni9F2i72a2+ywUAAJuRTGW1/axWtTY1+C4FAIC6RjiE0abU+50t7X2q9MCV0v2XSU/vL+19mnTYBVJbh+8Kt6q5MaQjd99BR+6+g9KbBvWLx9fo5oe7dfGKp3TJr57SW18/T8ve3KEjF2/PwycAAAGSZIw9AACBQDiEV7TMkg77jLTvmdIfvimt/J70+PLR4OjgT0qRmO8Ktyk6o1nvO2BHve+AHbX65Y265f+3d+fhcd31vcff39lntEsjWbYWW5bk7Ik3stoJ2UgClBBILgRoLxQuhct+aWmALjTtQ9pL4EIv3LZhLdzcBJqkeUJJSEJiwFmazdljx5YcR4udWPs22vW7f5yxNVocO4nlOdJ8Xs/ze3R05szoOz629NPHv2VbG//+ZDufvelJCqIh3nHqct67oZqNK0sws2yXKyIiktNae1KcVV+W7TJERERynsIhmSsvCZdeB2d8An57HTz0XXjip7Dpc965yOLYRr6hIp8vXXo8f/q24/jP3V3csq2NO57ey82PtVJbmuCKdVW8d301tWX6H0sREZFjbWR8klf6RzRySERExAe0W5kcWslKuOKf4ZMPwcqz4b5r4R/XwWM/hMnFs4V8IGCc3ZDkW/9lLY999SK+edVp1JTG+cf7d3HuN7Zw1T8/xM2PttA/snjek4iIyGLX3juMcygcEhER8QGFQ3J4y06ED9wMH/k1lNTBr/4HfO8MeO42mFpc28fnRUO8d0M1N37sTB748wv4s0uOo2tojGtue5a3/N1v+MxNT7Llxf1MTC6u9yUiIrLYaBt7ERER/9C0MjlyK8+CP/417Lwb7vsbuOUjsPw7cNFfQ/0F2a7udasqjvOp8xv472+t5+m2Pm5LTzv75dN7KS+I8u61K3jP+mpOWF6Y7VJFRESWnAPhkEYOiYiIZJ/CIXl9zOC4S6HxYm+x6i1fh59dAXXnwUVfg6r12a7wdTMz1tYUs7ammK++4wS27Ojg1m1t/PjBPXx/60ucuLyQ96yv4vK1VZQXRLNdroiIyJLQ2jNMJBSgPF8/W0VERLJN4ZC8MYEgrL0aTn4PPP4j+P034Pvnw4nvhgv+EpIN2a7wDYmGglx6ciWXnlxJ99AYdzzVzm1PtvN3v9rOdXft4Lw15bxnfRUXnbCMWDiY7XJFREQWrZauFDUlcQIB7R4qIiKSbQqH5M0JReHMT8LaD8LD3/V2Ntv+S1j/h3DeNVC4PNsVvmGleRE+fE4dHz6njl2vDnDrtnZuf7Kd+3fspyAW4p2nruDKDVWsry3BTB1bERGR16O1J6X1hkRERHxC4ZAcHbFCOP8r8JaPwe+v90YTPf1zOONPYNPnIV6S7QrflMZlBVxz2fH82SXH8VBzJ7elg6KbHm1hVVmC96yv5op1VerkioiIHAHnHC1dKTasXNz9AxERkaVC4ZAcXfkV8Pb/6Y0m2vJ1ePA78MRPYNMXvKAoHM92hW9KMGBsbixnc2M5f/vuCe56dh+3bWvnW/fu5Fv37uT0ulLevbaK844rp6p4cb9XERGRhdI3PM7A6IQWoxYREfEJhUOyMErr4L3fh3M+C/ddC7/5a3jkX+Ct13hT0IKL/69efjTEVRtruGpjDW09KW5/sp1bt7XzlX9/FoDVyTw2NSbZ3FjOmatLKYiFs1yxiIiIP7R2DwNQXaJwSERExA8W/2/o4m+Vp8AH/w32POgFRL/8LDz0v+HCv4QT3uXtfrYEVJck+PQFjXzq/AZ2vjrI1l0dPNDUyb893sZPH36ZYMBYV1PM5sZyNjUmOa26iFAwkO2yRUREsqJF29iLiIj4isIhOTZWnQMfvRdevNMbSfSLP4IV6+Gir8Hq87Jd3VFjZhxXWcBxlQV8bPNqRicmeeLlHh7Y1ckDTZ18+76d/K/f7KQgFuKs1WVsXlPO5oYkK8sSWtRaRERyRmuPFw7VlGoKtoiIiB8oHJJjxwyOfwesuRSevgm2XAc/fRfUX+CtSVR7FgSX1tSraCjI2fVJzq5P8iWgZ2iMB5s7eWBXJ1t3dXLPC68CUF0ST69llOTs+jKKE5HsFi4iIrKAWrpTlCTCmnItIiLiEwqH5NgLBGHdh+DkK+GxH8DW6+Ff/wCiRVB/PjS+DRov9ha3XmJK8iK889QVvPPUFTjn2NOVYuuuDrbu6uSXT+/lpkdbMINTq4oOTkFbX1tCJKQpaCIisnS0dmsbexERET9ROCTZE47B2Z+GDR+G3Vtg592w61544Xbv8RXr0kHR27zjQDCr5R5tZkZdMo+6ZB5/dNYqxieneLq1l63pKWj/9LtmvruliUQkyBl1pQdHFjVU5GsKmoiILGqt3SlOqirKdhkiIiKSpnBIsi+aDyf8gdecg1eehV33eEHR778Bv/sHSJRBw0VeUFR/ASRKs131URcOBti4qpSNq0r5wsVr6B8Z5+HmrvQUtA62vNgBQGVhLL0LWpJzGpIk86NZrlxEROTITU452nuHueyU5dkuRURERNIUDom/mMHyU7127p9Cqhua7/fCoqbfwDM/BwtA9ene1LPGt3k7oi3BkTSFsTCXnFTJJSdVAt7/sj7Q5K1XdO8Lr3LLE20AnLC8kHMbk2xqTPKWVaXEwktrhJWIiCwtr/SPMD7pqNE29iIiIr6hcEj8LVEKp1zptalJ2PukFxTtvBvu/1uvFSyfDopWvxWiBdmuekHUlCa4+vRarj69lskpx3PtfTzQ5I0q+tGDL/Evv99NNBTgLatK2ZwOi06oLCQQWHrBmYiILF4tXdrGXkRExG8UDsniEQhC9Uavnf8VGHjVG0206x54/nbY9lMIhGHlWdB4iRcWJRuX5D1BJ/QAABg9SURBVKiiYMA4raaY02qK+dT5DQyNTvDoS91sTU9Bu+6uHXAXlOVFOKfBm4K2ubGcyqJYtksXEZEcp23sRURE/EfhkCxeBctg3Qe9NjkOrY9Mr1V0z1e9VrxyelHrus0QXpod0bxoiPOPr+D8470d3l7pG0lPQevggaZO7nh6LwANFflsakhy7pokZ9SVkRfVtwARETm2WrtTBAxWFC/Nn8kiIiKLkX4zlKUhGIZVm7x28bXQ2wpN98LOe+CpG+Gx70MoBnXnpsOii6FkVbarXjCVRTGu3FDNlRuqmZpy7HhlgAeaOti6q5ObHm3hJw/tIRw01tWWpNcrKueUqiKCmoImIiILrLU7xfKiOOFgINuliIiISJrCIVmaimtg4x97bXwEXn7QG1G0625vdBFAcs30qKLasyAUyW7NCyQQME5cUciJKwr5+Ln1jIxP8vieHrY2dbB1ZyfX37OT6+/ZSVE8zNn1ZWxuLGdzY5IarQUhIiILoKU7pfWGREREfOaIwiEzuxT4DhAEfuCc+/tZj0eBnwIbgC7gfc65PWa2CtgOvJi+9D+dc584OqWLHKFwDBou9Nplfw9dzenpZ/fAozfAw9+FSL63mPWaS6DhYihcutvrxsJBNqUXrP7yZdA5OMqD6V3Qtu7q5K7nXgFgZVmCcxqSnFBZQH15Pg0V+ZQXRLEluIaTiIgcOy3dw1yYngYtIiIi/nDYcMjMgsD3gIuBNuAxM7vDOfdCxmUfBXqccw1m9n7gH4D3pR9rds6tPcp1i7xxZfVQ9kk485MwOgh7tnq7n+26F3b8h3dN5SnTo4qqNkJw6Q6yS+ZHuXxtFZevrcI5R3PHYHph607ueGov/2904uC1BdEQqyvyaSjPp74i72BoVFua0PQAERE5rOGxSToHR7UYtYiIiM8cyW+8pwNNzrndAGZ2M3A5kBkOXQ58LX18C/Bd0/ACWQyi+XDcZV5zDvZvn17U+oFvw9ZvQqwYGi7ygqKGCyEvme2qF4yZ0VBRQENFAR85pw7nHK/2j9K0f5DmDq817R/kgaYObt3WdvB54aCxsiyP+vLpwKi+PJ/V5XkUxMJZfEciIuIn0zuVaVqZiIiInxxJOFQFtGZ83gaccahrnHMTZtYHlKUfqzOzJ4F+4C+cc1tnfwEz+zjwcYDa2trX9QZEjhozWHai1zZ9HoZ7YfeW9FpF98JztwAGVRu8oGjN26DyNAgs3REzZkZlUYzKohibGmeGYv0j4+zuGKJ5/yBNHYPex/2D3Ld9PxNT7uB1lYUx6ivy0qON8g+GRxWaoiYiknNauxUOiYiI+NFCz5XZB9Q657rMbANwu5md5Jzrz7zIOXcDcAPAxo0b3TyvI3LsxYvhpCu8NjUF+55KB0X3wG+vg99+HfIqvJ3PGi+G+gsgVpTtqo+ZwliYtTXFrK0pnnF+bGKKlu7UjNFGzfsHuXVbO4PzTFGbPdpoZZmmqImILFUt6XBIC1KLiIj4y5GEQ+1ATcbn1elz813TZmYhoAjocs45YBTAOfeEmTUDa4DH32zhIsdUIABV67321j+HoU5ous8Linb8Cp66ESzo7XpWtxmKaqCoCgrTLZI7neBIKEBDhRf2ZDowRS1zelpzxyAPNnVy27bpbymhgLGyLDEjMKpPh0iaoiYisri1dg+TiAQpy1uaO4SKiIgsVkcSDj0GNJpZHV4I9H7gA7OuuQP4r8DDwJXA/c45Z2blQLdzbtLMVgONwO6jVr1ItuQl4bT3eW1yAtofn94B7bfXzb0+XjIdFBWumBkcHTi3xAOkzClq5zTMnKI2kJ6iNntto/t3zJyitqwwOh0YZYRHywo1RU1EZDFo6U5RU5LQ92wRERGfOWw4lF5D6NPA3Xhb2f/IOfe8mV0LPO6cuwP4IfAzM2sCuvECJIBzgWvNbByYAj7hnOteiDcikjXBENSe6bUL/wrGR6C/Hfr3pj+2Q9+Bz9u8ICnVNfd14iVQWD03PCrKCJDCS3N3l4JYmNNqijlt1hS18ckpXu5KzRptNMRts6ao5UdDB6enTa9rlMfKsjxNURMR8ZHW7pTWGxIREfGhI1pzyDl3J3DnrHN/lXE8Alw1z/NuBW59kzWKLC7hGJTVe+1QxoczwqO90NeWcdwObY/B8Dw5arx01qijFVBUPX1cWOV9/SUiHDz0FLX9A6MzF8PuGOSh5i5ue3LmFLXassTBxbBrSxNUFcepKolTVRwnFg4e67ckIpKznHO09qQ4u6Hs8BeLiIjIMbXQC1KLyHzC8cMHSGMpGNiXDo7So44OhEd97dD6CAz3zH1eomzWqKMVc0ckhaIL996OATNjWWGMZYUxzp41RW1wdILm/TPXNWruGJozRQ0gmR+hqiRBdUZgVFUcp7rU+6g1jkREjp6uoTFSY5NajFpERMSHFA6J+FUkcQQB0hD075sZHB2cytYKLQ/DSO/c5yWShxiBtGL680UaIOVHQ/NOUZuYnOKV/hHae4Zp7x2mvWeYtvTxC/v6uXf7q4xNTM14TmEsRFWJN9qousRrmSOPSvMiWjdDROQIHdzGvkThkIiIiN8oHBJZzCJ5kGzw2qGMDU1PYZsRHrVDz8vw8oMw0jf3eXnlXlBUuhrKGtJBVYP3eaJ04d7TAgkFA1SXJKg+xC8lU1OOzqFRLzDKCJDae4dp7U7xcHMnQ2OTM54TDwenRxylP1Yf/JigoiBKIKDwSEQEMraxL1M4JCIi4jcKh0SWukgeJBu9diijgzMX0D6wDlJfK+zdBi/cDi5jVE28NCMwqofSjOAomn/or+NjgYBRURCjoiDG+tqSOY875+gbHj842uhAcNTWk6K9d5hn2nrpSY3PeE44aCwvmg6PMkceVRcnqCyKEQlpwWwRyQ0aOSQiIuJfCodExAt0ytd4bT4To94oo64m6G72PnY1w+7fwdM3zby2YHk6LKrPGG1UD6V1i3aqGnjrHBUnIhQnIpxcVTTvNUOjE+ztHaatd3jWCKQUW3d1sH9gFOcyXxOWFcS80GjOCCRvOls8okWzRWRpaO0eJpkf1fc1ERERH1I4JCKHF4oeOjwaG4Lu3V5Y1NWUPm6CHb+CVOf0dRbw1jU6EBZljjwqqoXg4v92lBcN0bisgMZlBfM+Pjoxyb7ekek1jw6OQErxxMs9/OqZfXMWzS7Li1BbHGZNwRirEyOsjI+wIjJEUSJKfukyCksrCReUe6O5lsCfoYgsXS3dKWpL49kuQ0REROah3yRE5M2J5EHlKV6bbbg3PdKoOSM8aoZnfg6j/dPXBcJQsmrmukYHwqOCFRBYGlOvoqEgq8oSrMobh5IxSPVCqstrQ51MDXUx0vcqY/0dTA11ERjuIjrWS7xrELoO//qDls9QqJjRSDET0RJcIkkgr4xwQTnxogrySyuIFi7z1oxKlEGsyBu+JCJyDLT2pNiwcu60XREREck+hUMisnDixVC1wWuZnIOhzrnT1A5MVZsYnr42FE+HRbOmqZU1QF4y++HGxKj3Xg6EPJntUOenJuZ9qUAwSiIvSSJRCiVJqGr0Qpy8JCRKcfEy+gNF7BtP0D04wnBfB2P9HUwOdmCpbgIj3UTHeogP9VI00EKJPU8p/URt/q83QZBURphEooxAXpJwoRcmxYrKsYNfv8xrYf2vv4i8fuOTU+ztHeaKdVXZLkVERETmoXBIRI49M8gv99rKs2Y+NjUFA3vnTlPbvx1evHNmsBItnLkgduYC2fGZW9kfkakpGOmdJ9zphFT3/IHP2OCh3iTES6bDndLVUL0REsmMwKcsPYonfRzJe82wy4CidDucsYkpuoZG2dk/Sk9vDwM9rzDc28HYQAeTg50EUp0ER3qIjPWQN9BHyWA/pV17KbEBihkkYG7+1w3EGI2UMBEthUQpgfwkkYJyokUVBPLKpkOkA+8pXqLpbiLCvt4RppwWoxYREfEr9dhFxF8C6bWJiqph9XkzH5ucgL6WudPU2h6F524FMgKNRHLmNLXSOpianBXupEOfA+eGu2fuypYpnMgIPsqmRy5lhjuZoU+8BALZW3Q1EgqwvCjO8qI41BQDdYe8dnLK0ZMao3NwlO0DY3QOpBjo7WS4d396ilsnluomNNJFZKyH4vEBSoYGKO1pp5QdlNggcRs+5OuPhQuZiHkjn4L5ScIF5QTzMwKkgmXePVoia0+JyFwHtrGvKVU4JCIi4kfqhYvI4hEMpYOe1dB48czHxkegZ8/caWrN98NTN8681gLeAs4Hgpzy4+aGO7NDn8jS/YUmGDCS+VGS+VGoPHC2dt5rnXP0D0/QOTRK58Aozw56oVJv/wBDfZ2M9+9ncqiTQKqb4Eg3BVO9lEwMUDoyQEnvAGW2nRJ7lFIbIMLM6W6TFmIoUc140SoorSeyrJFE5XEEk/VeWJjFsE1E3pwD4VBt2dL9XioiIrKYKRwSkaUhHIOK47022+igFxyFoumFmIuXzCLXx5qZUZQIU5QIU1+ef9jrU2MTdA6MHQyTWtJhUtfACAMDfYz2dxAa3EvhcAvLx/eyqv8V6gZ2s7L9YRLPjR58nTHCdIRX0BOrIZW/kvHiOgLJBmLL1lC8rJbywjj5Uf1IE/Gr1p4U4aBRWRjLdikiIiIyD/WkRWTpi+ZD5cnZriInJSIhastCRzRaYHhsks7BUfYPjPL7/hEGO1uY6mwm1LubxMAeikdaqRh8mcb+R4juG59+nouwxy2j1ZbTEammL17LcMEqpkpWEy9ZTnlhjPKCKBUF3seyvAihoMJBkWOppTtFVXGcYEA7JIqIiPiRwiEREfGFeCRITWkiY02S5cAZc66bmpyk99U9DOx9kdFXd+G6mkj0vcT6oZcpHtlGaHwC+oF2GHQx9rhK9rhKtrhK9kxVsodK+uI1hPLL5wRH3nH04HFBNIRle0c8kSWgrTul9YZERER8TOGQiIgsKoFgkOIV9RSvqJ/74OQE9LWm155qJtbRREPHLhq7dxMefIyAm0xfB6mBPNqGqtjTXsmL4+U8M7mMPa6Sl1wl/XhT5mLhgBcU5UfnhEjl+VEqCqPp0UhRIiGNRhI5lJbuFJedsjzbZYiIiMghKBwSEZGlIxjydqYrrYOGiwiR8YNuchx6Xj4YHCW6m1nT1cya7mYu7t2KBad3uxuNFNMbq2F/uJpWW07zxDJ2vFrO/bvLaB8Oz/ulS/MiJPMjFMbCFMbDFMZC6Y9hCuMhCmLTxweuKYh5x74Jlpzz/pymxmFyDEIxCMezXZUscgMj4/SkxrWNvYiIiI8pHBIRkdwQDEOywWuz2Izd7pqJdjezrKuZZd3PcEr/XTOudWUVjBXVMZRXS1eshleCVbSwnF0TRewbDtA/PMH+gRGa9k/QPzLO4PAoQTdBhAnCB5pNTh8zQX5oiuKIoygKhWFHfhgKIo6C0BT5oSnyQlMkgl6LByaJBaeIBSaJ2iRRmyDoJrDJMS/YmRxLhzvp44Mfx2adm+f81PjMP5h3fBPe8rEFvCmSC1q7hwGo1bQyERER31I4JCIi8lq73Y2loOcl6GqG7masq5lo926i+7ZSOvgKjZnXJsq80TdTE17Y4sYgOnVkNUwCqSMvecIFGCdEihDjhJiwEFMWYioQZioQxgXCEIxAMIIFI1gon2AoTCAeJRSOEIrECIejhCNRQpEoFjxw/fTzqDnzyAsSOYSD29grHBIREfEthUMiIiKvJZKAZSd5bbbRAejefTA4oq8dAqF0uBKaDlkyA5dgGALzBDEzjqef6wIhRlyIgXGjf8zoH4O+MegfnaJ/ZIL+4XH6R8YZOHg861z/OKMTrx1QBQM2Yxrcgelu7y+p4K2VC/TnKjmjrccLh2pKNUVRRETErxQOiYiIvFHRAlh+mtcWiAHxdKt4g68xMj7pBUUj4/QPj2ccT5+bHTA1dwzSNzx++BcXOQwzo7Ein6L4/Ot1iYiISPYpHBIREVniYuEgsXCQ8oJotkuRHPTRTXV8dFNdtssQERGR1+CT7VFEREREZDYzu9TMXjSzJjO7Zp7Ha81si5k9aWbPmNnb0+fDZvavZvasmW03sy8f++pFRERksVA4JCIiIuJDZhYEvgdcBpwIXG1mJ8667C+AXzjn1gHvB/5P+vxVQNQ5dwqwAfgTM1t1LOoWERGRxUfhkIiIiIg/nQ40Oed2O+fGgJuBy2dd44DC9HERsDfjfJ6ZhfCWrBoD+he+ZBEREVmMFA6JiIiI+FMV0JrxeVv6XKavAR8yszbgTuAz6fO3AEPAPqAFuN451z37C5jZx83scTN7vKOj4yiXLyIiIouFwiERERGRxetq4CfOuWrg7cDPzCyAN+poElgB1AFfNLPVs5/snLvBObfRObexvLz8WNYtIiIiPqJwSERERMSf2oGajM+r0+cyfRT4BYBz7mEgBiSBDwC/ds6NO+f2Aw8CGxe8YhEREVmUFA6JiIiI+NNjQKOZ1ZlZBG/B6TtmXdMCXAhgZifghUMd6fMXpM/nAWcCO45R3SIiIrLIKBwSERER8SHn3ATwaeBuYDvermTPm9m1Zvau9GVfBP6bmT0N3AR82Dnn8HY5yzez5/FCph8755459u9CREREFoNQtgsQERERkfk55+7EW2g689xfZRy/AJwzz/MG8bazFxERETksjRwSEREREREREclhCodERERERERERHKYwiERERERERERkRymcEhEREREREREJIcpHBIRERERERERyWEKh0REREREREREcpjCIRERERERERGRHKZwSEREREREREQkhykcEhERERERERHJYQqHRERERERERERymMIhEREREREREZEcpnBIRERERERERCSHKRwSEREREREREclhCodERERERERERHKYOeeyXcMMZtYBvLyAXyIJdC7g68sbo/viT7ov/qV740+6L0dupXOuPNtFyLQF7oPp34Z/6d74k+6Lf+ne+JPuy5E5ZP/Ld+HQQjOzx51zG7Ndh8yk++JPui/+pXvjT7ovIvPTvw3/0r3xJ90X/9K98SfdlzdP08pERERERERERHKYwiERERERERERkRyWi+HQDdkuQOal++JPui/+pXvjT7ovIvPTvw3/0r3xJ90X/9K98Sfdlzcp59YcEhERERERERGRabk4ckhERERERERERNIUDomIiIiIiIiI5LCcCYfM7FIze9HMmszsmmzXIx4zqzGzLWb2gpk9b2afy3ZNMs3Mgmb2pJn9R7ZrkWlmVmxmt5jZDjPbbmZnZbsmATP7Qvr72HNmdpOZxbJdk4gfqA/mP+p/+Z/6YP6j/pd/qQ92dOREOGRmQeB7wGXAicDVZnZidquStAngi865E4EzgU/p3vjK54Dt2S5C5vgO8Gvn3PHAaegeZZ2ZVQGfBTY6504GgsD7s1uVSPapD+Zb6n/5n/pg/qP+lw+pD3b05EQ4BJwONDnndjvnxoCbgcuzXJMAzrl9zrlt6eMBvG+yVdmtSgDMrBp4B/CDbNci08ysCDgX+CGAc27MOdeb3aokLQTEzSwEJIC9Wa5HxA/UB/Mh9b/8TX0w/1H/y/fUBzsKciUcqgJaMz5vQz8AfcfMVgHrgEeyW4mkfRv4EjCV7UJkhjqgA/hxerj5D8wsL9tF5TrnXDtwPdAC7AP6nHP3ZLcqEV9QH8zn1P/yJfXB/Ef9L59SH+zoyZVwSHzOzPKBW4HPO+f6s11PrjOzdwL7nXNPZLsWmSMErAf+yTm3DhgCtIZHlplZCd5oiDpgBZBnZh/KblUiIq9N/S//UR/Mt9T/8in1wY6eXAmH2oGajM+r0+fEB8wsjNcxudE5d1u26xEAzgHeZWZ78KYAXGBm/ze7JUlaG9DmnDvwP7y34HVWJLsuAl5yznU458aB24Czs1yTiB+oD+ZT6n/5lvpg/qT+l3+pD3aU5Eo49BjQaGZ1ZhbBW6DqjizXJICZGd7c3e3OuW9lux7xOOe+7Jyrds6twvv3cr9zTgm8DzjnXgFazey49KkLgReyWJJ4WoAzzSyR/r52IVqoUgTUB/Ml9b/8S30wf1L/y9fUBztKQtku4Fhwzk2Y2aeBu/FWL/+Rc+75LJclnnOAPwSeNbOn0ue+4py7M4s1ifjdZ4Ab079o7QY+kuV6cp5z7hEzuwXYhrcL0JPADdmtSiT71AfzLfW/RF4/9b98SH2wo8ecc9muQUREREREREREsiRXppWJiIiIiIiIiMg8FA6JiIiIiIiIiOQwhUMiIiIiIiIiIjlM4ZCIiIiIiIiISA5TOCQiIiIiIiIiksMUDomIiIiIiIiI5DCFQyIiIiIiIiIiOez/A0Fnbqm+Lr8DAAAAAElFTkSuQmCC",
            "text/plain": [
              "<Figure size 1440x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "#wytrenowanie modeli\n",
        "modelNoPool = buildConvoNoPool()\n",
        "modelMaxPool = buildConvoMaxPool()\n",
        "modelAvgPool = buildConvoAvgPool()\n",
        "\n",
        "modelNoPool.fit(X_train,y_train,epochs=10,validation_data=[X_test,y_test])\n",
        "modelMaxPool.fit(X_train,y_train,epochs=10,validation_data=[X_test,y_test])\n",
        "modelAvgPool.fit(X_train,y_train,epochs=10,validation_data=[X_test,y_test])\n",
        "\n",
        "\n",
        "#dodać generowanie wykresów krzywych uczenia\n",
        "generateNetworkLearningCurve(modelNoPool,10)\n",
        "generateNetworkLearningCurve(modelMaxPool,10)\n",
        "generateNetworkLearningCurve(modelAvgPool,10)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XyFvDaAm3JoO"
      },
      "source": [
        "### **Wnioski zad. 7.3**\n",
        "Na podstawie wygenerowanych wykresów można zauważyć, że najbardziej stablilnymi sieciami są sieci w wykorzystaniem AvgPooling oraz MaxPooling w tych przypadkach funkcja strat w przypadku przetrenowania ma tendencję malejącą. Ponadto w zestawieniu funkcji strat dla AvgPooling oraz MaxPooling  lepszą jest AvgPooling, ponieważ zjawisko przetrenowania zachodzi na niższym poziomie niż w przypadku MaxPooling oraz przetrenowanie zachodzi dopiero dla epok wyższych niż 8 (gdzie w kontraście dla Maxpooling przetrenowanie zachodziło już w 3 epoce).\n",
        "\n",
        "W przypadku zestawienia wykresów funkcji dokładności możemy zaobserwować, że warstwy pooling wpływają pozytywnie na dokładność dla zbioru testowego. Ponadto najlepsze zestawienie dla sieci neuronowej wykorzystuje warstwy AvgPooling, ponieważ zjawisko przetrenowania zachodzi dopiero w 6 epoce i jest mniej odczuwalne niż w przypadku MaxPooling.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "N-PfZ5zAdac3"
      },
      "outputs": [],
      "source": [
        "\"\"\" (strona 74)\n",
        "Zadanie 7.4. Utworzenie modelu sieci konwolucyjnej do klasyfikacji zbioru danych\n",
        "CIFAR10\n",
        "Załaduj zbiór danych CIFAR10 za pomocą kodu, który przedstawia Listing 7.4. Zbuduj\n",
        "konwolucyjną sieć neuronową, która pozwala na jego klasyfikację. Znajdź optymalną liczbę\n",
        "warstw. Wytrenuj sieć o takiej liczbie warstw. Przetestuj działanie sieci na zbiorze testowym.\n",
        "Wygeneruj macierz pomyłek. Jakie klasy są klasyfikowane najłatwiej, jakie pary klas\n",
        "są mylone najczęściej?\n",
        "\"\"\"\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aJcYXd9x3_3i",
        "outputId": "899daa64-941e-4762-8f4c-dc8bc0fdda88"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170498071/170498071 [==============================] - 5s 0us/step\n"
          ]
        }
      ],
      "source": [
        "# załadowanie zbioru danych\n",
        "from keras.datasets import cifar10\n",
        "(x_train, y_train),(x_test,y_test) = cifar10.load_data()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "1_GlzT1KYcqT"
      },
      "outputs": [],
      "source": [
        "#funkcja budująca sieć konwolusyjną\n",
        "from keras.layers import Conv2D, Flatten, Dense,MaxPooling2D\n",
        "from keras.models import Sequential\n",
        "from keras.optimizers import Adam\n",
        "\n",
        "\n",
        "def buildConvo(layer_num=1):\n",
        "    #papametry sieci konwolucyjnej\n",
        "    class_cnt = np.unique(y_train).shape[0]\n",
        "    filter_cnt = 32\n",
        "    neuron_cnt = 32\n",
        "    learning_rate = 0.0001\n",
        "    act_func = 'relu'\n",
        "    kernel_size = (3,3)\n",
        "    model = Sequential()\n",
        "    conv_rule = 'same'\n",
        "    pooling_size=(2,2)\n",
        "    \n",
        "    #dodanie warstw do sieci konwolucyjnej\n",
        "    model.add(Conv2D(input_shape = x_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    for a in range(layer_num):\n",
        "      model.add(Conv2D(input_shape = x_train.shape[1:],filters=filter_cnt,kernel_size = kernel_size,padding = conv_rule, activation = act_func))\n",
        "    \n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(class_cnt, activation='softmax'))\n",
        "    model.compile(optimizer=Adam(learning_rate),loss='SparseCategoricalCrossentropy',metrics='accuracy')\n",
        "    return model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XR2NTV194t93",
        "outputId": "2806e91a-12bc-4f60-ee43-b328e56313a0"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "<ipython-input-7-48c397c061eb>:6: DeprecationWarning: KerasClassifier is deprecated, use Sci-Keras (https://github.com/adriangb/scikeras) instead. See https://www.adriangb.com/scikeras/stable/migration.html for help migrating.\n",
            "  keras_classifier = KerasClassifier(buildConvo)\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 11s 3ms/step - loss: 12.7383 - accuracy: 0.3302\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 3.5345 - accuracy: 0.4566\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.9357 - accuracy: 0.5231\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 1.4121 - accuracy: 0.5781\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.1070 - accuracy: 0.6431\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.9268 - accuracy: 0.6934\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.7819 - accuracy: 0.7332\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.6561 - accuracy: 0.7762\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.5799 - accuracy: 0.8025\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 0.4869 - accuracy: 0.8326\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.3946 - accuracy: 0.5075\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 8.8938 - accuracy: 0.3271\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 2.2510 - accuracy: 0.4495\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.4048 - accuracy: 0.5449\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.1286 - accuracy: 0.6182\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.9400 - accuracy: 0.6770\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.7984 - accuracy: 0.7229\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.6746 - accuracy: 0.7666\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.5750 - accuracy: 0.8037\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.4807 - accuracy: 0.8373\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.4249 - accuracy: 0.8577\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.3830 - accuracy: 0.4932\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 7.4130 - accuracy: 0.3317\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 2.0033 - accuracy: 0.4429\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.4666 - accuracy: 0.5266\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.2029 - accuracy: 0.5966\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 1.0263 - accuracy: 0.6508\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.8580 - accuracy: 0.7070\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.7267 - accuracy: 0.7498\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.6146 - accuracy: 0.7894\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.5188 - accuracy: 0.8224\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.4479 - accuracy: 0.8511\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.2954 - accuracy: 0.5043\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 6.7954 - accuracy: 0.3370\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.8819 - accuracy: 0.4603\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.3514 - accuracy: 0.5570\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.0870 - accuracy: 0.6320\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.9024 - accuracy: 0.6913\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.7533 - accuracy: 0.7394\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.6370 - accuracy: 0.7854\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.5394 - accuracy: 0.8165\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 0.4529 - accuracy: 0.8479\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 0.3901 - accuracy: 0.8713\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.3645 - accuracy: 0.5065\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 9.3500 - accuracy: 0.3306\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 2.3128 - accuracy: 0.4466\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.5076 - accuracy: 0.5260\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.2176 - accuracy: 0.5957\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 1.0149 - accuracy: 0.6539\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.8621 - accuracy: 0.7066\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.7436 - accuracy: 0.7454\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.6218 - accuracy: 0.7869\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 3s 2ms/step - loss: 0.5311 - accuracy: 0.8199\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 3s 3ms/step - loss: 0.4509 - accuracy: 0.8488\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.1857 - accuracy: 0.5193\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 4.4584 - accuracy: 0.3354\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.4180 - accuracy: 0.5148\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.1060 - accuracy: 0.6170\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.8841 - accuracy: 0.6926\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.7099 - accuracy: 0.7565\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.5716 - accuracy: 0.8084\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.4546 - accuracy: 0.8476\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.3638 - accuracy: 0.8818\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2866 - accuracy: 0.9095\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2239 - accuracy: 0.9312\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.1153 - accuracy: 0.5416\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 3.9788 - accuracy: 0.3372\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.4441 - accuracy: 0.5010\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.1425 - accuracy: 0.6006\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.9195 - accuracy: 0.6815\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.7371 - accuracy: 0.7484\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.5840 - accuracy: 0.8030\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.4632 - accuracy: 0.8459\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.3581 - accuracy: 0.8837\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2790 - accuracy: 0.9124\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2224 - accuracy: 0.9332\n",
            "313/313 [==============================] - 1s 2ms/step - loss: 2.2639 - accuracy: 0.5253\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 3.4208 - accuracy: 0.3605\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.2983 - accuracy: 0.5550\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.9822 - accuracy: 0.6643\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.7574 - accuracy: 0.7405\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.5717 - accuracy: 0.8074\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.4271 - accuracy: 0.8593\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.3146 - accuracy: 0.8989\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2261 - accuracy: 0.9310\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.1687 - accuracy: 0.9516\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.1249 - accuracy: 0.9674\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.1785 - accuracy: 0.5687\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 2.8809 - accuracy: 0.3667\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.3269 - accuracy: 0.5371\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.0474 - accuracy: 0.6357\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.8400 - accuracy: 0.7102\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.6824 - accuracy: 0.7673\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.5461 - accuracy: 0.8162\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.4261 - accuracy: 0.8615\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.3387 - accuracy: 0.8928\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2666 - accuracy: 0.9178\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2078 - accuracy: 0.9387\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.2800 - accuracy: 0.5311\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 3.3783 - accuracy: 0.3425\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.3928 - accuracy: 0.5195\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 1.1084 - accuracy: 0.6171\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.9053 - accuracy: 0.6890\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.7420 - accuracy: 0.7480\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.6044 - accuracy: 0.7958\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.4846 - accuracy: 0.8407\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.3869 - accuracy: 0.8758\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.3007 - accuracy: 0.9062\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 4s 3ms/step - loss: 0.2345 - accuracy: 0.9291\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.0161 - accuracy: 0.5451\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.9258 - accuracy: 0.3722\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.2901 - accuracy: 0.5493\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.9814 - accuracy: 0.6618\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.7499 - accuracy: 0.7448\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.5682 - accuracy: 0.8095\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.4222 - accuracy: 0.8614\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.3128 - accuracy: 0.8996\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.2333 - accuracy: 0.9282\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1767 - accuracy: 0.9482\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1346 - accuracy: 0.9617\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.4473 - accuracy: 0.5459\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 2.7290 - accuracy: 0.3686\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.2589 - accuracy: 0.5613\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.9837 - accuracy: 0.6603\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.7907 - accuracy: 0.7283\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.6254 - accuracy: 0.7900\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.4713 - accuracy: 0.8443\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.3448 - accuracy: 0.8901\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.2505 - accuracy: 0.9231\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1796 - accuracy: 0.9468\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1316 - accuracy: 0.9636\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.2307 - accuracy: 0.5580\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 2.0270 - accuracy: 0.3878\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.2055 - accuracy: 0.5780\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.9127 - accuracy: 0.6848\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.6984 - accuracy: 0.7625\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.5274 - accuracy: 0.8228\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.3926 - accuracy: 0.8735\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.2833 - accuracy: 0.9105\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.2010 - accuracy: 0.9414\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1459 - accuracy: 0.9590\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1087 - accuracy: 0.9705\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.1927 - accuracy: 0.5705\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 6s 4ms/step - loss: 3.1497 - accuracy: 0.3523\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.4021 - accuracy: 0.5115\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.1690 - accuracy: 0.5908\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.9806 - accuracy: 0.6575\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.8235 - accuracy: 0.7155\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.6856 - accuracy: 0.7630\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.5632 - accuracy: 0.8093\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.4525 - accuracy: 0.8488\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.3571 - accuracy: 0.8828\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.2734 - accuracy: 0.9132\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.1582 - accuracy: 0.5285\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 2.9399 - accuracy: 0.3675\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.3110 - accuracy: 0.5378\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 1.0695 - accuracy: 0.6268\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.8821 - accuracy: 0.6940\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.7220 - accuracy: 0.7531\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.5723 - accuracy: 0.8072\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.4376 - accuracy: 0.8564\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.3267 - accuracy: 0.8958\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.2421 - accuracy: 0.9259\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 5s 4ms/step - loss: 0.1703 - accuracy: 0.9511\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 1.9672 - accuracy: 0.5565\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 2.0787 - accuracy: 0.3977\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.1810 - accuracy: 0.5892\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.9273 - accuracy: 0.6813\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.7380 - accuracy: 0.7463\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.5734 - accuracy: 0.8069\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.4305 - accuracy: 0.8587\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.3062 - accuracy: 0.9048\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.2142 - accuracy: 0.9347\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1505 - accuracy: 0.9564\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1101 - accuracy: 0.9693\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.0987 - accuracy: 0.5673\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.9300 - accuracy: 0.4145\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.1589 - accuracy: 0.5986\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.9322 - accuracy: 0.6787\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.7526 - accuracy: 0.7430\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.5824 - accuracy: 0.8036\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.4368 - accuracy: 0.8555\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.3174 - accuracy: 0.9000\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.2177 - accuracy: 0.9351\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1523 - accuracy: 0.9575\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1177 - accuracy: 0.9678\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.0934 - accuracy: 0.5685\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 1.7726 - accuracy: 0.4240\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 6s 4ms/step - loss: 1.1121 - accuracy: 0.6142\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 6s 4ms/step - loss: 0.8432 - accuracy: 0.7112\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.6397 - accuracy: 0.7842\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.4685 - accuracy: 0.8453\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.3328 - accuracy: 0.8935\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.2270 - accuracy: 0.9311\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1586 - accuracy: 0.9535\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1149 - accuracy: 0.9679\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.0877 - accuracy: 0.9762\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.2281 - accuracy: 0.5897\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.7654 - accuracy: 0.4281\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.1325 - accuracy: 0.6085\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 6s 4ms/step - loss: 0.8871 - accuracy: 0.6959\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 6s 4ms/step - loss: 0.6995 - accuracy: 0.7624\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.5351 - accuracy: 0.8208\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.3905 - accuracy: 0.8737\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.2743 - accuracy: 0.9140\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1955 - accuracy: 0.9402\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.1458 - accuracy: 0.9566\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1062 - accuracy: 0.9692\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.3083 - accuracy: 0.5766\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.6151 - accuracy: 0.4514\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 1.0711 - accuracy: 0.6271\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.8280 - accuracy: 0.7170\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.6309 - accuracy: 0.7881\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.4639 - accuracy: 0.8499\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.3309 - accuracy: 0.8965\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.2268 - accuracy: 0.9327\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.1581 - accuracy: 0.9553\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1133 - accuracy: 0.9697\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 6s 5ms/step - loss: 0.0888 - accuracy: 0.9763\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.2834 - accuracy: 0.5689\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.7501 - accuracy: 0.4245\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 1.1835 - accuracy: 0.5844\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.9608 - accuracy: 0.6679\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 9s 7ms/step - loss: 0.7733 - accuracy: 0.7354\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 7ms/step - loss: 0.6069 - accuracy: 0.7959\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 9s 7ms/step - loss: 0.4545 - accuracy: 0.8475\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.3261 - accuracy: 0.8949\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2335 - accuracy: 0.9268\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1666 - accuracy: 0.9495\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1242 - accuracy: 0.9639\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 1.9853 - accuracy: 0.5760\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 10s 8ms/step - loss: 1.9434 - accuracy: 0.4096\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.1955 - accuracy: 0.5834\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.9586 - accuracy: 0.6691\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.7750 - accuracy: 0.7340\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 7ms/step - loss: 0.6103 - accuracy: 0.7923\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.4604 - accuracy: 0.8477\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.3353 - accuracy: 0.8904\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.2362 - accuracy: 0.9237\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.1666 - accuracy: 0.9481\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1279 - accuracy: 0.9598\n",
            "313/313 [==============================] - 1s 4ms/step - loss: 2.1255 - accuracy: 0.5710\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 1.8151 - accuracy: 0.4132\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 1.2094 - accuracy: 0.5760\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.9773 - accuracy: 0.6608\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.7803 - accuracy: 0.7305\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.5963 - accuracy: 0.7973\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.4345 - accuracy: 0.8530\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.3039 - accuracy: 0.9018\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.2126 - accuracy: 0.9320\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1485 - accuracy: 0.9562\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1091 - accuracy: 0.9673\n",
            "313/313 [==============================] - 1s 4ms/step - loss: 2.1965 - accuracy: 0.5703\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 1.6054 - accuracy: 0.4475\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 1.1130 - accuracy: 0.6108\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.8615 - accuracy: 0.7025\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.6522 - accuracy: 0.7801\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.4743 - accuracy: 0.8432\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.3345 - accuracy: 0.8924\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.2264 - accuracy: 0.9291\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1599 - accuracy: 0.9506\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 7s 6ms/step - loss: 0.1138 - accuracy: 0.9660\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.0909 - accuracy: 0.9724\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.2549 - accuracy: 0.5753\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.6594 - accuracy: 0.4333\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 1.1369 - accuracy: 0.6019\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.8909 - accuracy: 0.6938\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.6877 - accuracy: 0.7656\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.5079 - accuracy: 0.8305\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.3554 - accuracy: 0.8846\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.2470 - accuracy: 0.9222\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1704 - accuracy: 0.9470\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.1230 - accuracy: 0.9632\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 7s 5ms/step - loss: 0.0924 - accuracy: 0.9738\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.1696 - accuracy: 0.5801\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.7151 - accuracy: 0.4117\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.1993 - accuracy: 0.5776\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.9661 - accuracy: 0.6654\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.7715 - accuracy: 0.7355\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.5918 - accuracy: 0.7972\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.4281 - accuracy: 0.8587\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2985 - accuracy: 0.9018\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2009 - accuracy: 0.9368\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1423 - accuracy: 0.9562\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1130 - accuracy: 0.9651\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.1346 - accuracy: 0.5713\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.6528 - accuracy: 0.4322\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.1837 - accuracy: 0.5835\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.9626 - accuracy: 0.6663\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.7735 - accuracy: 0.7343\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.6012 - accuracy: 0.7957\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.4393 - accuracy: 0.8556\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.3068 - accuracy: 0.9002\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2055 - accuracy: 0.9349\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1588 - accuracy: 0.9485\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1118 - accuracy: 0.9661\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 2.0223 - accuracy: 0.5782\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.7120 - accuracy: 0.4175\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.2551 - accuracy: 0.5580\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.0424 - accuracy: 0.6369\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.8721 - accuracy: 0.6990\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.7191 - accuracy: 0.7544\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.5646 - accuracy: 0.8069\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.4265 - accuracy: 0.8581\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.3087 - accuracy: 0.8999\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2292 - accuracy: 0.9272\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1620 - accuracy: 0.9503\n",
            "313/313 [==============================] - 1s 3ms/step - loss: 1.8992 - accuracy: 0.5878\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.6314 - accuracy: 0.4381\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.1451 - accuracy: 0.5995\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.9248 - accuracy: 0.6826\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.7356 - accuracy: 0.7490\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.5648 - accuracy: 0.8105\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.4109 - accuracy: 0.8650\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2848 - accuracy: 0.9103\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1987 - accuracy: 0.9372\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1402 - accuracy: 0.9560\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1118 - accuracy: 0.9665\n",
            "313/313 [==============================] - 1s 4ms/step - loss: 2.1407 - accuracy: 0.5862\n",
            "Epoch 1/10\n",
            "1250/1250 [==============================] - 9s 6ms/step - loss: 1.9325 - accuracy: 0.4010\n",
            "Epoch 2/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.2947 - accuracy: 0.5415\n",
            "Epoch 3/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 1.0857 - accuracy: 0.6214\n",
            "Epoch 4/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.9156 - accuracy: 0.6820\n",
            "Epoch 5/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.7541 - accuracy: 0.7436\n",
            "Epoch 6/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.6061 - accuracy: 0.7939\n",
            "Epoch 7/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.4582 - accuracy: 0.8446\n",
            "Epoch 8/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.3287 - accuracy: 0.8905\n",
            "Epoch 9/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.2365 - accuracy: 0.9237\n",
            "Epoch 10/10\n",
            "1250/1250 [==============================] - 8s 6ms/step - loss: 0.1655 - accuracy: 0.9474\n",
            "313/313 [==============================] - 1s 4ms/step - loss: 1.9908 - accuracy: 0.5796\n",
            "Epoch 1/10\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 1.6159 - accuracy: 0.4494\n",
            "Epoch 2/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 1.1019 - accuracy: 0.6162\n",
            "Epoch 3/10\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.8973 - accuracy: 0.6901\n",
            "Epoch 4/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 0.7278 - accuracy: 0.7518\n",
            "Epoch 5/10\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.5647 - accuracy: 0.8073\n",
            "Epoch 6/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 0.4185 - accuracy: 0.8600\n",
            "Epoch 7/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 0.3021 - accuracy: 0.8999\n",
            "Epoch 8/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 0.2147 - accuracy: 0.9303\n",
            "Epoch 9/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 0.1561 - accuracy: 0.9501\n",
            "Epoch 10/10\n",
            "1563/1563 [==============================] - 10s 7ms/step - loss: 0.1165 - accuracy: 0.9632\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "{'layer_num': 5}"
            ]
          },
          "execution_count": 7,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# wyznaczenie optymalnej ilości warstw\n",
        "from keras.wrappers.scikit_learn import KerasClassifier #depricated\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "import numpy as np\n",
        "\n",
        "keras_classifier = KerasClassifier(buildConvo)\n",
        "param_distribs = {\n",
        "    'layer_num':[0,1,2,3,4,5]\n",
        "}\n",
        "\n",
        "grid_search_cv = GridSearchCV(keras_classifier,param_distribs)\n",
        "grid_result = grid_search_cv.fit(x_train,y_train,epochs=10)\n",
        "best_params_from_grid = grid_search_cv.best_params_\n",
        "best_model_from_grid = grid_search_cv.best_estimator_\n",
        "\n",
        "#najlepsze parametry znalezione\n",
        "best_params_from_grid"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "dCJiDmU_dBAd"
      },
      "outputs": [],
      "source": [
        "\"\"\"\n",
        "Epoch 8/10\n",
        "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0061 - accuracy: 0.9978\n",
        "Epoch 9/10\n",
        "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0060 - accuracy: 0.9982\n",
        "Epoch 10/10\n",
        "1875/1875 [==============================] - 12s 6ms/step - loss: 0.0038 - accuracy: 0.9987\n",
        "{'layer_num': 5}\n",
        "\"\"\"\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "M1NiQ_i9dKMY",
        "outputId": "ac0c4d03-39e6-4018-ec3a-15c6746f6ee5"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/8\n",
            "1563/1563 [==============================] - 12s 7ms/step - loss: 1.6495 - accuracy: 0.4426 - val_loss: 1.3442 - val_accuracy: 0.5229\n",
            "Epoch 2/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 1.1588 - accuracy: 0.5932 - val_loss: 1.1582 - val_accuracy: 0.5928\n",
            "Epoch 3/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.9465 - accuracy: 0.6749 - val_loss: 1.1277 - val_accuracy: 0.6029\n",
            "Epoch 4/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.7728 - accuracy: 0.7358 - val_loss: 1.1159 - val_accuracy: 0.6199\n",
            "Epoch 5/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.6108 - accuracy: 0.7926 - val_loss: 1.1929 - val_accuracy: 0.6096\n",
            "Epoch 6/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.4633 - accuracy: 0.8445 - val_loss: 1.2980 - val_accuracy: 0.6120\n",
            "Epoch 7/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.3380 - accuracy: 0.8875 - val_loss: 1.4364 - val_accuracy: 0.6018\n",
            "Epoch 8/8\n",
            "1563/1563 [==============================] - 11s 7ms/step - loss: 0.2439 - accuracy: 0.9185 - val_loss: 1.6058 - val_accuracy: 0.6018\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f69e676d700>"
            ]
          },
          "execution_count": 8,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# wytrenowanie model z najlepszymi parametrami\n",
        "best_model=buildConvo(**best_params_from_grid)\n",
        "best_model.fit(x_train,y_train,epochs=8,validation_data=(x_test,y_test)) #  8 epok to wartość optymalna (na podstawie zadania 7.2 dla sieci o takich samych parametrach)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sUWdT3GQ5Qnu",
        "outputId": "a1aa39cd-890b-4f75-9245-f834a135fb99"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "313/313 [==============================] - 1s 2ms/step\n",
            "[[696  26  67  28  27   6  16  14  74  46]\n",
            " [ 39 738  11  22   5   7  10  10  39 119]\n",
            " [107   5 452 101 100  67  93  44  18  13]\n",
            " [ 34  20 106 455  67 153  95  44  17   9]\n",
            " [ 27   4  98  90 518  44 102 102  10   5]\n",
            " [ 17   6 101 239  57 430  44  84  11  11]\n",
            " [ 10   6  73  81  53  20 726  20   7   4]\n",
            " [ 25   6  59  71  68  67  19 667   0  18]\n",
            " [171  42  13  24  16   8  14   7 657  48]\n",
            " [ 56 125  21  24   8   7  10  32  38 679]]\n"
          ]
        }
      ],
      "source": [
        "# wygenerowanie macierzy pomyłek\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "import numpy as np\n",
        "\n",
        "y_predict = np.argmax(best_model.predict(x_test),axis=1)\n",
        "print(confusion_matrix(y_test,y_predict))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "qggmcOXejGSF"
      },
      "outputs": [],
      "source": [
        "\"\"\"\n",
        "313/313 [==============================] - 1s 2ms/step\n",
        "[[696  26  67  28  27   6  16  14  74  46]\n",
        " [ 39 738  11  22   5   7  10  10  39 119]\n",
        " [107   5 452 101 100  67  93  44  18  13]\n",
        " [ 34  20 106 455  67 153  95  44  17   9]\n",
        " [ 27   4  98  90 518  44 102 102  10   5]\n",
        " [ 17   6 101 239  57 430  44  84  11  11]\n",
        " [ 10   6  73  81  53  20 726  20   7   4]\n",
        " [ 25   6  59  71  68  67  19 667   0  18]\n",
        " [171  42  13  24  16   8  14   7 657  48]\n",
        " [ 56 125  21  24   8   7  10  32  38 679]]\n",
        " \"\"\""
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kAsbqkLz_5Ng"
      },
      "source": [
        "### **Wnioski zad. 7.4**\n",
        "Optymalna liczba warstw to 5 warstw ukrytych (na podstawie grid search).\n",
        "\n",
        "Jakie klasy są klasyfikowane najłatwiej?\n",
        "\n",
        "Najłatwiej klasyfikowane klasy (kolumny w macierzy pomyłek) to te, których wartości na przekątnej macierzy pomyłek są największe. \n",
        "Są to kolejno (licząc od 1 do n):\n",
        " - klasa 2 (z wynikiem 738 klasyfikacji)\n",
        " - klasa 7 (z wynikiem 726 klasyfikacji)\n",
        " - klasa 1 (z wynikiem 696 klasyfikacji)\n",
        " - klasa 10 (z wynikiem 679 klasyfikacji)\n",
        " - klasa 8 (z wynikiem 667 klasyfikacji) \n",
        "\n",
        "Jakie pary klas są mylone najczęściej?\n",
        "\n",
        "Najbardziej mylone klasy to te, których wartości poza przekątną są największe.\n",
        "Są to kolejno (licząc od 1 do n):\n",
        " - klasa 4 (z wynikiem 239 błędnych klasyfikacji)\n",
        " - klasa 1 (z wynikiem 171 błędnych klasyfikacji)\n",
        " - klasa 6 (z wynikiem 153 błędnych klasyfikacji)\n",
        " - klasa 2 (z wynikiem 125 błędnych klasyfikacji)\n",
        " - klasa 10 (z wynikiem 119 błędnych klasyfikacji)\n",
        "\n",
        "Dodatkowo możemy zaobserwować, że klasa 1, która mimo wysokiej liczby najłatwiejszych klasyfikacji jest także klasą której klasyfikacje są najbardziej mylone.\n"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": []
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}