denoising_autoencoder_with_dropout.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/ali3brt/20-Newsgroups-Dataset/blob/main/denoising_autoencoder_with_dropout.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JafadovRyLhH"
      },
      "source": [
        "# Denoising Autoencoders"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xexL3mTzhkgj"
      },
      "source": [
        "This is based on [Geron's notebook, chapter 17](https://github.com/ageron/handson-ml3/blob/main/17_autoencoders_gans_and_diffusion_models.ipynb), in accordance with Apache License v2"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "1pxeWHUAhkgj"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "from scipy.spatial.transform import Rotation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N_7X_4h-hkgk"
      },
      "source": [
        "Let's load the fashion MNIST dataset, scale it, and split it into a training set, a validation set, and a test set:"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fVDOS30khkgk",
        "outputId": "719b67c2-ce09-4979-e0b3-a4fdb147a058"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "\u001b[1m29515/29515\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
            "\u001b[1m26421880/26421880\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "\u001b[1m5148/5148\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 1us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
            "\u001b[1m4422102/4422102\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 0us/step\n"
          ]
        }
      ],
      "source": [
        "# extra code – loads, scales, and splits the fashion MNIST dataset\n",
        "fashion_mnist = tf.keras.datasets.fashion_mnist.load_data()\n",
        "(X_train_full, y_train_full), (X_test, y_test) = fashion_mnist\n",
        "X_train_full = X_train_full.astype(np.float32) / 255\n",
        "X_test = X_test.astype(np.float32) / 255\n",
        "X_train, X_valid = X_train_full[:-5000], X_train_full[-5000:]\n",
        "y_train, y_valid = y_train_full[:-5000], y_train_full[-5000:]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NdMv5zGbyLhH"
      },
      "source": [
        "Using dropout:"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "ZRGJBLy4hkgr"
      },
      "outputs": [],
      "source": [
        "tf.random.set_seed(42)  # extra code – ensures reproducibility on CPU\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "djTZNib3hkgr"
      },
      "outputs": [],
      "source": [
        "dropout_encoder = tf.keras.Sequential([\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dropout(0.5),\n",
        "    tf.keras.layers.Dense(100, activation=\"relu\"),\n",
        "    tf.keras.layers.Dense(30, activation=\"relu\")\n",
        "])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "M-m52b9Zhkgr"
      },
      "outputs": [],
      "source": [
        "dropout_decoder = tf.keras.Sequential([\n",
        "    tf.keras.layers.Dense(100, activation=\"relu\"),\n",
        "    tf.keras.layers.Dense(28 * 28),\n",
        "    tf.keras.layers.Reshape([28, 28])\n",
        "])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "gKqPpv0mhkgr"
      },
      "outputs": [],
      "source": [
        "dropout_ae = tf.keras.Sequential([dropout_encoder, dropout_decoder])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "Ly0QSOM4hkgr"
      },
      "outputs": [],
      "source": [
        "dropout_ae.compile(loss=\"mse\", optimizer=\"nadam\")"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = dropout_ae.fit(X_train, X_train, epochs=10,\n",
        "                         validation_data=(X_valid, X_valid))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1OPjGJPSkhPO",
        "outputId": "399add11-1eff-4591-8747-704ca8bbe26c"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 4ms/step - loss: 0.0397 - val_loss: 0.0210\n",
            "Epoch 2/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 2ms/step - loss: 0.0220 - val_loss: 0.0192\n",
            "Epoch 3/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 2ms/step - loss: 0.0206 - val_loss: 0.0182\n",
            "Epoch 4/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 2ms/step - loss: 0.0197 - val_loss: 0.0176\n",
            "Epoch 5/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - loss: 0.0192 - val_loss: 0.0173\n",
            "Epoch 6/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - loss: 0.0188 - val_loss: 0.0170\n",
            "Epoch 7/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - loss: 0.0186 - val_loss: 0.0167\n",
            "Epoch 8/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - loss: 0.0184 - val_loss: 0.0165\n",
            "Epoch 9/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 2ms/step - loss: 0.0183 - val_loss: 0.0164\n",
            "Epoch 10/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - loss: 0.0181 - val_loss: 0.0164\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "wTKWjK3EyLhH",
        "outputId": "d02403bf-5ed8-4a83-b918-7cf6d59a07b8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - loss: 0.0180 - val_loss: 0.0163\n",
            "Epoch 2/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 2ms/step - loss: 0.0179 - val_loss: 0.0162\n",
            "Epoch 3/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - loss: 0.0179 - val_loss: 0.0161\n",
            "Epoch 4/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 2ms/step - loss: 0.0178 - val_loss: 0.0160\n",
            "Epoch 5/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - loss: 0.0177 - val_loss: 0.0159\n",
            "Epoch 6/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - loss: 0.0176 - val_loss: 0.0159\n",
            "Epoch 7/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 2ms/step - loss: 0.0176 - val_loss: 0.0158\n",
            "Epoch 8/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - loss: 0.0176 - val_loss: 0.0159\n",
            "Epoch 9/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 2ms/step - loss: 0.0175 - val_loss: 0.0158\n",
            "Epoch 10/10\n",
            "\u001b[1m1719/1719\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 2ms/step - loss: 0.0175 - val_loss: 0.0158\n"
          ]
        }
      ],
      "source": [
        "history = dropout_ae.fit(X_train, X_train, epochs=10,\n",
        "                         validation_data=(X_valid, X_valid))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "pbHCAPiZhkgs"
      },
      "outputs": [],
      "source": [
        "def plot_reconstructions(model, images=X_valid, n_images=5):\n",
        "    reconstructions = np.clip(model.predict(images[:n_images]), 0, 1)\n",
        "    fig = plt.figure(figsize=(n_images * 1.5, 3))\n",
        "    for image_index in range(n_images):\n",
        "        plt.subplot(2, n_images, 1 + image_index)\n",
        "        plt.imshow(images[image_index], cmap=\"binary\")\n",
        "        plt.axis(\"off\")\n",
        "        plt.subplot(2, n_images, 1 + n_images + image_index)\n",
        "        plt.imshow(reconstructions[image_index], cmap=\"binary\")\n",
        "        plt.axis(\"off\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "Pucj7MI3yLhI",
        "outputId": "125881d0-56ae-4c25-b84b-c1051de64080",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 326ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 750x300 with 10 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlkAAAD2CAYAAADlLZJVAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy81sbWrAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA/fElEQVR4nO2de7Bf0/n/H1VEiCSSyFVOLkTc4i6hkRBFqaZU0WpVMWoYjGqrZfh2OjoYM1XF1FQHNe2otNUJ7ShSl1RGc0FI0yCJRCKRe+IWWtr6/WX93s/7nM+z9/7ksz7nc+L9+mutWZ+999rrdvZZz3s9zzYfffTRRyaEEEIIIRrKpzq7AkIIIYQQWyP6yBJCCCGEyIA+soQQQgghMqCPLCGEEEKIDOgjSwghhBAiA/rIEkIIIYTIgD6yhBBCCCEyoI8sIYQQQogM6CNLCCGEECIDn+7sCgghPhls3LjR5b/2ta+5/F/+8peUvuKKK1zZzTffXPO+V199tctff/31KT106FBXdtVVV7n8RRddFNRYMP/6179cvlu3bp1Uk+rcdNNNLn/llVd2Uk3EJwntZAkhhBBCZEAfWUIIIYQQGdimqwaIPu2001z+gQceKHXdH/7wB5f/8pe/XPO3l1xyicsvXLjQ5R977LFSzxTik8rJJ5+c0n/+85/D306cODGlBw4c6Mruv//+htRn3LhxLj9z5syUPu+881zZZZddltIHHnhgQ54vmguaN1988UVXNnbs2GZXR3wC0U6WEEIIIUQG9JElhBBCCJGBljpdeNttt7n8pZdemtJsPli1apXL9+7du+Z9N23alNJsHuT7vv/++ym9zTbb1LwPP3P33Xd3ZfPmzUvp5cuXuzI+8dTqsEWZ2wX51re+5fJ33nlnqWfceOONLv+DH/zA5X/84x+n9P/+9z9X9n//93+lnmFm9v3vfz+lP/zwQ1cWnWDDsWjWfqwiF154ocuvXLkypc8++2xXduaZZ9aubBcBTffvvfeeK8MTg5dffrkru+WWW1x++vTpNZ+Bpw2jfirigAMOqFm2dOlSl49MhNOmTXP54447ru46dSX+85//uDyulz169Gh2dQrB04/7779/J9ZEfFLRTpYQQgghRAb0kSWEEEIIkQF9ZAkhhBBCZKDTXTiceOKJKY36DTOzYcOGpXT37t3D++BrsGZowYIFNa/bb7/9XB6v5fuwhgfL+Rn9+/dP6TVr1tR8fldnypQpLs8aozPOOCOl3333XVe2efPmlI70OMz48eNdfsaMGaWvxWPb3L+f+pT/nwPH1N///vfSzzjhhBNcHrV8s2fPLn2fViHyqG5m9p3vfCelt9tuO1eG+jn2uP3DH/7Q5X/0ox+l9PDhw13ZgAEDUrpKXzDnnnuuy++44441f4t6o7ffftuVsTuXe+65J6UPPfTQuuvX6jz11FMu/+lP/39Zb69evVwZr63Yb0cccUTD69YR6Gbn+OOPL33dP/7xD5eXnkvUi3ayhBBCCCEyoI8sIYQQQogMNN1cOHnyZJd/6KGHav4Wj1uzt95c4LYwbxnvvffeLv/SSy+VumdbW5vLL1u2rM7atR5HHnmky++6664ujyY5Nrc+8sgjpZ/zhS98ocN7cv7BBx8M7xOZp0866SSXj6YGX/tJ4tprr3V5NLOiyW9LYLMemuP22WcfVxbJAZhzzjnH5bfddtuUvvvuu6tU0YHe4W+99da679MZfPDBBy4/a9Ysl0dXNeiaxszsrLPOSunnnnvOlR1yyCEu//TTT6f0UUcdVbM+bJJkjj766JplU6dOdfk+ffrUfCa6YWEXLSxBYImCaH1eeeWVlOb+PP/88+u6J7sPYolJR2gnSwghhBAiA/rIEkIIIYTIgD6yhBBCCCEy0OkuHJCRI0e6/C677JLSbAuNNFp83Bbtplui7eIwG3hfjPZu5nUirFPauHFj3XXIBYfLwKPZzLhx41Ka9TGRruXwww93+cidwec+9zmXj/Rb+FseJ3iEu4hJkybVLHviiSdcfuLEiSmNup6O8hiChcO6NEtruCVccsklLn/77beXvvaqq65K6RtuuCH8LWq9ePz9+9//Tmlu3+uuu650fSIuuOACl0et0r333uvK2E3Ho48+2pA6tCKvv/56SnP4sCqgZov/7ERuLxrlTuGaa65x+YMOOiileUydcsopdT1DNBbUVZm1d4c0YcKElP7Zz37mylBbuH79elfGfxdQq4nul8y86xlpsoQQQgghWgR9ZAkhhBBCZEAfWUIIIYQQGagtvMkEhqowMxszZkxKs30TdRgcroO1QGhTZxt+FbA+7BPmhRdecPnDDjsspf/73//WvCdrsK688kqX53Ajrc6GDRtSmkOcoI3czOxvf/tbSrMG6/TTT09p1rSxBuu4445L6aitmUhnxbDuKgLrUCUkEPsk6goUabAwPA77yYp0WBxWJyrD0D3bb799WB/87U9+8hNXdtFFF7n8HXfckdI8Bn/961/XfAZrsFCveeGFF4bPbDVWrVrl8ji/zcx23nnnmteir0CelxxWh/1m1aKKBuvZZ591edZ2RX4PcT2v4mtNVAe1TPx3ntddXLP32msvV8Z5ZN26dS6P2qr77rsvrF+3bt1S+uSTT675uzIarHbXVL5CCCGEEEIUoo8sIYQQQogMdLoLh9GjR6c0mwHYXIewOwXc+uUyDLsyd+5cV4bHeDsqFzGjRo1y+YULF7o8mgg47AbCZj02q6GLiZkzZ7oyDLNRFJIDiUybZj4MBx/xxudwuA4MH8IcccQRLo/hYzisz5AhQ2reJzcYLqVoqx3hkDvoXiEKx2PWuJA8VcD2x9A9RZx33nkuvyUheTobNheyO5fIbcPixYtTeqeddnJl/Kdl0KBBNe+DR/UjkxD/tnv37qXrWoXHH3/c5Y899tiG3Fe0Z/Xq1S4/YMCA7M9kyc5bb72V0r/4xS9qXlfF1dHHaCdLCCGEECID+sgSQgghhMiAPrKEEEIIITLQdBcOzIcffpjSL7/8sivDI8DswoFBbdUOO+zgytC9/tixY10Z37esWwYzs+eff75mGR4dfvfdd10ZhmQxM/vVr34VPqcZROEC0F5tZtazZ8+UZg1WpNHitketXNGxfLR9jx8/vmbZMccc48qefPLJmvfkI/usrUIdVqT1Yg0WjiF+Dh8jx/AOnanBYlCHdfXVV7uy66+/vq57sn4hcuEQccUVV7g8Hr82q79+fHT7z3/+c0qzTgQ1a2ZmJ554YkqvXbvWlUVaxFaAdSa87qGbhBUrVrgyDD/zxhtv1F0HnGuvvvqqK2NtV5FmC9l7771TGt1NMA888IDL898QkY9cGiwMhYVhc8za/01rlJavI7STJYQQQgiRAX1kCSGEEEJkoOnmwjlz5rg8bg2zOwX0PDxy5EhXNmPGjJrPOPLII2uWoenQrP2Retya5mP7bO7A5zzzzDOubEu8zncGkSdb3s49//zzU/qyyy5zZYsWLap5n6jtI/cJZt4kN27cOFeGkQHQBNkRaMrj/mSzH5s3y8LmFxwLI0aMcGVLliyp6xm5ufjii1N68+bN4W/RRMvm2R/84AcpfeONN5Z+PnptN/MmrCLzIJoTb7755vA5uMawK4OhQ4emNB8zv+SSS1z+nXfeSelNmza5sj/84Q8p/eUvfzmsT2fAphI2+6HcoUePHq4MvdtHR98ZdMNg5k2A6BbCzGyPPfaoeR/+LfcTrjH8N+RrX/taSvOawu5URD4iqUpUxrCZHk2EPDb5Ptdcc02pupZx2cBoJ0sIIYQQIgP6yBJCCCGEyIA+soQQQgghMtD0sDrTpk1z+W9961spvXTpUleGWiC2tR9++OEuH2mpOAxLBN539uzZpa/r3bu3yw8cODClUTNk1v6IcqsT6YimTp3qyvBIN4Mhdsz80fYqoWlYc8d6uAiMzL5mzZrS120JqANbv369K5syZUpK85huJuhKxazYZUpZMFROvS4bGNR5mbU/jn3HHXeUvtcJJ5yQ0o8++qgra2trS2nW56Euke+zNfPQQw+5/OTJk0tf+9prr6X0sGHDXFmVsDroigFdNJi1d5eBaxev0QjrRXltOP7441OaNYFiy4h0V/WEsekIduGA7h2K+Otf/5rSrE9GN1O10E6WEEIIIUQG9JElhBBCCJEBfWQJIYQQQmSg6X6yjjvuOJdHHdY+++zjylDXxLBeCn1ssS8u1MSwfwy2xReF0qnFZz7zGZfHkByjR4+u656tAvoAYjjMBvupQq1cFF4k8n9i5rUXHA5n5513TmkOYcT06dMnpVm7x30/aNCglMZQQmaxLoQ1G3Pnzg3r1ApsiQYL243bFDUV1157rSu77rrrXL6sTy2uaxUN1pe+9CWX57GETJo0KaXvueee0s/oarBGlENcoR+tdevW1bwP+xnj9Zt1WAjqsLg+7N+K5xvCus8PPvggpTkcE/pQq9cnnqhOFd9XRRqs8847L6XvvvtuV/bwww+ndJEG6ze/+U1KL1++3JXdcsstKX3bbbe5MmmyhBBCCCE6CX1kCSGEEEJkoOnmQt5uu/TSS1P6/fffd2ULFiwofd/IJINbjrxVGR2bP/DAA12etzVXrlyZ0pE7gJdffrlmWauy//77p3RkIuDwIpyfMGFCSnPonPHjx6c0h0nCUC1m3jzH3HfffSnN4Yzuuusul8cx9cADD9S8p5kPLxKZrrlubLI4+OCDU/q9995zZV1hbLCXFzYJs4kQQdMTmwsZNBGymwY067E5i7noootqlv3xj38Mr0XY9UpZcrnDyAWb49DVgpnZU089ldLsuuLNN99MaZ4jkRmSQ/lE5neuT2R2fOyxx1we15++ffvWvA7f0az9PFWYncZRJA3Bv9FFv8Xxx+DfF+Zzn/ucyz/yyCM1f7vTTjul9JlnnhnWpyO0kyWEEEIIkQF9ZAkhhBBCZEAfWUIIIYQQGWi6Jgs1WGZeD8Au9BE+pv29733P5aOj+3jEvChUDupp2B7M9Yt0WN27d09ptu+zOwAOC9IZsO6KXTPU4te//rXLn3322S7P+hQEtSrogsOs/dHdhQsXpvSoUaNc2VlnnVXzGXwcFzVZp512mivbsGGDy+Ox3qiv+Rjv/Pnza/6WwXGLrii6KhdffLHL//znPy99LYbgYe0kzj12kXHBBRe4POq3Bg8e7MpY74OuAx588EFXVm/EsVbXYDEY0sbMbIcddgjzCOrxeNzzeEYdFuuskCIXDhEY/obhedqvX7+UjnSnorlEOqy3337b5VFjyXpLdNfCoZqGDBni8qgzZd346aefXlDjGO1kCSGEEEJkQB9ZQgghhBAZaLq5kD0u89Ywgl7T+chvlePVaCJkz75sloi8kvO29fDhw1MaPdebeRMhukMwa++qAj2q9+jRo+bzc4Jb52ZmmzZtKnUdH+fHqPdmZkuWLElp9Lxv5k2CbGJhE0WvXr1SushEifAW8jXXXJPS6FrBrL3JEs0L3D54bdExY4S3rdF1RauaC3mOsMsGdHVRxTx42WWXuTx642fv+7fffnvp+0awJ3k8ml+vebCrw2MyWpMZnLdsjuP7oImQ3TCgC4dINlIVfObkyZNd2fr161P6oYcecmX8W1HNU3suOELDb3/725S+8MILXRn+vfnxj3/sytjVx+OPP17zmegeCqOAlEU7WUIIIYQQGdBHlhBCCCFEBvSRJYQQQgiRgW0+aiEhAut78Nj+nnvuGV571FFHpfTTTz/tyo444oiU5tedOXOmy3NIFCTSa1Vh9OjRLt+KoVV23HHHlGYNWb1MnDjR5adPn17zt2z7Rt0aps3MFi1alNLf+MY3XNnf//73yvX8GDzmu3nzZleGfcj6QNYuIOwmYtq0aR3es5VgfRRrstDdBx+NRjhUDuujUDexceNGV4ZjsHfv3q7s+uuvr/nMKgwYMMDljz766JS+//77G/KMrsDixYtdfo899kjpe+65x5Wde+65Kc0hrViLGj2DxxjC+sworA5ra9BdzqGHHurK8O8N9y9rY7emsDqseWN3OY2+jonWR7NqWi/U47JWF0HdqFmsrXriiSdcHrWiBxxwQOm6fYx2soQQQgghMqCPLCGEEEKIDOgjSwghhBAiA9n9ZFXxrRHJw6688kqXv+mmm1w+sumjLgf1WR2Buiv2o1SFXXfdNaXb2tpcGb/nLbfcktKXX3553c9sJKhlQH9lZmYnn3xy6ftMmDAhpaP+ZZ9JbEPn/kdQL4H6kY746le/mtLoY8XM7Jvf/KbLR6GF0N/a888/Hz4TYW1hq+qwqhBpLFCHxT5ueKyj9oF1gBhW6+qrry5dNw65dN9997k86oZWr15d+r5dHdTXsLaG5xD6kNpll11q3pM1WFHoHH4Gh/aJQJ9aHCbp2GOPdXn2h4TgeoQhtMx8WDSuH/sV62rUq6Wq97oiovB1HE6Pw/JFOiyEQ9ex3z3UdbIeLwqpVgbtZAkhhBBCZEAfWUIIIYQQGchuLuTQCmxqwu3nZ5991pXhdu68efPC5zzzzDMpzaFzcDuSj/Tzb2fNmpXSbAY6/PDDXX7OnDkpzW4F+Ah6VwNdOFQxD7IbDuxDNpsit956a3jfRx55JKUxTJKZN80WuZuITIBohjDzYRmwr82q9S+andHdRFeBXTYw22+/fan7oFm8I370ox+lNJsJzjvvvJRmVwsRbB5kszS6HUBzJf/2hz/8YelndgVwnvAafeCBB7o8hphh8xuagLt16+bK2OSGocYYNMEVmQ7RRMh/M9hNA9Zh1apVrmzgwIEp/fWvfz18Jrun2FqZMmWKy5955pkpPX/+/PBaDqtUiyIXDbhOfPe73y11zyLYrMxuXzBE3u9+97uGPPNjtJMlhBBCCJEBfWQJIYQQQmRAH1lCCCGEEBloqbA6rQBqtFCf1RGoP+JmPP/881P6rrvualDtOodTTjnF5adOnVrXfVADYebDZfARW7bb4/Fhvg+672DtBNviEdZgRbBeB8PjjBgxwpXxWFi6dGlKo+3fzOySSy5J6e985zul69OqoHbKzGy33XZLaXb1wG5XevXqldLXXntt3XVADRG6HzDzYWDMfJgYDt/C4ZK2JtavX5/SrFNkTRayadMml8cQR/3793dlfPQ9mm84T/l3qA81M3v33XdTmrWwHPoFtWfoTsasvYYMmTFjhsvvvPPOKR21T06quENCXnzxRZf//e9/7/IYziqC22T8+PEu//DDD6c0hyFCTStqXRtJ1D64BpuZ3XzzzS5/2223pTT/DcExz2OoDNrJEkIIIYTIgD6yhBBCCCEyoI8sIYQQQogMNF2TxaEXUNfEIRumTZuW0myXj+DQOegba9y4ca5s5syZpe/LNmi2Uddi1KhRLs/aj4suuiilL7744tL1aQXYb84JJ5zg8ji8Bg0a5MowdA5rldi+vmzZspp1QD9U7NOJ/VsddNBBKT137lxXxiFv6vVpxWE30O8Pj4WFCxfW9YxWAn2Goc8yhn1UsaYE/eOcc845ruzee+9NaQ6rw/6PUGfF2hDUjRRxww03pPRVV13lylhPVuRLrJV54YUXXJ61VaiBjDRZVWCfS+hjictQD2VmNmzYsJTmNZg1Waid/OIXv+jKqoSJwXBrhxxySOnruhr4nmZeM8prcpFmuSw8R1Hzxv6smgFr2A444IAtup92soQQQgghMqCPLCGEEEKIDDTdXLj77ru7PIZlWLduXc3ronAtRaD5kMPqbMlvcRudTRZIUd3POOOMlG60S/9mwyZfbCMOlzF69OiU/vDDD13ZBx984PKvv/56SrPJGY/cVjHHsXsHPrKPY5NDx6AJg0P5cJiSTzLoluInP/mJK7v88stdPgq785WvfCWl77//flfGbiNwLv7lL38J64fuSR5//HFXhiE9trawOgibiHiNfuCBB1IapQ1mPtTYqaee6sq+/e1vu/zKlStTms16OP/LhmepCj+zirnw6aefTumjjjqqYXWqAq+JaG6P3oVlESiZMPPmWQ4Xhu4y2CXHTTfd5PI4blhGgvC6269fP5f/29/+VvPaemG3DA8++KDLX3PNNTWvxXc5/vjjKz9bO1lCCCGEEBnQR5YQQgghRAb0kSWEEEIIkYHyRuk64WOfqK0xK38MtEePHi7P7u3rteNGLh1Qb2DW/sg5h1pB0JbNGqw//elPLt/VdFioy3jiiSdc2dtvv136Pi+//HJKt7W1uTIeJwjb15EqLhE4fAdrFZYsWVKzfnzMHEGtmVnsfgR1YOimwCxf+Imq8PhljSG+A7snQR3WBRdc4Mp++ctflq4DryMIu1NAHdYxxxzjyp588kmXj0JEsf4IYQ3hdtttV/O3XQ0MhWQWt8Ps2bNL3xfHEbs5QTcSrBFjlwn4Ww5xw+4osLyKBotdQ1RxIZQL1oWWhdc15qmnnkppdNnA/PGPf3T5KmHJTjvttJTm9eOnP/1p6fvUCz8z0mCxdrjedv8Y7WQJIYQQQmRAH1lCCCGEEBlouguHetlnn31cnk11eMSWQZNgZHYw89uKRU1TZau8q4EmmC3xZl3WzQUzdOhQl1++fHmH9+T78vFzNjuit2jeQq7iZR7H1FtvveXK+L4LFiyoeR80Z9x5552ujM3VnUWRuTACzQ+333576evOOussl3/vvfdSmteCzvAKvTVRZJ5DeD6hR3U23VW5bxWP6s8++2xKH3rooeFvf/azn6U0u/pgCUoEuvPhiCKtzpQpU1ye109024AuTbYEjMJhZvb888835L6dAZoP2cxdBu1kCSGEEEJkQB9ZQgghhBAZ0EeWEEIIIUQGuowmqwqRW4ZWgMMjbOkR0c4k0k6Z+Qjm77zzjitDFwlVwuEMHz7c5TFcRuT6gRkyZIjLc7+gdqFv376uDEOwTJo0yZWxW4sIDBGDoWNaiSqaLD4CXkWHhSF4WL933333pfQVV1zhyjZt2uTyqJHjY+eimMWLF7v8tGnTUvrcc891Zd26dWv48zmcGbvOGTt2bF33Yi0VjqPNmze7sl/84hcu3wphddi1QFl9EIc34jUawyZF4FpuZjZ48GCXf/jhh0vdpwqsk+WxUJarr77a5VnHiW4uxo8f78rQ9Uc97ia0kyWEEEIIkQF9ZAkhhBBCZEAfWUIIIYQQGeiymiy2y5cNz1MF9lPEfrHQxs86gij0Q1cgCpUSseeee7p8r169Uhp9VJl5bc/q1atdGYdJGjlyZErzkMVncLgltunfeuutHVfc2murFi1alNLsWybykcY6JNR7fOYzn3FlbP9vRbbETxZy4YUXuvxOO+1U87c8Ht58882ULtJ+oLYN+9DMrGfPni6P+jkOJ7N27drwOV0ZDE2FPsjMzHbZZReXf+aZZ1L6nHPOcWVVQtW0MmeeeabLs7aws3RYOeDQZ6gJ5vWye/fuKc2aMJ5LAwYMaFQVs/PVr37V5X/729+mNIYAMiuvWauFdrKEEEIIITKgjywhhBBCiAw03VzYKNPDYYcd5vJz5syp+Vs0yWC4GLP2Zr4jjzyy5m+jkDwcguXmm29O6a9//es1r+uK4JH53r17N/35kVnnuuuuc2XXXnttU+q0NVNlzvIR/3vuuafmb+t198DhUe6++26Xx5A86PqhI0488cSURhOamdmKFStK1WdrA8PWmPl2aFTYlVaDwzidccYZLo9ygSouJITQTpYQQgghRAb0kSWEEEIIkQF9ZAkhhBBCZKClz9+ef/75KX3XXXe5skiDtd1227k8akhYZ8XgcWV2XTBixAiXf+mll2reZ/78+TXLMAyMWWseg0YdDmtwGqXDwj7FI/pmPsSKmQ+BwVqeX/7ylynNep1WY2sKqfQxGC4p0mBx2KAqrkFOOumklEaXHR3BR9QjUGf5SdJg4XH8bbfd1pUdeuihLv/GG2/U9YxI7strSrTeRPetV9NrZnb55Zen9H777efKOKzT1qpFE/nRTpYQQgghRAb0kSWEEEIIkYEu6/E9Ak1+Zt7jNm+Ns5dvZN68eS7PrgO6kodbIZoBu8xAlxpV3GtMnjzZ5R966KGURtOxmXeXYmZ29tlnpzSaMs3Mpk6d6vJoBtqwYYMrO+SQQ1L6lltuqVnXrg57tF6zZo3LozuDvn37NqVOnc1NN93k8uiGZ9CgQc2ujujCaCdLCCGEECID+sgSQgghhMiAPrKEEEIIITKwVWqyhBBCCCE6G+1kCSGEEEJkQB9ZQgghhBAZ0EeWEEIIIUQG9JElhBBCCJEBfWQJIYQQQmRAH1lCCCGEEBnQR5YQQgghRAb0kSWEEEIIkQF9ZAkhhBBCZEAfWUIIIYQQGdBHlhBCCCFEBvSRJYQQQgiRAX1kCSGEEEJkQB9ZQgghhBAZ0EeWEEIIIUQG9JElhBBCCJEBfWQJIYQQQmRAH1lCCCGEEBnQR5YQQgghRAb0kSWEEEIIkQF9ZAkhhBBCZEAfWUIIIYQQGdBHlhBCCCFEBvSRJYQQQgiRAX1kCSGEEEJk4NOdXQHkv//9r8t/+OGHKf3ee++5so0bN7p8z549U7pbt26urHv37im97bbburKPPvrI5V966aWUfuONN1zZgAEDXL6trS2lt99+e1eGz/n0p1uqmbeY//3vfynN7Vf2OjOz7bbbrmF1aiX4PXFcf+pT/v8azG+zzTZ5K1YB7FeuF7/fBx98kNJvvvmmK1u7dm1Kz58/35XhXDMz69+/f0oPGzbMle2yyy4pzePmP//5j8vjvH300UddWY8ePVx+7NixKT1mzBhXhvN9p512cmU831u1HxsN9rWZHwu87jJvvfVWSj///POurG/fvinN/fvvf//b5XFtHTFihCvDtZ7hcctzUVQD/z6b+X7acccdXRn/3Y3YvHlzSvPf/d69e7t89Lf1/fffr1nG9cH5zOME53M9c1ujTAghhBAiA/rIEkIIIYTIQKfbsXD7eeHCha5s2rRpKT1nzpzwPmhq+Ne//uXKcBubtzF5O/L1119P6U2bNrkyNoWgSeMLX/iCK5s0aVJK9+vXz5VxHVrRvIDmIt52xW1h3spncwKy6667ujz205IlS1zZunXrXB5Ns2yWwPpxW/bq1cvlsQ/5mWymGDVqVErvttturgy3tN9++21XxuYrhNtghx12SOnI1JEbNvuiiZPfZ/369S7/3HPPpfSsWbNc2YIFCzpMm7Xf6kfzA8sBcFxx/0fzndv7nXfecfl77703pXfffXdXhnP4pJNOcmVsWkRzIpskW2F+c/9GdYokADzfsc9wLHOZmdnixYtTevr06a4M11025XB7jhw5MqXRjGzWfq199913a9Z95513Tmk2B1eRMkSm9a4Ozn2e99xG2BdVzIMM9kUkBTDzY4PHCa5hvEazNAnNkDguzNqbD5Ey/a2dLCGEEEKIDOgjSwghhBAiA9t8VOV4WAbmzp2b0jfccEPNMt5+RPOgmTcR8G9xG5u3OPH0k5k/cYJbzWZmq1evrpkfNGiQKzvllFNS+tJLL3VlbJbA+m7JNmsjiU6MocmF2w9PEPFvV61a5crQzMemWd6yxdOjURuxSYDB/l2xYoUr27Bhg8uj+Y5PluKYYpMk1wG3v/faay9XhuYNfgZvf+ckMnHyuJ86darLP/bYYzXvgyZBnnv8fjgGeO6hebboFBua49mcxGYhHNs8PvG92ez47W9/2+XRtIgn5cx8GzTzRBsu7dGJKTbrsdkPTXloGmb69Onj8iwzwPnE5psnn3wypXlNOfjgg13+sMMOS2k2FbMZCMcRvyf2BZogzdr3IY4bHkPYlmwCb8UTjPwnH9uF5x2u33zdkCFDXL4Zp+h5jcY1g0/84hjjtYfHDb4Lj+MtpfVGgBBCCCHEVoA+soQQQgghMqCPLCGEEEKIDDTdhQNrNn71q1+l9DPPPOPKUBc0ePBgV8a6HLQrsxsB/G3RUWa01bL2g+30qL1ZtmyZK5syZUpKs17r4osvdnm0JbeKJgvfjfVSK1euTGk8lm1m9tprr7k8uklgXQvqZbjP+L7YL6yzwWtZk8EuOrB9WX/AYP/z8WXUjbDehd8T3T+wC4Fx48bVvK4zwXdiHc7jjz/u8jj3hg8f7sqwP/jdWbuGGi2ea9Ec4fvgM1l7wXkcS6ytGzp0aErPmzfPld15550uj57HUT/Idc9J5IaD5wX+lsu4jX73u9+lNOqzzPyYHT16tCtjzR2u/ezGZv/9909p1N+Ztdew4jjidYPHAroFYg0m/nbfffd1ZawDQ70ka3bwXaKIDp0F9y/PQ2xDXluxD1lz1RnuKrjtcW6x5g41lqwrjrS7PH/xPev5+9z5I0AIIYQQYitEH1lCCCGEEBlourmQg8KiOYmPWeJWIG+FR6Ye3qKNtvvYnITw9iMTBaXG46NPPfWUK5s8ebLLo6mhimfmRsJbyri9ysdm0TT6z3/+05W98sorLo/156PhaBZgD+9sdsT2ZfMLbv1y33N/4zFyNkvwFjJunbPLEByrbEJZs2aNy2Nbcv3wvuzCodFHiSO4ndDj+j/+8Q9Xxi4dMPIBzwOcX5EbAYb7GK9lExGbnrCN2fTF4H35mXhUH73/m5nNnj3b5Z999tmU3nPPPWvep5ngWsLrCprueExy5I3o3dCUV+SJH9cYbmucl1yfKCA5r9H8nmiq537Av0UcvJzN3mi+5vfCtaCTPSIlsM14neM6orucIvco9YLuPIrWaDRLct3Z7IflPKawjPuX12iWpyAo96kSoPpjtJMlhBBCCJEBfWQJIYQQQmRAH1lCCCGEEBnIosmKopJzFG2007N7ByxjvRbfF8ujI6usC2J7P+a5PmxLRnss25XRto36FjPvAsHM2/8j3UozjwNH0deXL1+e0qzXYl0T6mVYu4S6Oj5WzLZ3bE/WFGC7cP+yzRyvLdIGoN6DtXuoIeEjv9wmUWgP1Hd05nFvblPU3S1atCi8FvuY5x7qZ3jOcj5y04Djkfs40m/xusHgtTxW8Jncxxz2acaMGSl94oknujJ0DVFUny0hWhOjUCqsVWJdJY511gliG7FONtL3cF2xT3ndje7Dv+X7Yj9FIZZYs8NaIBzj2B5clrN/qxDpkLkdmkG0tnEoNlxrWX/JfxfQvRD/Fl12sM4Q/4aZeU0wj3F0UyIXDkIIIYQQLYI+soQQQgghMqCPLCGEEEKIDGTRZEX+b9g/BfqTYps+2unZ3srPiPRbmC+6D/6W9VGRf5TIpxY/gzVa+J6d5WeFbc347vxuaEPn9uQ2Qp0L9y+2S1Fboy6DNRpo7+eyyB8P64cifUekNWONEGs2IvA9O9PHDrcF+o1BX3Zm7etZ5IvqY9ifFT8T2401TziOWN/B7Y9jmbV0ke6Ty1D/wXotrh/6EmPNJfo/axXNDvYhvxv7qMNy1sSU9cVlFoefwd9GfcTlvDbxeoTP4TUG+4L7hUOJIbxWtkqfIrh2VVmPGgW3NbYnrxe8tqI+l/924veCmdcLR+HfeFzwWMU8hkwz83/vOMRTGbSTJYQQQgiRAX1kCSGEEEJkoCHmwiqhYNhUhu72Mc35KkfzeXuZty4jeKsaYVMU1oFdEETP5DZAuO5l3PY3Aj5Gi0da2e0GvmvUXma+HbiNcJxw30dtzUeQcfuexx5GkTfz5gPeQo7CO/BvcZu6yMyHbYmR4c28CbWs2S0HHCoHzV9FR/NxC5/NSdim3BdscsOxEs2fKMyKWWwKYzMVmpvY9QYeAecxx/MU+3XBggWubMyYMR3WzSxv2KwoFBG2A5treCxgHy5dutSVoSlqyJAhrmzgwIEuj+OITWyR6xwef+hegfshMi3yGodjg6/jNRrzXPco5FdnuWXB92GTapXQOTg2+G9wBK9zr776as3f8tjE9YTbk/9O4Fjl+Yt/b7gfOIQZzks2O6K7h7322suVlTEVaydLCCGEECID+sgSQgghhMiAPrKEEEIIITLQEMFPFV0B2//RXhwd6+Uy1mWgTTU6/s82VLbx4n2KtFwYloHDMERHh6PQQp11jJ81Cahf4DLUaBTpmrBdomOzPC7YFo/3ZW0Fhj1gvQH3N9a36Kg45llPhvfl8ca/xWfyOMHxx+4GcoNzL9Lk8RzBMDFmXgsRhUvhOcwuHSJdBM4h1llFWjrWo0SuVlhPhjocHhtcd3xv1qNgn7OuhUPaNJJo/cR5wuM1CiEVabJwHprF+ijWPEU6IdZk4VpRtEbjc9AtiZlZ//79a9aV9T3YJtGY4jWOx0mzwLbmOcl1xDbkdqiiycJxFLl9KQpNU1ZLyOX8NwPXKe4HHlN4H64f6sn23XdfV9bW1tau/ox2soQQQgghMqCPLCGEEEKIDDTHPwDAZolo+x63+3g7OTI9sRkIn8Gmhsi0yERbwbyVisdQ+R35iDTeNzKZ5nTnwFvKuGXKz8X3ZjMfb+9GLh7wWr4uMsdE5mnuB74vUmR2ilyIYJ/y1jOPcXRdwHXHLW7e7s4N9g2bytCMFfWFmX9/Huv47uwGgedTtGWPZdF1Zt41RGT+MPPzn81mOIe5DfgIOOajOVHlGHxVuF2wDdmshuscyxfYXIh1ZlMortHc1gyOm8iUyHXlMYXvVeQSA+diJNPo27evK2NzJvYpr9F4n2a53GGiaBVFUhuEf1tFChRFr9htt91SmucA153XCYTnKJqreR3AtaeK13ue6zhuOBqCzIVCCCGEEJ2EPrKEEEIIITKgjywhhBBCiAxkD6vDZaxXQPssa1IwTAPbkdk2i9eyHRntzNFRZjNv7+fj/5HehG36eASe9SVcd7RJc30iTVEjidwXcP3LaivMvL6DtR6oeeJxwvcpq01jjR3roxAeb9zWeC27LUAdEteVxxjWN/pts49749zj4/cYbZ6PLUeaFG5v7A/uGx4P+P4837ENo5AsZr6N+bfsGgK1aKyVxPHBrhbw+L+Z13ywXgvndLPC6JjFWk+kSA+JedYfYjvw89ldSaRHw3bh+3AfRr9ldyNYHoWHwvFu1n5dwzbiNRrrUxRmLBe8RkehxiJ4jlYJSYdtxvODw9EgPCdxnHDduT6Y57pHmmkeUxgmjZ+JOkTWZE2cONGK0E6WEEIIIUQG9JElhBBCCJEBfWQJIYQQQmSgbk0W2vvZpo/2T9Yxsc0cdRmsyxk8eHBKr1y50pWxngbtqGxvxd8W+VFCuy5rE9gWj9oLtsVjWAHWEHH9ytrx2R4d6Syqwm2P2prIlxj7NGHdQ/QMbBduW9a4oZYv8sfDbcLjLdK/8ZhCf0v828g3DvtkwXJuS8wXhZtoNNhWAwcOdGWf/exnU3qPPfZwZdymc+bMSen58+e7Mpy3rHXg+RXpK5Ai3WKkY+TxgesBj93x48enNOvSJk2a5PI4HgYNGuTKWO+Wi8h3YDS2eP2JQlPxM1DLFIXCMvP9i3PLzOtbi9ZDHH/RWs/PZF0llvE787qGczj6u1DkK2xLiMKbNUrrV6R3jMA6DB8+vPR13N/Yvry28lzCtTby8cX6zxUrVrg8aq14LGD9eO0rg3ayhBBCCCEyoI8sIYQQQogM1G0ujI7RInykm4/14hYob9P169cvpdesWePKeBsR7xuZ59h9QpUjq9HR4mHDhrmy6dOn17wPb12imYK30XELNKe5kLei8Wg7tyeagLkOnMej71E0eN6mjvqXTb742ygcBj+zKMI7wnVHcwLXh5+J281sksI+zWlqKKJ3794uj23BY5v7Zu+9907p3//+964MzYVFIVBwbvJYxzbGcDdmsWmRxxXXIXIxMXbs2JQ+9dRTXdmYMWNcHs0IbFKIXFPkpKy5kEMqsTkM78NmfDyaH4VIM/NjivsF1/5ojpj58cf9yaCkg90K4NF8NjNGoV/4mXhtTpc7zZAT8DOqhAnCsVGlrpGZOQoXZObHDY8TdA2xaNEiV4YSBzM/5ljugbIlDA9UFu1kCSGEEEJkQB9ZQgghhBAZ0EeWEEIIIUQGGhJWJ6LoeDXa8VmP1KdPn5QuCruCtlnW1qCttsh9AtqDe/bs6cpYC4L35WPbWF/WPPBR4uhYaBTeppGwTiyyi2MZH3VmDV5k6+ZnIjxOsH8jvRbr+lgngjZ9vk90DJrt/ai74jLuJ2wjHgvYBs3U6xSBR+qLtH9tbW0pzZoFHL88jlifiXOT+w3nE/cx3wfLeb5HY45BPQ/PbwZ1TKwxQRqpo2R4/ET1QFiLxvMCtS0cXihys8P1wWv5KH6ka4rmJbcna25xLvK6++qrr6Y0z0sef9hG0dgs0oh1NcqOIbP6/z7x31mE5yvPdRybGBrHzP/tefHFF2uWmfm683zANaSKa4qP0U6WEEIIIUQG9JElhBBCCJGBhpgLo23C6LirWfljvXwsPqoDb3fjtnWRaQfL+T68NY1bjuhuwiyOJF5k0kLw2pymBu4nNPuwORO3/XkLnl1t4H35Prj1i6Zhs/ZbttFxdGw/bkt2ixCZCzmPz+T64H24f6t418YxVOW4dG6qjDU0lUWmFDYXM/jbyDzLdeNnRs/hvsFnRmOHzWQRzfbc/zH1ev2OxraZ2bp162qWoakHzW9m8RhiFwm4FqAJyKy9JCGqazT++G/P2rVrUxrdOZi1d2OB5kR+RvQ3o6uDfcHzivswAqU23A/8dx/nHZt/ly5dWjO/bNmyms8vWnvQLMmmY6wfmsfLop0sIYQQQogM6CNLCCGEECID+sgSQgghhMhAQwzIke2dj0qyRgZt33yUE23x0ZF+M28fjuz9bE/n30a6ELYlo1aBjySjOwp+Btc9OlYehS9qpEaLbdaRjg31MezWgl04YP1ZV4dtxG3C98Xfcj+gDoKfwfoYtK+znoPbEzVkUZgN1m9wW+J9eRzjM3Jq7nISzT0M18N9zG2B13IfYzvxdawnjHSA/FscO1H7cx+3IvXqgViDwrpBnIvcDrhm89zjsYBuQRi8tig0TRQaKwqTxmsBrru83kRuQiK9aFfXZPH8QE0Wt18VTRau39x+7JYB+3TJkiWu7LnnnnN5LOf1fPTo0SnNmmnU45mZrVixIqXx77qZ1x1z+LEydM1VXQghhBCixdFHlhBCCCFEBvSRJYQQQgiRgboNyGjrjkIJsN6I9T1ow2ZdE9r/2fbO+q1I+xPpiyIdRlGYE9QfRXoyttOzXxC2USN4bc6wK9wO2Ibcv5HOIfJhFLU92+VZ+4H9y2MBdX5FvoJQ68N15ftiOd8Xf8vhoFh3iPeJfHFF/tK6CqzTiEJmMPj+rAtCquheeHzy2oD+r9gXVjQvW5EqfrKwfdkXEc9T1OWMGjXKlaFehdcJnsM4Z1iri2V8H+6HaGzwb7G/WeeHz+H1h9sS78v1wzFWr6+yzgTXK9QmmXnfU6y3HTFihMtH/sIwz+ss+yjDZ86aNcuVLViwwOVRazVmzBhXtscee9S8jn26oQ6L+xfXsHp84GknSwghhBAiA/rIEkIIIYTIQOl9dzaBYD4yuXEZh2FBU0vfvn1LPz8KrRKZ1dhUx1vPuK3JZfxM3PLmrfEBAwakNJtMeSu1bFidnFvRUfvye+P2Kb8bt2/kPiNykcDPxDybfLDu/PzoWDlvC/O1kRkA+5Drzs/E37J7B35mVydyi1F0xB+vjeZekZk3mv8897AObNosciXQavB4jtaLyBTOZjU0lbGkY+PGjSkdhZsxa2+WrFXGpjt+L5zvPE742miNwbnHYzMK18NlkVymFWHz3MqVK1N63rx5ruz1119P6T333NOV8dqFEhmeo1GIsuXLl7v89OnTU/rFF190ZW1tbS7/pS99KaXRPGjmx9+GDRtc2erVq10e/37z+Mc1m8dX5JbkY7STJYQQQgiRAX1kCSGEEEJkQB9ZQgghhBAZKK3JYjs02sWjI9Vsp+dwBXhf1jWhnTyy8Zp523x0hJ7vw/ZhrE9kwzfzeiTWH6Ami4+L8n3KHt3PqcniOkWao+i9+bfRsV58ZqR3499GYXV4XHBYFex/riuPGxx/PDaj0B7cJvhubMNHe39XPP5t5t830jFFOkqz+P2xjMcqtzeOnSIXDnhf1stFoVR4HWsFqoyfKNQTvyu6cOD3Rk0W66P4vrhu8NzDscB9FGmyeH5X0WviusFhdfhdIjdArajJwvd57bXXXNkLL7zg8osWLUppDrGEbcZlHH4G+zf6O4/jycxs5syZLj937tyUHjhwoCs7/fTTXX7SpElWizfeeKPDupnFbp54HuHaX48bJe1kCSGEEEJkQB9ZQgghhBAZ0EeWEEIIIUQGSmuyWMuEttrITxZrIthWi3bxPn36uDK0i7ONnHUDaBePwjKwXT7SR7FGh8OnRLoG9OeBvkbM2ms/Ii0FvneRX6ctgbVCvXv3Tmn2b4OaF24TfjdsT/YxgtoqLot0alVCIXGoFNSCsJ6DdSL4btxHkXaP+wnbhH+L4z96r2aD464ojA2Oj0ijx2Wcx/EQ6XuivuBrozBenOexi7oW1ghGmizWe+AaWCUkUE4i7VwUQgrXBTPfvqyV5PkVaalQ71OkccT1PPJJx/CagmMz0uZynv8u4H2b2b/o64k1R6hHYk0Wap7MvK6O+wXnBIfVeemll1we10v2OYdzYO3ata5szpw5Lo/t+/nPf96VTZgwwcqC44j7jPsJdVf8Nw3zvH6XoXVWdSGEEEKIrQh9ZAkhhBBCZKDuvc2y26JFIRvQnMPHNdEMUcVUFh0B5vrwVjSWs6kzMlmyOQHNhbNnz3ZlbMKKjgBjHXjLs5FE26lsxsI2YzMf53FLm7eiMQwDtwG7TMA8b/vjtbzVG8HvxWOT8wi2AY8T7l8MJcXvifWtJ8J7LnAOFc117A+el9gWbHbm9sVr2WQUuXDg+mEbF/0W82wCjEyUEdyPrWIiRCJTKJue0JTHYxvbiPuM5zuW8zzA+hS5y8CxyWtBFLKMxxuuefxMXuvxPpF7hyqhjarC4WcwBA6vZShRmT9/vitjl0LYnmwOi6Qh0XyJ/layywb+7amnnprS7KKhylzC8cfX8btELoKwbeXCQQghhBCiRdBHlhBCCCFEBvSRJYQQQgiRgbrFApEOA2HbJx8BRrcNu+22myvDI6NFITmiMrSpRmFWOF9kp8c868D233//lJ4xY4YrQy2SWaz9iTRijSQKO8DtgFoG1hxEmgTWWaGdnNsvOmLO4DFj7l/WiKGOgOvK2hTUn7CLC7xPFCrHzGuy2Ka/6667pnRnhtVhTR62Y5HLEXwnnk/RmOU+x/nE98GxU6T9wfqyFoPnMNaB3xPrXsW9RtFx8Vxwu2A9eO5FGqNI08ptFPU9j5NI54ll0TprFod0i9wrcBvg/OY1jsE1mp8ZaXYauWb/85//rJnnvymoceN+4P5FrSTrJsu6Z+E6cJgidCPxyiuvuDJ2y3DMMcekNK6PVcE1mrWEkeY3Wr/rcbOjnSwhhBBCiAzoI0sIIYQQIgOl97Hr3QblbTk2rfTr1y+lcVvOzG/L8nVshsStcq4rlhVt3ePWeHTc28y/G29pDxs2LKXZNUXfvn1dvqzbgZzmJL43tjebcXlLGeFtYty+j/qFvQlH5gQG68PmAjYBRp59I/Mwb8fjc3j7nbeb8V342DNuY3emx/fI7UDRuMM5E5kioigNDM8nNAlyXSMP61FEh47yCI5XfkYEj0HM5zT5cz9Fz8K2Z7kC9wuO0f79+7syfLdBgwa5Ml7nsK15HmBb81yP6sf9G8kM+G8R/u3hsRmZl7g+aDbLOYdZaoMSjxUrVrgydLvBY5elITi3uP3QGzy3H/cv/pbX81WrVqX0yJEjXRm7aRg1apTVgv+GRO2N84Gv4zmK6020FtZj+tdOlhBCCCFEBvSRJYQQQgiRAX1kCSGEEEJkoLSBsV5bc1HIEXThwDbfyPYehXpheyvamfk+nI/CibAuB8v5mahjYk0T6xEibUqziNqT+wzboSjUQqSPiY4+V9HVYZ/xdZGbgEivZeZ1adwGWAcuY60C5nks7L777ikduULpTHhs13v8vijkCGpF+D4IHx1nXVWkn2I9TVS/yI1JRNFalYsq4ydyp8C6Spx7PNZxPHO/DBgwwOWxDbnvcd1gXVCk7eKxyXXActQpmfm1i58ZhXXiZzSrfw855BCXx/Vq8eLFrmzu3Lk1y/jdNm3alNLsDgV1Vhs2bKh5nVn7tR9Bvd7RRx/tyvbdd1+Xj/qX5y/+lvsX61tUd3T1w+2Da4TC6gghhBBCtAj6yBJCCCGEyIA+soQQQgghMpA93gP79mCfUWhjjXwR8X1Y6xXZSiOtAtvTsT5s/2XNDtaBbccI6m7M2muK2MbfGXAd0L7OehT0A8VtxH04ePDglGatDD6jyH8R2sW5rbEf+BlR6AfWhUThFNCnDt+HxyLqDM3MhgwZUrMMw0Y0K/xKR3CbRr6xWC+F9Y76kedaFH6GnxGFuOF2w/leJcwPazFwvvO4quKvB7VeOX3dcZ3wufzeWA+ew9wvuPayL8PRo0enNPoG5OvMvN6H2wG1PzyG+D74nnyf6F0WLFjgyvBv0aJFi1wZrxtYB14nmgWPc/y7wv2CvqZ4/eZQY6jBW7t2rSvDfFFbow81/juw3377pfSIESOsUeBc5zUa8zyG+D0POuiglGZt11e+8pWUjkLg1UI7WUIIIYQQGdBHlhBCCCFEBkrbJ6JQEdEWGh/r5LAMUbgChLdK+bg1bofztnlksig6Vo7w8Vb8bWSSZHMhmx7YDFmLKnWtCpsTcCudTVx4HJe3YaNtdg7ngNu5vEXL5iLsw8iUVGR6xW11Po7O1/bs2TOluQ1wjPG4RfOgmW+DKKxObnMhzwscP9HY4rHBeXwnnl9oNqjiXiMyqUfHuPlaNqPwPMU6sQmh1j3NYvMg/zaniRDh9o3GE7YD9z27NsH+jkxGRetYVI5mcx5fjXJtws/HukdhVrhOkXuRnPD6iWOQTZhR3/P8wfnLz1izZk1K83oZ/Q0uCsHTbNhEyX+T8buE1xMcm/WgnSwhhBBCiAzoI0sIIYQQIgP6yBJCCCGEyEBpEQjbyesNBcO20HfeeSel2Vbc1taW0qyJWbFihcvjMVQOC4FHWIvs/Wjn5nfkEDJo5z344IOtFnic1sxs2bJlLs82YAR1FlU0F1VhrVB0xBvbhfVH3L94NJb7Ba/lZ7BeJzrCj3Uv0oWgDozfucrxXDzuzzqVKhoS1ILkDq/E9cTxxM/GOcNaNX6/t956K6Vfe+01VzZr1qyULtJSIdzH+Ex+Dw6RgfVlrQrXHcck6wIxZEyVcBr820gv2siQLHyvSOOG447rxEf+I+0S/nZLXNNg3RupYcP68tqP9eV3xr9LZn496qwwaFHYGn43fO9I32rmtVasu8L1ku/DbYZrIv8Wxxi7d+C/q/jbRs0Pntv8twhdiPDfEHwvuXAQQgghhGgR9JElhBBCCJGBuu1NkefkCN6+j7znnn322Sn9+c9/3pWtXr3a5aOI23g0m7d6eWsQty7ZPQG7n8CtVPY2i/Axcj4izduntWimR/Boqx+383kLe++993Z59AjNW/Bl37sVqWfbuCPqnUf1wGYYfDbXA8ds0fF2nFNjx451ZXgkfNWqVa6Mt/AjMwyaa4pM/ljOc4bzOG95fk+cODGleZ2IzH6RtKKR5kGG7x09C9e2CRMmuDLuB8wPHTrUlbF5qV7wGY1c5/C+vEbjexe5YUHZA/9daBY8rqK279GjR82yKq6A2JSHRFEitmSc55gjPKZ4PuPc579pVaQCHaGdLCGEEEKIDOgjSwghhBAiA/rIEkIIIYTIwDYfddZ5VCGEEEKIrRjtZAkhhBBCZEAfWUIIIYQQGdBHlhBCCCFEBvSRJYQQQgiRAX1kCSGEEEJkQB9ZQgghhBAZ0EeWEEIIIUQG9JElhBBCCJEBfWQJIYQQQmTg/wHAhzAqxmSM5AAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "tf.random.set_seed(42)\n",
        "dropout = tf.keras.layers.Dropout(0.5)\n",
        "plot_reconstructions(model=dropout_ae, images=dropout(X_valid, training=True))\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hGT0IRlthkgs"
      },
      "source": [
        "If you want, you can try replacing the `Dropout` layer with `tf.keras.layers.GaussianNoise(0.2)`."
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.11.3"
    },
    "nav_menu": {
      "height": "381px",
      "width": "453px"
    },
    "toc": {
      "navigate_menu": true,
      "number_sections": true,
      "sideBar": true,
      "threshold": 6,
      "toc_cell": false,
      "toc_section_display": "block",
      "toc_window_display": false
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}