lectures/cap09_3_multilayer_perceptron_keras.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "cap7_2_multilayerperceptron.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyMUURajqNoP9irvqgjSolYX",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/viniciusrpb/cic0269_natural_language_processing/blob/main/lectures/cap09_3_multilayer_perceptron_keras.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Capítulo 9 - Redes Neurais Artificiais\n",
        "\n",
        "Neste notebook Python, iremos descrever sobre as redes multilayer perceptron e como podemos gerar redes neurais profundas.\n",
        "\n",
        "Para isso, vamos precisar instalar as bibliotecas keras, tensorflow:lioteca numpy:"
      ],
      "metadata": {
        "id": "IipByK9A-0GD"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AuV9oScUNg6G"
      },
      "source": [
        "!pip install -U keras\n",
        "!pip install tensorflow"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.models import Sequential\n",
        "from keras.layers import Dense,Activation\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "import numpy as np"
      ],
      "metadata": {
        "id": "rX3Wa1BUB0bx"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 9.3. Multilayer Perceptron\n",
        "\n",
        "Vamos agora trabalhar com o conjunto de dados XOR por ele ser não-linearmente separável:"
      ],
      "metadata": {
        "id": "5y6N5CPhBqQR"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZM_Ry07qQsWZ"
      },
      "source": [
        "X = np.array([[0,0],[0,1],[1,0],[1,1]])\n",
        "labels = np.array([0,1,1,1])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Vamos implementar a seguinte arquitetura baseada em uma Multilayer Perceptron para a tarefa de classificação da tabela XOR.\n",
        "\n",
        "![1_B5uB2VJdROs3JT10YQy87w.jpeg](data:image/jpeg;base64,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)\n",
        "\n",
        "Fonte: https://medium.com/analytics-vidhya/coding-a-neural-network-for-xor-logic-classifier-from-scratch-b90543648e8a\n",
        "\n",
        "Primeiramente, vamos criar o modelo que compreende a nossa rede multilayer perceptron. Como se trata de uma rede do tipo feed forward, utilizaremos o modelo [Sequential](https://keras.io/guides/sequential_model/), em que será possível ir enfileirando (ou empilhando) camadas de neurônio desde a camada de entrada até a camada de saída:\n"
      ],
      "metadata": {
        "id": "2HyxTYcyCmAH"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "32HeC0aLOvIZ"
      },
      "source": [
        "model = Sequential()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "No Keras-Tensorflow, a camada de entrada não é diretamente declarada. Assim, vamos diretamente incluir a primeira camada intermediária. Temos algumas decisões a serem tomadas:\n",
        "\n",
        "*   a quantidade de neurônios dessa camada intermediária\n",
        "*   o tipo da função de ativação\n",
        "\n",
        "Podemos ver na figura acima que são necessários dois neurônios na camada intermediária. Utilizaremos a tradicional **tangente hiperbólica** como função de ativação.\n",
        "\n",
        "No Keras-Tensorflow, a classe [Dense](https://keras.io/api/layers/core_layers/dense/) implementa uma camada de neurônios totalmente conexa (fully connected). Precisamos passar como argumentos:\n",
        "\n",
        "*   ```number_of_neurons```: a quantidade de neurônios nessa camada intermediária.\n",
        "*   ```input_dim```: a dimensão do *embedding* (espaço de valores) que chegam à essa camada. Como se trata da primeira camada intermediária, a dimensão de entrada é a própria dimensionalidade (quantidade de atributos) dos dados de entrada da rede.\n",
        "*   ```activation```: a função de ativação. Consulte a documentação da classe [Dense](https://keras.io/api/layers/core_layers/dense/) para outras possibilidades.\n"
      ],
      "metadata": {
        "id": "t72AglHhC8qO"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "number_of_neurons = 2\n",
        "\n",
        "model.add(Dense(number_of_neurons,input_dim=2, activation=\"tanh\"))"
      ],
      "metadata": {
        "id": "HCO4F0HdC8xV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Por fim, temos que modelar a camada de saída. Como o problema XOR consiste em instâncias das classes \"0\" ou \"1\", temos um problema de classificação binária. Para isso, basta termos um único neurônio na camada de saída.\n",
        "\n",
        "Assim, para incluir a camada de saída, utilizamos novamente uma camada do tipo Dense, mas agora a função de ativação é a sigmoide. Não se preocupe com as conexões entre as camadas do Keras... o próprio Keras se encarrega de conectar todas as camadas Dense (Fully Connected) entre si! "
      ],
      "metadata": {
        "id": "DvhB-TmbDtQL"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.add(Dense(1,activation=\"sigmoid\"))"
      ],
      "metadata": {
        "id": "XaRimKCRDtXr"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Uma visão geral de como está a arquitetura construída para essa Feed Forward Multilayer Perceptron:"
      ],
      "metadata": {
        "id": "jELZeuofHnMa"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8CCi5jxDHlfK",
        "outputId": "3fec3366-5f4a-4b7a-9462-4f31d0fdc691"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense (Dense)               (None, 2)                 6         \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 1)                 3         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 9\n",
            "Trainable params: 9\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Em relação à quantidade de parâmetros. Como cada neurônio possui 2 pesos, teremos no total: \n",
        "\n",
        "*   Cada neurônio contém $2$ pesos. Sendo $3$ neurônios, temos $6$ pesos;\n",
        "*   Temos também um bias para cada neurônio. Sendo $3$ neurônios, temos mais $3$ pesos;\n",
        "\n",
        "Isso totaliza $9$ parâmetros cujos valores serão ajustados durante o treinamento, realizado pelo algoritmo backpropagation. Lembre-se:\n",
        "\n",
        "\n",
        "**Parâmetros do modelo $\\neq$ Hiperparâmetros do modelo**\n",
        "\n",
        "Agora vamos treinar nossa rede utilizando o método de otimização [Stochastic Gradient Descent](https://keras.io/api/optimizers/sgd/), uma implementação do BackPropagation que suporta os modelos do Keras-Tensorflow.\n",
        "\n",
        "Precisamos ajustar um argumento relacionado ao hiperparâmetro da taxa de aprendizado. Colocaremos a *learning rate* igual a $0,01$:"
      ],
      "metadata": {
        "id": "HhOFSCFLEbnx"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TpCbr2gNQ5k2"
      },
      "source": [
        "sgd = SGD(learning_rate=0.01)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Agora vamos \"compilar\" nosso modelo de forma a prepará-lo para o processo de treinamento utilizando o método *compile*. Precisamos definir uma função *loss*, que nada mais é do que a função Erro que avalia como está a performance do modelo criado na classificação dos labels da tabela XOR. Só que agora estamos com uma Multilayer Perceptron que tem em sua camada de saída uma função sigmóide, que gera valores reais: a probabilidade da instância $\\mathbf{x_i}$ processada ser da classe $1$, isto é, $P(y_i=1|\\mathbf{x_i})$.\n",
        "\n",
        "No caso, estaremos utilizando a *binary_crossentropy* (BICE) como a função erro por termos um problema de classificação binária (em duas classes: labels $0$ ou $1$):\n",
        "\n",
        "\\begin{equation}\n",
        "BICE = - \\sum_{i=1}^N y_i \\log(p_i) + (1-y_i) \\log (1-p_i)\n",
        "\\end{equation}\n",
        "\n",
        "em que $y_i \\in \\{0,1\\}$ é a classe predita pela rede, $p_i \\in [0,1]$ é a estimativa de probabilidade do modelo para classificar a instância $\\mathbf{x_i}$ como sendo da classe $1$ -- valor este gerado pela função sigmóide.\n",
        "\n"
      ],
      "metadata": {
        "id": "PFIuc4ZRJc8V"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1HZ7xvZqQ6Ju"
      },
      "source": [
        "model.compile(loss=\"binary_crossentropy\",optimizer=sgd,metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Por fim, vamos iniciar o processo de treinamento. Aqui vamos inserir a quantidade de épocas igual a $10$ -- é outro hiperparâmetro importante!"
      ],
      "metadata": {
        "id": "prEoRRYSKISb"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KzwJLL_6SE8K",
        "outputId": "b4cafdbd-4925-4f33-f875-30fb49d26942"
      },
      "source": [
        "history_fine = model.fit(X,labels,epochs=10)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1/1 [==============================] - 1s 1s/step - loss: 0.9489 - accuracy: 0.2500\n",
            "Epoch 2/10\n",
            "1/1 [==============================] - 0s 15ms/step - loss: 0.9450 - accuracy: 0.0000e+00\n",
            "Epoch 3/10\n",
            "1/1 [==============================] - 0s 11ms/step - loss: 0.9411 - accuracy: 0.0000e+00\n",
            "Epoch 4/10\n",
            "1/1 [==============================] - 0s 7ms/step - loss: 0.9373 - accuracy: 0.0000e+00\n",
            "Epoch 5/10\n",
            "1/1 [==============================] - 0s 7ms/step - loss: 0.9335 - accuracy: 0.0000e+00\n",
            "Epoch 6/10\n",
            "1/1 [==============================] - 0s 6ms/step - loss: 0.9297 - accuracy: 0.0000e+00\n",
            "Epoch 7/10\n",
            "1/1 [==============================] - 0s 6ms/step - loss: 0.9260 - accuracy: 0.0000e+00\n",
            "Epoch 8/10\n",
            "1/1 [==============================] - 0s 14ms/step - loss: 0.9223 - accuracy: 0.0000e+00\n",
            "Epoch 9/10\n",
            "1/1 [==============================] - 0s 7ms/step - loss: 0.9186 - accuracy: 0.0000e+00\n",
            "Epoch 10/10\n",
            "1/1 [==============================] - 0s 6ms/step - loss: 0.9150 - accuracy: 0.0000e+00\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "loss = history_fine.history['loss']\n",
        "\n",
        "plt.plot(history_fine.history['loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['training set'], loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "sx7rVwgpKwtn",
        "outputId": "67f2cebb-1220-42cc-9dff-3fdce44c8a47"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Observe que o erro da rede foi diminuindo com o passar das épocas, indicando que a rede gradativamente foi aprendendo os padrões nos dados e ganhando confiança na predição das instâncias de $X$. Entretanto, os valores da função de erro (*loss*) ainda estão altos, já que existem erros na classificação das instâncias de $X$ em relação aos $labels$. Mas veremos isso mais tarde...\n",
        "\n",
        "Agora podemos apenas classificar as instâncias utilizando um modelo recém-treinado. Vamos criar uma instância fictícia x' (```xl```), em que ```xl = [1,0]``` e vamos prever qual é sua classe.\n",
        "\n",
        "Para isso, utilizamos o método *predict* e passamos como argumento o conjunto de dados a ser predito. Como esse método obrigatoriamente recebe um conjunto de instâncias a serem classificadas, vamos criar uma variável associada a conjunto de dados ```XL``` que contenha apenas a instância ```xl```:"
      ],
      "metadata": {
        "id": "DWJ-O_ryKnWJ"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QszCknOASgNJ",
        "outputId": "724c7b9c-4704-4e0c-9512-3094bf18f740"
      },
      "source": [
        "xl = [1,0]\n",
        "\n",
        "XL = np.array([xl])\n",
        "\n",
        "XL"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[1, 0]])"
            ]
          },
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "e agora executamos o método *predict*:"
      ],
      "metadata": {
        "id": "rIAvGO9GOQzL"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y = model.predict(XL)\n",
        "y"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WwAtM9NeOQ6L",
        "outputId": "24bc3918-9110-4d18-ac97-bd504ccfd214"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0.429945]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Observe que a saída obtida se trata de uma probabilidade as saídas que obtivemos para cada instância de $X$. Agora vamos definitivamente decidir qual é a classe da instância ```xl``` de acordo com o procedimento abaixo:"
      ],
      "metadata": {
        "id": "y_TBR_gsMFRm"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UFJQtZofStuy",
        "outputId": "66a8f3a1-c1ef-4f97-a5f8-666a975c3663"
      },
      "source": [
        "for i in range(len(XL)):\n",
        "  if y[i] >= 0.5:\n",
        "    print(f\"Classe: 1\")\n",
        "  else:\n",
        "    print(f\"Classe: 0\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Classe: 0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "y_pred = np.argmax(y,axis=1)\n",
        "y_pred"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "prVT7PXYqAui",
        "outputId": "b295f1ad-bfb8-46da-82e0-d6a7606a3938"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0])"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "repare que o resultado \n",
        "\n",
        "**e se alterarmos a learning rate? Como será que o modelo vai se comportar?**\n",
        "\n",
        "Bom, vamos recomeçar todo o processo: recriamos o modelo de classificação e realizamos o treinamento mas com um valor de taxa de aprendizado maior:"
      ],
      "metadata": {
        "id": "cIiAYO6SMbR6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model = Sequential()\n",
        "model.add(Dense(number_of_neurons,input_dim=2, activation=\"tanh\"))\n",
        "model.add(Dense(1,activation=\"sigmoid\"))\n",
        "sgd = SGD(learning_rate=0.5)\n",
        "model.compile(loss=\"binary_crossentropy\",optimizer=sgd,metrics=['accuracy'])\n",
        "history_fine = model.fit(X,labels,epochs=10)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wVURZnbmRQbp",
        "outputId": "2c541206-63f1-41f3-d74e-f1a080155b7f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1/1 [==============================] - 1s 734ms/step - loss: 0.6705 - accuracy: 0.7500\n",
            "Epoch 2/10\n",
            "1/1 [==============================] - 0s 20ms/step - loss: 0.6099 - accuracy: 0.7500\n",
            "Epoch 3/10\n",
            "1/1 [==============================] - 0s 20ms/step - loss: 0.5652 - accuracy: 0.7500\n",
            "Epoch 4/10\n",
            "1/1 [==============================] - 0s 22ms/step - loss: 0.5314 - accuracy: 0.7500\n",
            "Epoch 5/10\n",
            "1/1 [==============================] - 0s 15ms/step - loss: 0.5051 - accuracy: 0.7500\n",
            "Epoch 6/10\n",
            "1/1 [==============================] - 0s 24ms/step - loss: 0.4839 - accuracy: 0.7500\n",
            "Epoch 7/10\n",
            "1/1 [==============================] - 0s 22ms/step - loss: 0.4660 - accuracy: 0.7500\n",
            "Epoch 8/10\n",
            "1/1 [==============================] - 0s 31ms/step - loss: 0.4502 - accuracy: 0.7500\n",
            "Epoch 9/10\n",
            "1/1 [==============================] - 0s 11ms/step - loss: 0.4355 - accuracy: 0.7500\n",
            "Epoch 10/10\n",
            "1/1 [==============================] - 0s 22ms/step - loss: 0.4215 - accuracy: 0.7500\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "loss = history_fine.history['loss']\n",
        "\n",
        "plt.plot(history_fine.history['loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['training set'], loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "MQcZ4f8PS6zK",
        "outputId": "5e56a1c4-5c12-4076-84e1-8a924f4b5f66"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3dd3yV9f3+8dc7i5AQVsJMwpAhewYFceBAUQSpKGpdtFWrVq2t1eq3rbbWX2trW22tG21dVRFQQVEUBSvKCnvLkJEQ9oaQ+f79cY40YsAASe7k5Ho+HnmYc49zrpyWXLnvz7k/t7k7IiIih4sKOoCIiFRNKggRESmVCkJEREqlghARkVKpIEREpFQqCBERKZUKQqQcmNm/zeyhMm671szOO9HnEaloKggRESmVCkJEREqlgpAaI3xq524zW2hm+83seTNrYmbvm9leM5tsZg1KbD/UzJaY2S4zm2pmHUus62lmc8P7vQHEH/ZaF5vZ/PC+X5hZt+PMfKOZrTKzHWY23syah5ebmT1qZlvMbI+ZLTKzLuF1F5nZ0nC2bDP7xXG9YVLjqSCkphkODATaA0OA94H/AxoR+vdwB4CZtQdeA+4Mr5sITDCzODOLA94GXgYaAm+Gn5fwvj2BF4AfA8nAM8B4M6t1LEHN7Bzgj8AIoBmwDng9vPp84Mzwz1EvvM328LrngR+7exLQBfjkWF5X5GsqCKlpHnf3ze6eDXwGzHT3ee5+EHgL6Bne7grgPXf/yN0LgL8AtYHTgL5ALPCYuxe4+xhgdonXuAl4xt1nunuRu78I5IX3OxZXAy+4+1x3zwPuA/qZWSugAEgCOgDm7svcPSe8XwHQyczquvtOd597jK8rAqggpObZXOL73FIe1wl/35zQX+wAuHsxsAFIDa/L9m/OdLmuxPctgbvCp5d2mdkuID2837E4PMM+QkcJqe7+CfBP4Algi5k9a2Z1w5sOBy4C1pnZp2bW7xhfVwRQQYgcyUZCv+iB0Dl/Qr/ks4EcIDW87GstSny/Afh/7l6/xFeCu792ghkSCZ2yygZw93+4e2+gE6FTTXeHl89290uAxoROhY0+xtcVAVQQIkcyGhhsZueaWSxwF6HTRF8A04FC4A4zizWzS4FTSuz7HHCzmZ0aHkxONLPBZpZ0jBleA35gZj3C4xd/IHRKbK2Z9Qk/fyywHzgIFIfHSK42s3rhU2N7gOITeB+kBlNBiJTC3VcA1wCPA9sIDWgPcfd8d88HLgVGAjsIjVeMK7FvJnAjoVNAO4FV4W2PNcNk4DfAWEJHLW2AK8Or6xIqop2ETkNtBx4Jr7sWWGtme4CbCY1liBwz0w2DRESkNDqCEBGRUqkgRESkVCoIEREplQpCRERKFRN0gPKSkpLirVq1CjqGiEi1MmfOnG3u3qi0dRFTEK1atSIzMzPoGCIi1YqZrTvSOp1iEhGRUqkgRESkVCoIEREpVcSMQZSmoKCArKwsDh48GHSUGiM+Pp60tDRiY2ODjiIiJyiiCyIrK4ukpCRatWrFNyfelIrg7mzfvp2srCxat24ddBwROUERfYrp4MGDJCcnqxwqiZmRnJysIzaRCBHRBQGoHCqZ3m+RyBHxBfFdioudnN255BcWBR1FRKRKqfEFUVhczI59+WzYkUt5T32+a9cunnzyyePa96KLLmLXrl1H3eb+++9n8uTJx/X8J+Ltt99m6dKllf66IlK5anxBxMVE07x+bfbnF7J1X165PvfRCqKwsPCo+06cOJH69esfdZsHH3yQ884777jzHS8VhEjNUOMLAqB+Qiz1aseyeU8euflH/8V9LO69915Wr15Njx49uPvuu5k6dSpnnHEGQ4cOpVOnTgAMGzaM3r1707lzZ5599tlD+7Zq1Ypt27axdu1aOnbsyI033kjnzp05//zzyc3NBWDkyJGMGTPm0PYPPPAAvXr1omvXrixfvhyArVu3MnDgQDp37swNN9xAy5Yt2bZt2zdyFhUVMXLkSLp06ULXrl159NFHAVi9ejWDBg2id+/enHHGGSxfvpwvvviC8ePHc/fdd9OjRw9Wr15dbu+XiFQtEf0x15J+N2EJSzfuOeJ6B3LzizCgdlx0mZ6zU/O6PDCk8xHXP/zwwyxevJj58+cDMHXqVObOncvixYsPfQz0hRdeoGHDhuTm5tKnTx+GDx9OcnLyN55n5cqVvPbaazz33HOMGDGCsWPHcs0113zr9VJSUpg7dy5PPvkkf/nLXxg1ahS/+93vOOecc7jvvvv44IMPeP7557+13/z588nOzmbx4sUAh05t3XTTTTz99NO0a9eOmTNncuutt/LJJ58wdOhQLr74Yi677LIyvU8iUj3VmIL4LgbUioniYEER+YXFxMVUzMHVKaec8o1rBP7xj3/w1ltvAbBhwwZWrlz5rYJo3bo1PXr0AKB3796sXbu21Oe+9NJLD20zblzoFsnTpk079PyDBg2iQYMG39rvpJNOYs2aNdx+++0MHjyY888/n3379vHFF19w+eWXH9ouL698T8GJSNVWYwriaH/pl7RxVy7b9uXRKiWRuvHlfzVwYmLioe+nTp3K5MmTmT59OgkJCQwYMKDUawhq1ap16Pvo6OhDp5iOtF10dPR3jnGU1KBBAxYsWMCkSZN4+umnGT16NI899hj169c/dPQjIjWPxiAO07RuPPGx0WTtzKWwqPiEnispKYm9e/cecf3u3btp0KABCQkJLF++nBkzZpzQ65Wmf//+jB49GoAPP/yQnTt3fmubbdu2UVxczPDhw3nooYeYO3cudevWpXXr1rz55ptA6CrpBQsWlOnnEpHIoII4TFSUkd4ggaJiJ2vniX30NTk5mf79+9OlSxfuvvvub60fNGgQhYWFdOzYkXvvvZe+ffueSPRSPfDAA3z44Yd06dKFN998k6ZNm5KUlPSNbbKzsxkwYAA9evTgmmuu4Y9//CMAr776Ks8//zzdu3enc+fOvPPOOwBceeWVPPLII/Ts2VOD1CIRzMr7s/9BycjI8MNvGLRs2TI6dux4XM+3dW8eObtzSWtQm4aJtb57hyoqLy+P6OhoYmJimD59OrfcckuFnzY6kfddRCqXmc1x94zS1tWYMYhjlVInjr0HC9i46yCJcTHUii3bJ5uqmvXr1zNixAiKi4uJi4vjueeeCzqSiFQTKogjMDPSGiSwcsteNuzMpU2jxGo5z1C7du2YN29e0DFEpBqK+DGIEzmFFhcTRWr92hzIL2TLXn3Esywi5ZSliER4QcTHx7N9+/YT+qVVPyGOBglxbNmTx4G88rvKOhJ9fT+I+Pj4oKOISDmo0FNMZjYI+DsQDYxy94dL2WYE8FtCFzMvcPfvh5cXAYvCm61396HH+vppaWlkZWWxdevW4/wJQord2bYnj21Z0LhuLaKq4ammyvL1HeVEpPqrsIIws2jgCWAgkAXMNrPx7r60xDbtgPuA/u6+08wal3iKXHfvcSIZYmNjy+3OZvvWbOfK52ZwZZ90/nhpt3J5ThGRqqwiTzGdAqxy9zXung+8Dlxy2DY3Ak+4+04Ad99SgXlOyKknJXPzWW14bdYGPlyyKeg4IiIVriILIhXYUOJxVnhZSe2B9mb2uZnNCJ+S+lq8mWWGlw+rwJxl9rPz2tO5eV3uHbeILXt1W00RiWxBD1LHAO2AAcBVwHNm9vVNEFqGL974PvCYmbU5fGczuylcIpknOs5QFnExUfz9yh7szyvknjEL9YkdEYloFVkQ2UB6icdp4WUlZQHj3b3A3b8CviRUGLh7dvi/a4CpQM/DX8Ddn3X3DHfPaNSoUfn/BKVo2ziJXw3uyNQVW3llxrpKeU0RkSBUZEHMBtqZWWsziwOuBMYfts3bhI4eMLMUQqec1phZAzOrVWJ5f6DK3MLs2r4tOat9Ix56bxmrtmjSOhGJTBVWEO5eCNwGTAKWAaPdfYmZPWhmX39kdRKw3cyWAlOAu919O9ARyDSzBeHlD5f89FPQzIxHLu9GYq0Y7nxjPvmFJzbrq4hIVRTRk/VVtElLNvHjl+dw64A23DOoQ6W+tohIeTjaZH1BD1JXaxd0bsoVGek89elqZn21I+g4IiLlSgVxgu4f0okWDRP42Rvz2XOwIOg4IiLlRgVxghJrxfDoFT3YtOcgv31nSdBxRETKjQqiHPRq0YDbzm7LuHnZvLtwY9BxRETKhQqinNx2Tlt6pNfnV28tJmd3btBxREROmAqinMRGR/HoFT0oKCrmF28uoLg4Mj4dJiI1lwqiHLVOSeT+izvx+artvPD5V0HHERE5ISqIcnZFn3QGdmrCnz9YwfJNe4KOIyJy3FQQ5czMePjSrtStHcudr8/nYEFR0JFERI6LCqICJNepxSOXd2P5pr38ZdKKoOOIiBwXFUQFOfvkxlzXryWjpn3F56u2BR1HROSYqSAq0H0XdqRNo0TuGr2AXQfyg44jInJMVBAVqHZcNH+/sifb9uXxq7cW6wZDIlKtqCAqWJfUevxsYHveW5TDW/MOv1+SiEjVpYKoBDef1YY+rRpw/ztL2LDjQNBxRETKRAVRCaKjjL+N6AHAXaMXUKSrrEWkGlBBVJL0hgk8eElnZq3dwTP/XR10HBGR76SCqETf65nK4G7N+NuHX7I4e3fQcUREjkoFUYnMjP83rAspdWrx09fnkZuvq6xFpOpSQVSy+glx/HVEd1Zv3c8f318WdBwRkSNSQQSgf9sUbji9NS9NX8eU5VuCjiMiUioVREB+ccHJdGiaxN1jFrJ9X17QcUREvkUFEZD42Ggeu7IHe3ILuHfcIl1lLSJVjgoiQB2a1uWeQSfz0dLNvDF7Q9BxRES+QQURsB/2b03/tsn8bsJSvtq2P+g4IiKHqCACFhVl/OXy7sTFRHHnG/MpKCoOOpKICKCCqBKa1avNH77XlQUbdvGLNzUVh4hUDTFBB5CQwd2asX5HB/70wXJqx0bzx0u7YmZBxxKRGkwFUYXcMqANB/ILefyTVdSOi+b+izupJEQkMCqIKubnA9uzP6+IFz7/isS4GH5xwclBRxKRGkoFUcWYGb+5uCO5BYX8c0roSOInZ7cNOpaI1EAVOkhtZoPMbIWZrTKze4+wzQgzW2pmS8zsPyWWX29mK8Nf11dkzqrGzHhoWFeG9WjOI5NW8O/Pvwo6kojUQBV2BGFm0cATwEAgC5htZuPdfWmJbdoB9wH93X2nmTUOL28IPABkAA7MCe+7s6LyVjXR4Y+/5hYU8dsJS0mIi2FEn/SgY4lIDVKRRxCnAKvcfY275wOvA5ccts2NwBNf/+J3969nrrsA+Mjdd4TXfQQMqsCsVVJMdBT/uKonZ7ZvxC/HLWT8go1BRxKRGqQiCyIVKDl/RFZ4WUntgfZm9rmZzTCzQcewL2Z2k5llmlnm1q1byzF61VErJppnrulNn1YN+fkb8/lo6eagI4lIDRH0hXIxQDtgAHAV8JyZ1S/rzu7+rLtnuHtGo0aNKihi8GrHRfPCyD50Tq3HT16dy2crI7MMRaRqqciCyAZKnjRPCy8rKQsY7+4F7v4V8CWhwijLvjVKnVoxvPiDPpzUKJEbX8pk1lc7go4kIhGuIgtiNtDOzFqbWRxwJTD+sG3eJnT0gJmlEDrltAaYBJxvZg3MrAFwfnhZjVY/IY6Xf3QqzevX5of/ns3CrF1BRxKRCFZhBeHuhcBthH6xLwNGu/sSM3vQzIaGN5sEbDezpcAU4G533+7uO4DfEyqZ2cCD4WU1XqOkWrx6w6k0SIzluhdmsXzTnqAjiUiEski5UU1GRoZnZmYGHaPSbNhxgMufnk5hsTP6x305qVGdoCOJSDVkZnPcPaO0dUEPUstxSm+YwCs3nIq7c/WomWzYcSDoSCISYVQQ1VjbxnV4+Uensj+vkKtHzWTznoNBRxKRCKKCqOY6Na/Liz88he378rh61Ey278sLOpKIRAgVRATo2aIBz4/sw4YdB7j2+Vnszi0IOpKIRAAVRIToe1Iyz1zbm5Vb9jLyX7PYn1cYdCQRqeZUEBFkwMmNefyqXizM2s0NL2ZysKAo6EgiUo2pICLMoC5N+evl3Znx1XZueWUO+YXFQUcSkWpKBRGBhvVM5Q/f68qUFVv56evzKCxSSYjIsVNBRKirTmnBby7uxPuLN3HPmIUUF0fGBZEiUnl0y9EI9qPTW3Mgr5C/fvQlteOieWhYF8ws6FgiUk2oICLcbee0ZX9+EU9/upqEuGj+76KOKgkRKRMVRIQzM3456GQO5Bfy3GdfkVgrhjvPax90LBGpBlQQNYCZ8dshnTmQX8Rjk1eSEBfNTWe2CTqWiFRxKogaIirK+NPwbuQWFPGHicupHRfDtX1bBh1LRKowFUQNEh1lPHZFD/IKivjN24upHRvNZb3Tgo4lIlWUPuZaw8RGR/HP7/eif9tk7hmzgImLcoKOJCJVlAqiBoqPjea56zLo1aIBd7w2j0+Wbw46kohUQSqIGiohLoYXftCHDs2SuPmVuXyxalvQkUSkilFB1GB142N56Yen0io5gRteymTOup1BRxKRKkQFUcM1TIzjlR+dSuOkWoz81ywWZ+8OOpKIVBEqCKFx3XhevbEvdeNjue6FWazcvDfoSCJSBaggBIDU+rV59YZTiY4yrnpuJnPX63STSE2ngpBDWqUk8tqNp1I7LoornpnOf2auDzqSiARIBSHf0LZxEhNuO51+bVL4v7cWce/YheQV6s50IjWRCkK+pX5CHP8a2YefnN2G12dvYMQzM8jZnRt0LBGpZCoIKVV0lHH3BR14+prerNq8lyGPT2PGmu1BxxKRSqSCkKMa1KUp79zWn7q1Y7l61ExemPYV7ro7nUhNoIKQ79S2cRLv/KQ/53RozIPvLuVnb8wnN1/jEiKRTgUhZZIUH8sz1/TmroHteWfBRoY/9QUbdhwIOpaIVCAVhJRZVJRx+7nteOH6PmTtPMDFj0/jv19uDTqWiFQQFYQcs7M7NGb8bafTrF481/9rFk9MWaVxCZEIVKaCMLOfmlldC3nezOaa2fll2G+Qma0ws1Vmdm8p60ea2VYzmx/+uqHEuqISy8cf248lFa1VSiLjbj2NwV2b8cikFdzyylz25RUGHUtEylFZjyB+6O57gPOBBsC1wMNH28HMooEngAuBTsBVZtaplE3fcPce4a9RJZbnllg+tIw5pRIlxMXw+FU9+dVFHflw6SaGPfE5q7fuCzqWiJSTshaEhf97EfCyuy8psexITgFWufsad88HXgcuOb6YUlWZGTeeeRKv/OhUduzPZ9g/P+ejpboBkUgkKGtBzDGzDwkVxCQzSwKKv2OfVGBDicdZ4WWHG25mC81sjJmll1geb2aZZjbDzIaV9gJmdlN4m8ytWzVYGqTT2qYw4fbTaZWSyI0vZfK3D1dQXKxxCZHqrKwF8SPgXqCPux8AYoEflMPrTwBauXs34CPgxRLrWrp7BvB94DEza3P4zu7+rLtnuHtGo0aNyiGOnIjU+rV58+Z+XNY7jX98soofvTib3bkFQccSkeNU1oLoB6xw911mdg3wa+C77iyTDZQ8IkgLLzvE3be7e1744Sigd4l12eH/rgGmAj3LmFUCFB8bzSOXdeP3w7owbdU2hv5zGis26f4SItVRWQviKeCAmXUH7gJWAy99xz6zgXZm1trM4oArgW98GsnMmpV4OBRYFl7ewMxqhb9PAfoDS8uYVQJmZlzbtyWv39SXA/lFDHvic95duDHoWCJyjMpaEIUe+qD7JcA/3f0JIOloO7h7IXAbMInQL/7R7r7EzB40s68/lXSHmS0xswXAHcDI8PKOQGZ4+RTgYXdXQVQzvVs25L3bT6dT87rc9p95/GHiMgqLvmvoSkSqCivLBU5m9inwAfBD4AxgC7DA3btWbLyyy8jI8MzMzKBjSCnyC4v5/btLeXnGOvq3Tebxq3rRMDEu6FgiApjZnPB477eU9QjiCiCP0PUQmwiNJzxSTvkkwsXFRPH7YV145LJuzF67kyGPT2NR1ncNYYlI0MpUEOFSeBWoZ2YXAwfd/bvGIES+4fKMdMbc3A93Z/jTXzBmTlbQkUTkKMo61cYIYBZwOTACmGlml1VkMIlM3dLqM+H20+ndogG/eHMB97+zmPxCjUuIVEUxZdzuV4SugdgCYGaNgMnAmIoKJpEruU4tXv7RKfzpg+U899lXLN24hyev7kXjuvFBRxOREso6BhH1dTmEbT+GfUW+JSY6il8N7sQ/rurJko17uPjxacxZtyPoWCJSQll/yX9gZpPCs6+OBN4DJlZcLKkphnZvzrhbTyM+Nporn53ByzPWaepwkSqirIPUdwPPAt3CX8+6+y8rMpjUHB2b1WXCbafTv20Kv3l7MfeMWcjBAt3SVCRoZboOojrQdRDVX1Gx89jkL3n8k1V0S6vH09f0pnn92kHHEolox30dhJntNbM9pXztNbM9FRNXaqroKOOu80/mmWt7s2brfgY99l9enblOs8KKBOSoBeHuSe5et5SvJHevW1khpWa5oHNTJoSn6PjVW4sZ/vQXLN2ov0dEKps+iSRVUuuURF67sS9/G9Gd9dsPMOSf03jo3aXs121NRSqNCkKqLDPj0l5pfHzXWYzISGfUtK8472+f8sHiTfqkk0glUEFIlVc/IY4/XtqVsbecRr3asdz8yhxueDGTDTsOBB1NJKKpIKTa6N2yAe/efjq/HtyR6Wu2M/DRT3lq6moKNIW4SIVQQUi1EhMdxQ1nnMTkn5/Fme0a8acPljP4H58xe62uwhYpbyoIqZaa16/Ns9dl8Nx1GezPK+Lyp6dzz5gF7NifH3Q0kYihgpBqbWCnJnz08zO5+aw2jJubzbl/ncrozA0axBYpByoIqfYS4mK498IOvHfHGbRpVId7xizkimdm8OXmvUFHE6nWVBASMU5umsToH/fjz8O78eWWvVz098/48wfLyc3XvE4ix0MFIRElKsoY0SedT+4awLCeqTw5dTUDH/2UT5ZvDjqaSLWjgpCI1DAxjr9c3p3Xb+pLfGw0P/x3Jje/PIec3blBRxOpNlQQEtH6npTMxDvO4J5BJzP1yy2c99dPGfXZGgp17YTId1JBSMSLi4ni1gFt+ehnZ3FK64Y89N4yhv7zc+at3xl0NJEqTQUhNUZ6wwReGNmHp67uxfb9eVz61Bf8+u1F7M4tCDqaSJWkgpAaxcy4sGszPr5rAD84rTX/mbmec/86lbfnZevaCZHDqCCkRqpTK4b7h3Ri/G2nk9oggTvfmM81z89kzdZ9QUcTqTJUEFKjdUmtx7hbTuOhYV1YmLWbQY99xqMffal7YougghAhOsq4pm9LPr7rLC7q2pS/f7ySQY/9l89Wbg06mkigVBAiYY2T4nnsyp68esOpmBnXPj+L21+bx+Y9B4OOJhIIFYTIYfq3TeH9n57Bnee1Y9LiTZz55yk8OGEpW1QUUsNUaEGY2SAzW2Fmq8zs3lLWjzSzrWY2P/x1Q4l115vZyvDX9RWZU+Rw8bHR3Hleeyb//CyGdm/Oi9PXcsafp/C7CUtUFFJjWEV9tM/MooEvgYFAFjAbuMrdl5bYZiSQ4e63HbZvQyATyAAcmAP0dvcjXtmUkZHhmZmZ5f1jiACwbvt+npiyirFzs4mJMq46pQW3DGhDk7rxQUcTOSFmNsfdM0pbV5FHEKcAq9x9jbvnA68Dl5Rx3wuAj9x9R7gUPgIGVVBOke/UMjmRP1/WnSl3DeCSHs15ecY6zvjzFH47fonGKCRiVWRBpAIbSjzOCi873HAzW2hmY8ws/Rj3FalULZITDhXF93qk8kqJoti0W0UhkSXoQeoJQCt370boKOHFY9nZzG4ys0wzy9y6VR9JlMrTIjmBP13WjSm/GMClPUNFceYjU3jgncUqCokYFVkQ2UB6icdp4WWHuPt2d88LPxwF9C7rvuH9n3X3DHfPaNSoUbkFFymr9IYJPDw8VBTDe6Xy6sz1nPnnKdz/zmJNLS7VXkUOUscQGqQ+l9Av99nA9919SYltmrl7Tvj77wG/dPe+4UHqOUCv8KZzCQ1S7zjS62mQWqqCDTsO8OTU1byZuYEoM67ok84tA9rQvH7toKOJlOpog9QxFfWi7l5oZrcBk4Bo4AV3X2JmDwKZ7j4euMPMhgKFwA5gZHjfHWb2e0KlAvDg0cpBpKpIb5jAHy/tyq0D2vDk1NW8Pns9b8zewIg+adw6oK2KQqqVCjuCqGw6gpCqKGvn/44oAK7ok66ikCrlaEcQKgiRSpC9K5cnp6xidLgoRmSkc+vZbUlVUUjAVBAiVUT2rlyemrqKN2aHiuLyjHRuHdCGtAYJASeTmkoFIVLFbNyVy1NTV/PG7A04zmW90/nJ2SoKqXwqCJEq6ttFERrMTm+oopDKoYIQqeJydoeK4vVZGyj2UFH85GwVhVQ8FYRINZGzO5enp67mNRWFVBIVhEg1s2n3QZ7+dDX/mbWe4mJneK80bjrrJNo0qhN0NIkwKgiRamrznoM8NTVUFPmFxZzRLoXr+rXinA6NiY6yoONJBFBBiFRzW/fm8cbs9bwyYz2b9hwktX5trunbkiv6pNMwMS7oeFKNqSBEIkRhUTEfLd3MS9PXMX3NduJiohjSrTnXn9aSbmn1g44n1ZAKQiQCfbl5Ly9NX8u4udkcyC+iR3p9rj+tJRd1bUatmOig40k1oYIQiWB7DhYwbk4WL81Yx5qt+0lOjOPKU9K5+tSWmvNJvpMKQqQGcHc+X7WdF6ev5eNlmwEY2KkJ1/drRb82yZhpUFu+LZDpvkWkcpkZp7dL4fR2KWTtPMCrM9fz+qz1TFqymbaN63Bdv5Zc2iuNOrX0z17KRkcQIhHsYEER7y3M4aXpa1mQtZs6tWK4tFcq1/VrSdvGSUHHkypAp5hEhPkbdvHS9LW8uyCH/KJiTmuTzHX9WnFex8bERAd9e3oJigpCRA7Zvi+PNzI38OqM9WTvyqV5vXiuDl9TkVKnVtDxpJKpIETkWwqLivl4+RZenr6Oaau2ERcdxeBuzbiuX0t6pNfXoHYNoUFqEfmWmOgoLujclAs6N2XVlr28PH0dY+dm89a8bLql1eO6fq24uFsz4mN1TUVNpSMIEfDtnEEAAAwgSURBVDlkX14hb83N4sXp61i1ZR8NEmK5ok8Lrj61hWaUjVA6xSQix8Tdmb5mOy99sY6Plm3G3TmnQxOu69eS09umEKWJAiOGTjGJyDExM05rk8JpbVLYuCuX/8xcz2uz1jN52WZS69fm0l6pDO+VRquUxKCjSgXSEYSIlEleYREfLN7E2LnZTFu5lWKH3i0bMLxXGoO7NaNe7digI8px0CkmESlXm3Yf5O352YyZk8WqLfuoFRPF+Z2bMrxXKme0a6R7VVQjKggRqRDuzsKs3Yydm8X4BRvZdaCAxkm1+F7PVIb3TqN9E12tXdWpIESkwuUVFjFl+RbGzMlm6ootFBY7XVPrMbxXKkN7pOrGRlWUCkJEKtW2fXm8M38jY+dksTRnD7HRxjkdGjO8VxoDTm5MXIym9qgqVBAiEphlOXsYOyeLt+dvZNu+PBomxjG0e3Mu651G5+Z1dcV2wFQQIhK4wqJi/rtyK2PnZPPR0s3kFxXToWkSw3ulcUnP5jROig86Yo2kghCRKmXXgXwmLMxh7Jws5m/YRXSUcWa7FIb3TuO8jk00vUclUkGISJW1ass+xs3N4q152eTsPkjd+Bgu7t6c4b3S6NVCkwZWtMAKwswGAX8HooFR7v7wEbYbDowB+rh7ppm1ApYBK8KbzHD3m4/2WioIkeqtqNj5YvU2xs7J4oMlmzhYUMxJKYkM753G93qm6v7aFSSQgjCzaOBLYCCQBcwGrnL3pYdtlwS8B8QBt5UoiHfdvUtZX08FIRI59h4s4P1FmxgzN4tZX+3ADE5rk8zwXmkM6tKUhDjNElRegpqL6RRglbuvCYd4HbgEWHrYdr8H/gTcXYFZRKQaSYqPZUSfdEb0SWf99gOMm5fF2LlZ/Hz0An7z9mIu7NqMod2bc1qbZN0NrwJVZEGkAhtKPM4CTi25gZn1AtLd/T0zO7wgWpvZPGAP8Gt3/+zwFzCzm4CbAFq0aFGe2UWkimiRnMCd57XnjnPaMXvtDsbOzWLiok2MmZNFcmIcF3ZtytDuqWS0bKBZZstZYMdpZhYF/A0YWcrqHKCFu283s97A22bW2d33lNzI3Z8FnoXQKaYKjiwiAYqKMk49KZlTT0rmwUu6MHXFViYs3MiYOVm8MmM9TevGc3G3Zgzp3pxuafU0uF0OKrIgsoH0Eo/Twsu+lgR0AaaG/4dsCow3s6HungnkAbj7HDNbDbQHNMggIsTHRjOoS1MGdWnK/rxCJi/bzIQFObw4fS2jpn1Fi4YJDOnejIu7NadD0ySVxXGqyEHqGEKD1OcSKobZwPfdfckRtp8K/CI8SN0I2OHuRWZ2EvAZ0NXddxzp9TRILSK7DxQwaekmJizYyBert1NU7LRtXIch3ZozpHszTmpUJ+iIVU4gg9TuXmhmtwGTCH3M9QV3X2JmDwKZ7j7+KLufCTxoZgVAMXDz0cpBRASgXkIsIzLSGZGRzrZ9eby/OFQWj338JY9O/pLOzesypHtzLu7WjLQGuoXqd9GFciIS8XJ25/LewhwmLMxhwYZdAPRqUZ8h3ZszuGszGtetudN86EpqEZGw9dsP8O6ijUxYkMOynD2YQd/WyQzp3pwLuzSlQQ2bllwFISJSilVb9jJhQQ4TFmxkzbb9xEQZp7dLYUi35gzs3IS68ZF/G1UVhIjIUbg7S3P2HCqL7F25xMVEcfbJjRjSvTnndmhC7bjInEBQBSEiUkbuzrwNu5iwYCPvLcxhy948EuKiOa9jE4Z0b86Z7VOoFRM5ZaGCEBE5DkXFzqyvdjBh4UbeX5TDzgMFJMXHMKhzU4ZEyFQfKggRkRNUUFTM56u2MWFBDh8u2cTevEIaJsZxQecmXNilGf3aJBNbDctCBSEiUo4OFhQxdcVWJi7K4eNlm9mfX0T9hFjO79SEC7s2o3+blGpz320VhIhIBTlYUMR/vwyVxeRlW9iXV0jd+BgGdmrK4G5N6d+2ao9ZqCBERCpBXmER01Zu471FOXy0dDN7DxaSVCuG8zo14aKuzTijXUqVu51qUPeDEBGpUWrFRHNuxyac27EJ+YWhMYuJi3L4cOlm3pqXTZ1aMZzbsTEXdmnGgJMbVbmyOJyOIEREKlhBUTFfrN7O+4tymLRkEzsPFJAQF805HRpzUddmnH1y48Cus9ApJhGRKqKgqJiZa3YwcXEOkxZvYvv+fGrHRnN2h0Zc2KUZ53RoTGKtyju5o4IQEamCCouKmbV2BxMX5fDB4s1s25dHrZgoBpzciIu6NuPcjk2oU8FloYIQEaniioqd2Wt38P6iHN5fvIkte/OIi4nizHaNGNytKed2rJi5oVQQIiLVSHGxM2f9TiYuyuH9RZvYtOcgcdFRnN4uhYu6NmNgxybUSyifslBBiIhUU8XFobmhvj6yyN6VS2y00b9tChd1acbATk1OaIpyFYSISARwdxZk7WbiohwmLsoha2cuMVHGBV2a8sT3ex3Xc+o6CBGRCGBm9EivT4/0+tx3YQcWZ+/hvUU5VNQUUCoIEZFqyMzomlaPrmn1Kuw1qsdsUiIiUulUECIiUioVhIiIlEoFISIipVJBiIhIqVQQIiJSKhWEiIiUSgUhIiKlipipNsxsK7DuBJ4iBdhWTnGqO70X36T345v0fvxPJLwXLd29UWkrIqYgTpSZZR5pPpKaRu/FN+n9+Ca9H/8T6e+FTjGJiEipVBAiIlIqFcT/PBt0gCpE78U36f34Jr0f/xPR74XGIEREpFQ6ghARkVKpIEREpFQ1viDMbJCZrTCzVWZ2b9B5gmRm6WY2xcyWmtkSM/tp0JmCZmbRZjbPzN4NOkvQzKy+mY0xs+VmtszM+gWdKUhm9rPwv5PFZvaamcUHnam81eiCMLNo4AngQqATcJWZdQo2VaAKgbvcvRPQF/hJDX8/AH4KLAs6RBXxd+ADd+8AdKcGvy9mlgrcAWS4excgGrgy2FTlr0YXBHAKsMrd17h7PvA6cEnAmQLj7jnuPjf8/V5CvwBSg00VHDNLAwYDo4LOEjQzqwecCTwP4O757r4r2FSBiwFqm1kMkABsDDhPuavpBZEKbCjxOIsa/AuxJDNrBfQEZgabJFCPAfcAxUEHqQJaA1uBf4VPuY0ys8SgQwXF3bOBvwDrgRxgt7t/GGyq8lfTC0JKYWZ1gLHAne6+J+g8QTCzi4Et7j4n6CxVRAzQC3jK3XsC+4EaO2ZnZg0InW1oDTQHEs3smmBTlb+aXhDZQHqJx2nhZTWWmcUSKodX3X1c0HkC1B8YamZrCZ16PMfMXgk2UqCygCx3//qIcgyhwqipzgO+cvet7l4AjANOCzhTuavpBTEbaGdmrc0sjtAg0/iAMwXGzIzQOeZl7v63oPMEyd3vc/c0d29F6P8Xn7h7xP2FWFbuvgnYYGYnhxedCywNMFLQ1gN9zSwh/O/mXCJw0D4m6ABBcvdCM7sNmEToUwgvuPuSgGMFqT9wLbDIzOaHl/2fu08MMJNUHbcDr4b/mFoD/CDgPIFx95lmNgaYS+jTf/OIwGk3NNWGiIiUqqafYhIRkSNQQYiISKlUECIiUioVhIiIlEoFISIipVJBiFQBZjZAM8ZKVaOCEBGRUqkgRI6BmV1jZrPMbL6ZPRO+X8Q+M3s0fG+Aj82sUXjbHmY2w8wWmtlb4fl7MLO2ZjbZzBaY2VwzaxN++jol7rfwavgKXZHAqCBEysjMOgJXAP3dvQdQBFwNJAKZ7t4Z+BR4ILzLS8Av3b0bsKjE8leBJ9y9O6H5e3LCy3sCdxK6N8lJhK5sFwlMjZ5qQ+QYnQv0BmaH/7ivDWwhNB34G+FtXgHGhe+fUN/dPw0vfxF408ySgFR3fwvA3Q8ChJ9vlrtnhR/PB1oB0yr+xxIpnQpCpOwMeNHd7/vGQrPfHLbd8c5fk1fi+yL071MCplNMImX3MXCZmTUGMLOGZtaS0L+jy8LbfB+Y5u67gZ1mdkZ4+bXAp+E79WWZ2bDwc9Qys4RK/SlEykh/oYiUkbsvNbNfAx+aWRRQAPyE0M1zTgmv20JonALgeuDpcAGUnP30WuAZM3sw/ByXV+KPIVJmms1V5ASZ2T53rxN0DpHyplNMIiJSKh1BiIhIqXQEISIipVJBiIhIqVQQIiJSKhWEiIiUSgUhIiKl+v/HyL/E2LE+3QAAAABJRU5ErkJggg==\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Hummmm... o modelo está um pouco melhor (mais confiante) após apenas ajustarmos o valor da taxa de aprendizado. Exercício: otimize os hiperparâmetros desse modelo para melhorar ainda mais esse modelo, fazendo-o acertar todas as predições durante o treinamento nas últimas épocas."
      ],
      "metadata": {
        "id": "MIIDOY8nS8mK"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Salvando os pesos do seu modelo"
      ],
      "metadata": {
        "id": "PZyJCmmmqTbC"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SYVCPd6AUEzV"
      },
      "source": [
        "import h5py\n",
        "model_structure = model.to_json()\n",
        "\n",
        "with open(\"arquitetura_1_mlp.json\",'w') as json_file:\n",
        "  json_file.write(model_structure)\n",
        "\n",
        "model.save_weights(\"pesos_1_modelo.json\")"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}