03_mnist_classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "DpII-qgno6EX"
   },
   "source": [
    "# First MindSpore Neural Network - MNIST Classifier\n",
    "\n",
    "Make Your First GAN With MindSpore, 2022"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7bkrfSHHAPUB"
   },
   "outputs": [],
   "source": [
    "# import libraries\n",
    "\n",
    "import mindspore\n",
    "import mindspore.nn as nn\n",
    "import mindspore.dataset as ds\n",
    "\n",
    "import pandas\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.6.1 Ascend\n"
     ]
    }
   ],
   "source": [
    "print(mindspore.__version__, mindspore.context.get_context('device_target'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "o91c-ALVmw0n"
   },
   "source": [
    "## DataSet Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "VImRycohm3Ct"
   },
   "outputs": [],
   "source": [
    "# dataset class\n",
    "\n",
    "class MnistDataset:\n",
    "    \n",
    "    def __init__(self, csv_file):\n",
    "        self.data_df = pandas.read_csv(csv_file, header=None)\n",
    "    \n",
    "    def __len__(self):\n",
    "        return len(self.data_df)\n",
    "    \n",
    "    def __getitem__(self, index):\n",
    "        # image target (label)\n",
    "        label = self.data_df.iloc[index,0]\n",
    "        target = np.zeros((10), dtype=np.float32)\n",
    "        target[label] = 1.0\n",
    "        \n",
    "        # image data, normalised from 0-255 to 0-1\n",
    "        image_values = np.array(self.data_df.iloc[index,1:].values, dtype=np.float32) / 255.0\n",
    "        \n",
    "        # return image data tensor and target tensor\n",
    "        return image_values, target\n",
    "    \n",
    "    def plot_image(self, index):\n",
    "        img = self.data_df.iloc[index,1:].values.reshape(28,28)\n",
    "        plt.title(\"label = \" + str(self.data_df.iloc[index,0]))\n",
    "        plt.imshow(img, interpolation='none', cmap='Blues')\n",
    "    \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "AJ_E9wgYQybx"
   },
   "source": [
    "## Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wioFGZ4mKUqj"
   },
   "outputs": [],
   "source": [
    "mnist_train_dataset = MnistDataset( './dataset/mnist/mnist_train.csv')\n",
    "\n",
    "batch_size = 16\n",
    "train_dataset = ds.GeneratorDataset(mnist_train_dataset, ['image', 'target']).batch(batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 281
    },
    "colab_type": "code",
    "id": "Bnq2wNUhLZRg",
    "outputId": "b5b64d42-aacc-403b-e599-f73ccbaa6e33"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAEICAYAAACZA4KlAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8+yak3AAAACXBIWXMAAAsTAAALEwEAmpwYAAAQEUlEQVR4nO3de5CV9X3H8ffHW4x37K4EkYJVHKuMQWeLOlqlmnih46B2aqSNhcYGL9iYGDOx9qK9TKq2onZitChWrNZLqgy0pVFDqiY2WhdFLqHxNotCueyWKDCGKvjtH/sQl3XP7xzOZc9Zfp/XzJl99vk+l+8e/fCcc57nOT9FBGa269ut2Q2Y2eBw2M0y4bCbZcJhN8uEw26WCYfdLBMO+xAkqUvS5ypcNiQdWeV+ql7XWo/Dbk0haaykLZIebHYvuXDYrVnuBF5qdhM5cdiHOEkTJP1Y0ruS1kj6tqS9+i02SdJbknok/Y2k3fqs/yVJKyT9TNKTkkYPQs8XA+8CCxu9L/uYwz70bQO+BrQBJwNnAlf2W+YCoAM4AZgMfAlA0mTgeuBCoB34IfBwJTuV9J3iH5iBHksS6x0A/AVwzU78jVYHDvsQFxGLIuKFiNgaEV3A3wOn91vs5ojYEBFvA7cDU4r5lwN/HRErImIr8C1gfCVH94i4MiIOKvE4LrHqXwKzI2LVTv6pVqM9mt2A1UbSUcBMeo/c+9D733RRv8Xe6TO9Eji0mB4N3CHp1r6bBEYWy9W71/HA54Dj671tK89H9qHvLuC/gbERcQC9L8vVb5lRfaZ/GfifYvod4LJ+R+VPR8R/ltuppLslbS7xWF5itYnAGOBtSWuBa4HfkvRypX+sVc9hH/r2BzYCmyUdDVwxwDLfkDRM0ijgauDRYv7dwB9JOhZA0oGSfruSnUbE5RGxX4nHsSVWmwUcAYwvHncD/wacXdFfajVx2Ie+a4HfATYB9/BxkPuaR+9L+8X0hms2QETMBW4GHpG0EVgGnNuoRiPi/YhYu/0BbAa2RER3o/ZpH5O/vMIsDz6ym2XCYTfLhMNulgmH3SwTg3pRTVtbW4wePWYwd2mWlZUru+jp6el/nQVQY9glnQPcAewO3BsRN6WWHz16DM+/2FnLLs0s4ZQTO0rWqn4ZL2l3em9TPBc4Bpgi6Zhqt2dmjVXLe/YJwBsR8VZEfAA8Qu8dVWbWgmoJ+0h2vMFiVTFvB5KmS+qU1Nnd4wulzJql4Z/GR8SsiOiIiI72tvZG787MSqgl7KvZ8W6qw4p5ZtaCagn7S8BYSYcXX4N0MTC/Pm2ZWb1VfeotIrZKugp4kt5Tb/dFRKn7mM2syWo6zx4RC4AFderFzBrIl8uaZcJhN8uEw26WCYfdLBMOu1kmHHazTDjsZplw2M0y4bCbZcJhN8uEw26WCYfdLBMOu1kmHHazTDjsZplw2M0y4bCbZcJhN8uEw26WCYfdLBMOu1kmBnXIZrOhYtx1/56sR0SyvvzmSfVspy58ZDfLhMNulgmH3SwTDrtZJhx2s0w47GaZcNjNMuHz7JalaQ+9kqyvfv7ZZP2MqRfWs51BUVPYJXUBm4BtwNaI6KhHU2ZWf/U4sv9GRPTUYTtm1kB+z26WiVrDHsBTkhZJmj7QApKmS+qU1Nnd013j7sysWrWG/dSIOAE4F5gh6bT+C0TErIjoiIiO9rb2GndnZtWqKewRsbr4uR6YC0yoR1NmVn9Vh13SvpL23z4NnAUsq1djZlZftXwaPxyYK2n7dv4pIr5Xl67M6uCK7y4pWZt396Pplff8VLI87aTDqmmpqaoOe0S8BXy2jr2YWQP51JtZJhx2s0w47GaZcNjNMuGwm2XCt7jaLuuZ/3qndPHDLcl1R/36xGT9vHGHVtFRc/nIbpYJh90sEw67WSYcdrNMOOxmmXDYzTLhsJtlwufZd3GLu95N1qfd+2Ky/uz1ZyTrB+6z5862VDdzl6xK1tcuLf31CnsceXxy3ce/cmpVPbUyH9nNMuGwm2XCYTfLhMNulgmH3SwTDrtZJhx2s0z4PPsu7qw/nZ+sf/jaomS968snJuufHX3QzrZUN1felh5Wmf8tfT/7gzOnJlcd+5n9qmmppfnIbpYJh90sEw67WSYcdrNMOOxmmXDYzTLhsJtlwufZd3F777N3sv5h75DbJW3Z+lE929kpr63ZlKxvWflaegO77V6y9P7WrdW0NKSVPbJLuk/SeknL+sw7WNLTkl4vfg5rbJtmVqtKXsbfD5zTb951wMKIGAssLH43sxZWNuwR8Rywod/sycCcYnoOcH592zKzeqv2A7rhEbGmmF4LDC+1oKTpkjoldXb3dFe5OzOrVc2fxkdEAJGoz4qIjojoaG9rr3V3ZlalasO+TtIIgOLn+vq1ZGaNUG3Y5wPb7xGcCsyrTztm1ihlz7NLehiYCLRJWgXcANwEPCbpUmAlcFEjm7S0q+cuL1nbtOSF5Lp7/+qvJetHNfC+7p9/sC1Zv+KRV9IbeP+9ZLnt5NLfeX/esUNvfPValQ17REwpUTqzzr2YWQP5clmzTDjsZplw2M0y4bCbZcJhN8uEb3EdAta9tyVZf2D2U6WLe+yVXHfBDb+ZrA/bN71+Lb74QPprrF9+9In0BkYenSy/fvvknW1pl+Yju1kmHHazTDjsZplw2M0y4bCbZcJhN8uEw26WCZ9nbwFvrtucrJ/4tX9Ob6BnZcnS5D9MD018/JiD0tuu0bcWlv665x/8w3dr2vYt3zi7pvVz4yO7WSYcdrNMOOxmmXDYzTLhsJtlwmE3y4TDbpYJn2evg63b0sMaP/bqO8n6jCtmpnfwUforl1NDE//wha7kqn9y8KeT9T/7/FHJ+saff5is3zt3WelipJ+3U6Z9IVn/8kmHJ+u2Ix/ZzTLhsJtlwmE3y4TDbpYJh90sEw67WSYcdrNM+Dx7HcxdujpZn3HlbekNSOl64jw6wB5jTyhZ2/DSs8l17yxTf3jB6cn6hlVrk3XW/LR0rX1MctV/vfzk9LZtp5Q9sku6T9J6Scv6zLtR0mpJi4vHpMa2aWa1quRl/P3AOQPMvy0ixhePBfVty8zqrWzYI+I5YMMg9GJmDVTLB3RXSVpSvMwfVmohSdMldUrq7O7prmF3ZlaLasN+F3AEMB5YA9xaasGImBURHRHR0d7WXuXuzKxWVYU9ItZFxLaI+Ai4B5hQ37bMrN6qCrukEX1+vQBI3MdoZq2g7Hl2SQ8DE4E2SauAG4CJksYDAXQBlzWuxdYwL3EuffoVZc6j77l3un7QZ5Ll7/9d+rvf99+79H/GC+84JLnu6meeTNbLnacnIl1PXUPQ83Zy1WEX3JWsL509LVk/rMy9+rkpG/aImDLA7NkN6MXMGsiXy5plwmE3y4TDbpYJh90sEw67WSZ8i2uFrn/w1ZI1jR6XXPeOa89M1i/pGF1VT5X4l2tOS9bP/r/0V0F3//gH9WxnR2W+SvrEM49L1n1qbef4yG6WCYfdLBMOu1kmHHazTDjsZplw2M0y4bCbZcLn2St06aTSQxf/7vgzkusOP7DMLa4N9N77Zc6jL1tS0/YfmP3NZH3cIQdWve1DDvhU1evaJ/nIbpYJh90sEw67WSYcdrNMOOxmmXDYzTLhsJtlwufZK3TN6Uc2u4WSNm/ZWrJ28zNvpFfemB6Sa48jSw8HDXDeuEPT27eW4SO7WSYcdrNMOOxmmXDYzTLhsJtlwmE3y4TDbpaJSoZsHgU8AAynd4jmWRFxh6SDgUeBMfQO23xRRPysca1aKbc882bJ2vfufjC98iGHJ8vLv/OFalqyFlTJkX0r8PWIOAY4CZgh6RjgOmBhRIwFFha/m1mLKhv2iFgTES8X05uAFcBIYDIwp1hsDnB+g3o0szrYqffsksYAxwMvAsMjYk1RWkvvy3wza1EVh13SfsDjwFcjYmPfWkQEve/nB1pvuqROSZ3dPenrsM2scSoKu6Q96Q36QxHxRDF7naQRRX0EsH6gdSNiVkR0RERHe1t7PXo2syqUDbskAbOBFRExs09pPjC1mJ4KzKt/e2ZWL5Xc4noKcAmwVNLiYt71wE3AY5IuBVYCFzWkQ2PNu1uS9W/f/3zpotL/nv/+H5yVrB/SxK/BtvoqG/aI+BGgEuX0wONm1jJ8BZ1ZJhx2s0w47GaZcNjNMuGwm2XCYTfLhL9Kegg4bsajyXp0lR52+eTfS1/+MHPyMVX1ZEOPj+xmmXDYzTLhsJtlwmE3y4TDbpYJh90sEw67WSZ8nn0IuOyLJybrd/75KyVrX5mY/qpoy4eP7GaZcNjNMuGwm2XCYTfLhMNulgmH3SwTDrtZJnyefQj4q3OPLlO/fXAasSHNR3azTDjsZplw2M0y4bCbZcJhN8uEw26WCYfdLBNlwy5plKT/kPQTScslXV3Mv1HSakmLi8ekxrdrZtWq5KKarcDXI+JlSfsDiyQ9XdRui4i/bVx7ZlYvZcMeEWuANcX0JkkrgJGNbszM6mun3rNLGgMcD7xYzLpK0hJJ90kaVmKd6ZI6JXV293TX1q2ZVa3isEvaD3gc+GpEbATuAo4AxtN75L91oPUiYlZEdERER3tbe+0dm1lVKgq7pD3pDfpDEfEEQESsi4htEfERcA8woXFtmlmtKvk0XsBsYEVEzOwzf0SfxS4AltW/PTOrl0o+jT8FuARYKmlxMe96YIqk8UAAXcBlDejPzOqkkk/jfwRogNKC+rdjZo3iK+jMMuGwm2XCYTfLhMNulgmH3SwTDrtZJhx2s0w47GaZcNjNMuGwm2XCYTfLhMNulgmH3SwTDrtZJhQRg7czqRtY2WdWG9AzaA3snFbtrVX7AvdWrXr2NjoiBvz+t0EN+yd2LnVGREfTGkho1d5atS9wb9UarN78Mt4sEw67WSaaHfZZTd5/Sqv21qp9gXur1qD01tT37GY2eJp9ZDezQeKwm2WiKWGXdI6kn0p6Q9J1zeihFEldkpYWw1B3NrmX+yStl7Ssz7yDJT0t6fXi54Bj7DWpt5YYxjsxzHhTn7tmD38+6O/ZJe0OvAZ8HlgFvARMiYifDGojJUjqAjoioukXYEg6DdgMPBAR44p5twAbIuKm4h/KYRHxzRbp7UZgc7OH8S5GKxrRd5hx4HxgGk187hJ9XcQgPG/NOLJPAN6IiLci4gPgEWByE/poeRHxHLCh3+zJwJxieg69/7MMuhK9tYSIWBMRLxfTm4Dtw4w39blL9DUomhH2kcA7fX5fRWuN9x7AU5IWSZre7GYGMDwi1hTTa4HhzWxmAGWH8R5M/YYZb5nnrprhz2vlD+g+6dSIOAE4F5hRvFxtSdH7HqyVzp1WNIz3YBlgmPFfaOZzV+3w57VqRthXA6P6/H5YMa8lRMTq4ud6YC6tNxT1uu0j6BY/1ze5n19opWG8BxpmnBZ47po5/Hkzwv4SMFbS4ZL2Ai4G5jehj0+QtG/xwQmS9gXOovWGop4PTC2mpwLzmtjLDlplGO9Sw4zT5Oeu6cOfR8SgP4BJ9H4i/ybwx83ooURfvwK8WjyWN7s34GF6X9Z9SO9nG5cCvwQsBF4Hvg8c3EK9/SOwFFhCb7BGNKm3U+l9ib4EWFw8JjX7uUv0NSjPmy+XNcuEP6Azy4TDbpYJh90sEw67WSYcdrNMOOxmmXDYzTLx/1sB8NvPNg/WAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# check data contains images\n",
    "\n",
    "mnist_train_dataset.plot_image(9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KG5Xf_6CQ6h7"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.00784314, 0.07058824, 0.18039216, 0.53333336,\n",
       "        0.53333336, 0.95686275, 1.        , 0.94509804, 0.40392157,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.05882353, 0.36862746, 0.6392157 ,\n",
       "        0.99215686, 0.99215686, 0.99215686, 0.99215686, 0.93333334,\n",
       "        0.85490197, 0.8       , 0.13725491, 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.5137255 , 0.99215686, 0.99215686, 0.99215686, 0.99215686,\n",
       "        0.92941177, 0.78431374, 0.22352941, 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.60784316, 0.9647059 , 0.99215686,\n",
       "        0.96862745, 0.42352942, 0.25490198, 0.1764706 , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.8117647 , 0.99215686, 0.99215686, 0.9019608 , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.6156863 , 0.99215686,\n",
       "        0.99215686, 0.49019608, 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.34901962, 0.99215686, 0.98039216, 0.22352941,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.34901962,\n",
       "        0.99215686, 0.96862745, 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.34901962, 0.99215686, 0.96862745,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.34901962, 0.99215686, 0.96862745, 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.08235294, 0.90588236,\n",
       "        0.9764706 , 0.13333334, 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.88235295, 0.99215686, 0.90588236,\n",
       "        0.8352941 , 0.8352941 , 0.48235294, 0.0627451 , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.6745098 , 0.99215686, 0.99215686, 0.99215686, 0.99215686,\n",
       "        0.99215686, 0.74509805, 0.24705882, 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.00784314, 0.45490196,\n",
       "        0.28235295, 0.4862745 , 0.81960785, 0.99215686, 0.99215686,\n",
       "        0.5529412 , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.09803922, 0.85882354, 0.99215686, 0.80784315, 0.01176471,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.40784314,\n",
       "        0.9647059 , 0.99215686, 0.01960784, 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.8352941 , 0.99215686,\n",
       "        0.01960784, 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.10196079, 0.8862745 , 0.99215686, 0.01960784, 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.5176471 , 0.99215686,\n",
       "        0.81960785, 0.01176471, 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.30588236, 0.99215686, 0.3372549 , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        , 0.        ,\n",
       "        0.        , 0.        , 0.        , 0.        ], dtype=float32),\n",
       " array([0., 0., 0., 0., 0., 1., 0., 0., 0., 0.], dtype=float32))"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# check Dataset class can be accessed by index, returns image values and target tensor\n",
    "\n",
    "mnist_train_dataset[100]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "R1VyK4VQpNYA"
   },
   "source": [
    "## Neural Network Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MrgKlTsRAfSk"
   },
   "outputs": [],
   "source": [
    "from datetime import datetime\n",
    "from mindspore.train.callback import Callback\n",
    "\n",
    "# classifier class\n",
    "class Classifier(nn.Cell):\n",
    "    \n",
    "    def __init__(self):\n",
    "        # initialise parent pytorch class\n",
    "        super().__init__()\n",
    "        \n",
    "        # define neural network layers\n",
    "        self.model = nn.SequentialCell(\n",
    "            nn.Dense(784, 200),\n",
    "            nn.Sigmoid(),\n",
    "            nn.Dense(200, 10),\n",
    "            nn.Sigmoid()\n",
    "        )\n",
    "        \n",
    "        # create loss function\n",
    "        loss_fn = nn.MSELoss()\n",
    "        # create optimizer, using simple stochastic gradient descent\n",
    "        optimizer = nn.SGD(self.model.trainable_params(), learning_rate=0.1)\n",
    "\n",
    "        self.model_train = mindspore.Model(self.model, loss_fn=loss_fn, optimizer=optimizer)\n",
    "\n",
    "        # accumulator for progress\n",
    "        self.progress = []\n",
    "    \n",
    "    def construct(self, inputs):\n",
    "        # simply run model\n",
    "        return self.model(inputs)\n",
    "\n",
    "    def train(self, epochs, dataset):\n",
    "        class LossCallback(Callback):\n",
    "            def __init__(self, progress):\n",
    "                super().__init__()\n",
    "                self.progress = progress\n",
    "\n",
    "            def epoch_begin(self, run_context):\n",
    "                cb_params = run_context.original_args()\n",
    "                print(f'{datetime.now()}： start epoch {cb_params.cur_epoch_num} of {epochs}:')\n",
    "            \n",
    "            def step_end(self, run_context):\n",
    "                cb_params = run_context.original_args()\n",
    "                step_num = cb_params.cur_step_num\n",
    "                if step_num % 10 == 0:\n",
    "                    loss = cb_params.net_outputs\n",
    "                    self.progress.append(loss.asnumpy())\n",
    "                if step_num % 1000 == 0:\n",
    "                    print(f'{datetime.now()}: counter = {step_num}')\n",
    "\n",
    "        self.model_train.train(epochs, dataset, callbacks=[LossCallback(self.progress)], dataset_sink_mode=False)\n",
    "\n",
    "    def plot_progress(self):\n",
    "        df = pandas.DataFrame(self.progress, columns=['loss'])\n",
    "        df.plot(ylim=(0, 1.0), figsize=(16,8), alpha=0.1, marker='.', grid=True, yticks=(0, 0.25, 0.5))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "cwGBnd4xSLBL"
   },
   "source": [
    "## Train Neural Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 527
    },
    "colab_type": "code",
    "id": "VLS3pXEevWq9",
    "outputId": "b416bcc9-4268-49a6-d7ad-3a4139228d47"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2022-05-21 16:11:49.080888： start epoch 1 of 8:\n",
      "2022-05-21 16:12:23.994045: counter = 1000\n",
      "2022-05-21 16:12:33.799769: counter = 2000\n",
      "2022-05-21 16:12:43.566110: counter = 3000\n",
      "2022-05-21 16:12:51.001757： start epoch 2 of 8:\n",
      "2022-05-21 16:12:53.489642: counter = 4000\n",
      "2022-05-21 16:13:03.440270: counter = 5000\n",
      "2022-05-21 16:13:13.324890: counter = 6000\n",
      "2022-05-21 16:13:23.245056: counter = 7000\n",
      "2022-05-21 16:13:28.218773： start epoch 3 of 8:\n",
      "2022-05-21 16:13:33.175945: counter = 8000\n",
      "2022-05-21 16:13:43.108555: counter = 9000\n",
      "2022-05-21 16:13:53.040750: counter = 10000\n",
      "2022-05-21 16:14:02.959323: counter = 11000\n",
      "2022-05-21 16:14:05.457163： start epoch 4 of 8:\n",
      "2022-05-21 16:14:12.920380: counter = 12000\n",
      "2022-05-21 16:14:22.816719: counter = 13000\n",
      "2022-05-21 16:14:32.739297: counter = 14000\n",
      "2022-05-21 16:14:42.701727: counter = 15000\n",
      "2022-05-21 16:14:42.703441： start epoch 5 of 8:\n",
      "2022-05-21 16:14:52.640150: counter = 16000\n",
      "2022-05-21 16:15:02.572089: counter = 17000\n",
      "2022-05-21 16:15:12.457931: counter = 18000\n",
      "2022-05-21 16:15:19.887973： start epoch 6 of 8:\n",
      "2022-05-21 16:15:22.352460: counter = 19000\n",
      "2022-05-21 16:15:32.258052: counter = 20000\n",
      "2022-05-21 16:15:42.126624: counter = 21000\n",
      "2022-05-21 16:15:52.037904: counter = 22000\n",
      "2022-05-21 16:15:56.999324： start epoch 7 of 8:\n",
      "2022-05-21 16:16:01.951402: counter = 23000\n",
      "2022-05-21 16:16:11.857559: counter = 24000\n",
      "2022-05-21 16:16:21.795060: counter = 25000\n",
      "2022-05-21 16:16:31.745695: counter = 26000\n",
      "2022-05-21 16:16:34.238597： start epoch 8 of 8:\n",
      "2022-05-21 16:16:41.645516: counter = 27000\n",
      "2022-05-21 16:16:51.485526: counter = 28000\n",
      "2022-05-21 16:17:01.332720: counter = 29000\n",
      "2022-05-21 16:17:11.138300: counter = 30000\n",
      "CPU times: user 5min 45s, sys: 53.7 s, total: 6min 39s\n",
      "Wall time: 5min 22s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "from datetime import datetime\n",
    "\n",
    "# create neural network\n",
    "\n",
    "C = Classifier()\n",
    "\n",
    "# train network on MNIST data set\n",
    "\n",
    "epochs = 8\n",
    "C.train(epochs, train_dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 483
    },
    "colab_type": "code",
    "id": "G046xt9ySheM",
    "outputId": "4641693c-fced-48a5-fdf5-f4d2aadb03f5"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA6gAAAHSCAYAAADhZ+amAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8+yak3AAAACXBIWXMAAAsTAAALEwEAmpwYAACDP0lEQVR4nOz9WYxl253f+X3X2uMZY875jiyyZqlKNciGoRIld7f0pLYgA24BNqoEl2QYkAzYgIBuCGgLsgAZqgc9NaAWDAmqB1kFA3ooA2q1hW4TVWqprGKxyBpIFnmnvDenyJjOuOe1lh/2ybh5k5l5M8nLuofk70NcMCLOPvvsOLkjYv/2f63/MiEERERERERERD5t9tM+ABERERERERFQQBUREREREZEtoYAqIiIiIiIiW0EBVURERERERLaCAqqIiIiIiIhsBQVUERERERER2Qrxp30ATzo8PAyvv/76p30Yz7VerxmNRp/2YcgW0rkhz6PzQ55F54Y8i84NeR6dH/Is235u/M7v/M5pCOHoaY9tXUB9/fXX+eIXv/hpH8ZzfeELX+Dzn//8p30YsoV0bsjz6PyQZ9G5Ic+ic0OeR+eHPMu2nxvGmNvPekxDfEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2wdXNQRUREREREfpC0bcudO3eoquoT2d/Ozg5f+9rXPpF9fSfyPOfWrVskSfLCz1FAFRERERER+RTduXOHyWTC66+/jjHmO97fcrlkMpl8Akf27QshcHZ2xp07d3jjjTde+Hka4isiIiIiIvIpqqqKg4ODTyScbgtjDAcHBy9dFVZAFRERERER+ZR9P4XTR76d70kBVURERERE5AfceDz+tA8BUEAVERERERGRLaGAKiIiIiIi8j2m6TzruqPp/Ce63xACf/tv/21+4id+gp/8yZ/k137t1wC4f/8+v/ALv8BP/dRP8RM/8RP85m/+Js45fumXfuly23/0j/7Rd/z66uIrIiIiIiKyJRZVS+fCc7dpneferCQEMAZu7A5Iog9rj6uipbXN5edxZJjmL7bUy7/6V/+KL3/5y3zlK1/h9PSUn/u5n+MXfuEX+Bf/4l/wF/7CX+Dv/J2/g3OOoij48pe/zN27d/mDP/gDAGaz2ct/w09QBVVEREREROR7SOs8IcAoiwih//yT8u/+3b/jr/7Vv0oURVy9epU/+2f/LL/927/Nz/3cz/HP/tk/4+/+3b/L7//+7zOZTHjzzTd55513+Ft/62/xb/7Nv2E6nX7Hr6+AKiIiIiIisiWmecL+KH3uf1cmOXvDhCSy7A0Trkzyjzy+N/zoPl60evo8v/ALv8Bv/MZvcPPmTX7pl36JX/3VX2Vvb4+vfOUrfP7zn+cf/+N/zC//8i9/x6+jgCoiIiIiIvI9JI0tN/eGXJ3m3NwbksafXKz7M3/mz/Brv/ZrOOc4OTnhN37jN/j5n/95bt++zdWrV/nrf/2v88u//Mt86Utf4vT0FO89f+Wv/BX+/t//+3zpS1/6jl9fc1BFRERERES+x6Sx/USD6SN/+S//Zf7Df/gP/Mk/+ScxxvAP/+E/5Nq1a/zzf/7P+ZVf+RWSJGE8HvOrv/qr3L17l7/21/4a3vdDjP/BP/gH3/HrK6CKiIiIiIj8gFutVgAYY/iVX/kVfuVXfuUjj//iL/4iv/iLv/gtz/skqqaP0xBfERERERER2QoKqCIiIiIiIrIVFFBFRERERERkKyigioiIiIiIfMpCCJ/2IXzivp3vSQFVRERERETkU5TnOWdnZ99XITWEwNnZGXmev9Tz1MVXRERERETkU3Tr1i3u3LnDycnJJ7K/qqpeOhh+N+R5zq1bt17qOQqoIiIiIiIin6IkSXjjjTc+sf194Qtf4Kd/+qc/sf39cdIQXxEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRraCAKiIiIiIiIltBAVVERERERES2ggKqiIiIiIiIbAUFVBEREREREdkKCqgiIiIiIiKyFRRQRUREREREZCsooIqIiIiIiMhWUEAVERERERGRrfBCAdUY8xeNMX9kjHnLGPNfPuXxXzLGnBhjvrz575cfe+wXjTHf3Pz3i5/kwYuIiIiIiMj3j/jjNjDGRMB/A/ynwB3gt40xvx5C+OoTm/5aCOFvPvHcfeD/CvwsEIDf2Tz34hM5ehEREREREfm+8SIV1J8H3gohvBNCaIB/CfznL7j/vwD82xDC+SaU/lvgL357hyoiIiIiIiLfzz62ggrcBD547PM7wJ9+ynZ/xRjzC8A3gP9zCOGDZzz35pNPNMb8DeBvAFy9epUvfOELL3Twn5bVarX1xyifDp0b8jw6P+RZdG7Is+jckOfR+SHP8r18brxIQH0R/2/g/xlCqI0x/wfgnwN//kWfHEL4J8A/AfjZn/3Z8PnPf/4TOqzvji984Qts+zHKp0PnhjyPzg95Fp0b8iw6N+R5dH7Is3wvnxsvMsT3LvDKY5/f2nztUgjhLIRQbz79fwA/86LPFREREREREYEXC6i/DXzWGPOGMSYF/gvg1x/fwBhz/bFP/xLwtc3H/z3wnxlj9owxe8B/tvmaiIiIiIiIyEd87BDfEEJnjPmb9MEyAv5pCOEPjTF/D/hiCOHXgf+TMeYvAR1wDvzS5rnnxpj/G33IBfh7IYTz78L3ISIiIiIiIt/jXmgOagjhXwP/+omv/dePffxfAf/VM577T4F/+h0co4iIiIiIiPwAeJEhviIiIiIiIiLfdQqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrvFBANcb8RWPMHxlj3jLG/JdPefz/Yoz5qjHm94wx/4Mx5rXHHnPGmC9v/vv1T/LgRURERERE5PtH/HEbGGMi4L8B/lPgDvDbxphfDyF89bHNfhf42RBCYYz5PwL/EPjfbB4rQwg/9cketoiIiIiIiHy/eZEK6s8Db4UQ3gkhNMC/BP7zxzcIIfx/QwjF5tPfAm59socpIiIiIiIi3+9eJKDeBD547PM7m689y/8e+O8e+zw3xnzRGPNbxpj/1csfooiIiIiIiPwgMCGE529gzP8a+IshhF/efP6/A/50COFvPmXb/y3wN4E/G0KoN1+7GUK4a4x5E/gfgf9lCOHtJ573N4C/AXD16tWf+Zf/8l9+59/Zd9FqtWI8Hn/ahyFbSOeGPI/OD3kWnRvyLDo35Hl0fsizbPu58ef+3J/7nRDCzz7tsY+dgwrcBV557PNbm699hDHmPwH+Do+FU4AQwt3N/79jjPkC8NPARwJqCOGfAP8E4Gd/9mfD5z//+Rc4rE/PF77wBbb9GOXToXNDnkfnhzyLzg15Fp0b8jw6P+RZvpfPjRcZ4vvbwGeNMW8YY1LgvwA+0o3XGPPTwH8L/KUQwsPHvr5njMk2Hx8C/wvg8eZKIiIiIiIiIsALVFBDCJ0x5m8C/z0QAf80hPCHxpi/B3wxhPDrwK8AY+D/ZYwBeD+E8JeAHwX+W2OMpw/D//cnuv+KiIiIiIiIAC82xJcQwr8G/vUTX/uvH/v4P3nG8/498JPfyQGKiIiIiIjID4YXGeIrIiIiIiIi8l2ngCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKCKiIiIiIjIVlBAFRERERERka2ggCoiIiIiIiJbQQFVREREREREtoICqoiIiIiIiGwFBVQRERERERHZCgqoIiIiIiIishUUUEVERERERGQrKKC+pKbzVF2g6fynfSgiIiIiIiLfVxRQX0LTeb5+f8HdpefuRaGQKiIiIiIi8glSQH0JrfM4H0hjCJvPRURERERE5JOhgPoSkshiLVQdmM3nIiIiIiIi8slQwnoJaWy5sTtgNzXc3BuSxnr7REREREREPilKWC8pjSKy2CicioiIiIiIfMKUsr4N4dM+ABERERERke9DCqgvy3zaByAiIiIiIvL9SQH1JRkDQTVUERERERGRT5wCqoiIiIiIiGwFBdSXZECTUEVERERERL4LFFBFRERERERkKyigviRjjAqoIiIiIiIi3wUKqCIiIiIiIrIVFFBfUtt56i5Qt+7TPhQREREREZHvKwqoL6HpPPdnJbM6cHdW0HT+0z4kERERERGR7xsKqC+hdR4M5LHB+83nIiIiIiIi8olQQH0JSWSxxlB1gUAgifT2iYiIiIiIfFKUsF5CGltu7A4YJ7CTp5/24YiIiIiIiHxfUUB9SdYYig7Oy5q7F5qHKiIiIiIi8klRQH1Jznt8gEESE9A8VBERERERkU+KAupLypOIEAIni5rOec1DFRERERER+YTEn/YBfK8xxtA4z1lREUf5p304IiIiIiIi3zcUUF/SrGg4XkPycM2X3p1x53zNwTgjjWKu7eTsjlKSyJLGlqbztI9VWU+WFfOyIbGWJLYMk5hhFmMMhADGQNN6irYjsfbysUfPf7SvNH561XZVddyfF7Rd4PrOgL3xRxs5NZ1nXXcAjLL4mft5tO3Hvd7L+qT3+bT9Pes1vp3Xft7+Q/hwePfz3stV1VE0HcO03+bx8+GTOk4RERERke8XCqgvqXWeOnT8x3dOeP+04P/zVUdiE67sZoySiJ967YBhZtlNYs6rDkwgAFUdeOdkwbrqKFrP/iBlksXsTBMmg4TlumM8iLl/URBFFucDkyxif5zjCcTGEtmAB64M001A6pe6SSLDumr5+vGaO4sSXGBvlPHTr0zJkwjnAyHAedkyK1piaxkPIg4GKXFkiQw0zhNZgzWGtnM8XLfEkSGyME1j4sgSW0PrPMYYYmtx3hPg8uvOByJrSSJL5x0hGGJrcMHTucCs7rBA5T1XhxlZHG32YYgt/XFu9ud8wIc+nDvvaV24PFYfAq0LrGpHGzzBBA4HKd57Hq46kgQiY5imCZE1JJHlZF33NwIITJLkcl/9sYG1hjSyl8dDCMyrjsj2Nw+y2GINrFoPwXGy6IgiQ5padrKY6zsDJoOEt89rsrdOcM5TNR3vnBVEsaFzgVEWczTKSCND1Tl8CKRRxJVpX4lflC0nqxoDjLOUm3s587Kh6wKDLKKoO+rWsT8ecDBOKdqOtvMM082PsYFhEoMBArTeUzQdO3nKMI1p/YfbFk1/EyRN7OVNkSefszNMaZ2/DOL9uWYvb5o8eRPm8UC9qrrLmzGP32hJY/uRx3ZH6UvdMJgVDa3z7AxSxnl8+fWPC/5PbqfwLyIiIrKdFFBf0s4gJTeWzhuGecxF4SjrhlXT0TYdv3v7AmsMDo8FgjcEuwlexhJFhsYFYgIhGJzz2AicB2stBE8cRwQf6Jwnjiw+9OuuRqaf/+rxGMB5Q2QDAMEYQoDYQmwtdeP4736Pzb4N1gYMEJkIQqDDYzb/C/Shk9B/HAAXDJkBR4BnbPf4xzxxPE9uF4IBPAZL5wPWBIzpuyKHp+zv0cds9t09tm9rDD4YgvHEWHwAv9nOe0isxW2+P2ssxgaMN/1733kC4fL7edZxYwKJtWDAu37LgMFaT4TFEYiMJYsMXeiXIEoiw3zeMXrv93Cd3wTuQBbHtM4RjGF3mBCACIOxhrbzRMbQ4mmbQNc5bBwxHUSsa0ccWeqmAwKtA2zEJDfEJmKYWeo2MMojug6ORil1F9gdJ1SVY1E3WGMYxBHDYcy13QGu8aRJTJYZHlxUHEwy7s9Kgodl2bLuWjIbMR1mvLKb43xgVrdEBg4mOcYafuTKhINxxiiL+ebDBW0He8OEnUFC6zxF43jvbM266qg7x94o4/pOxiRP2RkkfOWDC5a1J0sMP3Zth9cORySR/UhFGrisVD8aYfDu6ZIv3b4gjix7g5gfu7HLziDleFmxrjuyxDAvOwZxxN4o5ebe8DJEr+uO43lFEhmsNZePfZzHK+CPArGIiIiIfPfoiusljfOYn7wSkbgxX20di1UNkaX1jiSJsNbgQ4AQAdAFT2ojQmDT9Tf0V9uRxRIINsIQsBZMCGAsnQtgwMQRnoA1li4EvIE4MjgfYQFr+qt3HwKpjTBA2Toa68jTiMb3r2lN2ARBcAb6AqHB2ghLoPWG2EaYEGiDJTIGA7QECIbU9iHZBYO1FhPAB0tsLWbzfIwBIFhDCIHY9kHYBdtXZgO0oY+FNjZYLC54zGY7HwzR5fFs9h0CbfjWfRsbERNonKUzffUzBNOHy81xG9PvjxDofB8+Gx/AhMuv+6fs+9H3F0KgCeAeqxjbEOi8ofEOaywtnrLtK7tN52h9oGmgK1t84HL7ZdMQbyqIdy8qQvBEUQQE8AFv+pKnMQa8wVUts9LinGOQxDgDbev7iq3xHFeBzm8q3ra/uWEwvB2Bc4E06QO88YHIGCoX8M6RxjGdgRxIEkPdBToXcCaQGIsxgYBlOoyp78/5reCxxhI252dkYwZpzO9OM3ZHGbVzOBfAw7rpaLxjWTsGcX/u7w1Tqs7RdB0Hw5QkTRmnlnXdsT8ZMEgMX+kcJ4s1t2cVReu4Msq4sTOgcg7Xee4tal7dG2CM4aJocR664Pjd2ytO1w2RCYRgubKT82BR4pxnOsxI4+gy8N4+W3G2ajleVNzYzRmmMeu6I42fvpbxo0BbNB2/f2dGwJBHhh+/ufvUiu/jw74fH67/okPzX7QC/DwvWx1WNVlERES2lQLqt+HaOOXP//Rn+cN7F3zt3pK683z9/gXr1nF8UdI66ELAWoMNljSyJNYwHicMo4hhFnO+rqk7T9s6mi5cBtFBbi8vcF3rqbu+8mmCx5k+tBH6i+EQ+gASYbAYxrnl+l6GxbBqHYtVg3MBFzZDb2NDvAnGjevDSQiBZBNmIJD02RgIRFE/rLVzAXu5Xf+YDTz2HAOB/nV8n4A/sj/ffz9pZIF+uGu/jw+3s08ew8fs2xgYRQa/CY6t+7DyaaOIaHPcPgRiLIPEXgbVfvhw/5rfuu/++wtYUmuwcYQP0Hb9G5Nbg40TAGIbCB4chjZ44s0NAEsfUpIoIkv6zyNraANk1mN8jLHQBYPZ3GSA/niTCFoTkVrwNqbuPNYaEmOIgqH2AUM/XLhsO4Lvg6U1m2o8hs6Zzc2B/n1KrMHZiNZ7bDBUBtZFR57GRJG5rBCbAM53nC46oK80GhwRBhcCgZqm7Xi4KgjBMkksNrZ0XaDsHHXnSCPLMLE0Du6cr2k6j7GW96OSQRYxymOCh+G8JDKWnTzh35YNq7rDh35Y+94gIzaGJjiWTUeMZZREpHFElka44Clrx7rteP98jTGGSRKRxjGTcQzHa5ZlQxL1Ny/ePy8YJBHrpuWbxy0H4/QjHbgfr5Cuqo6vfHDOvXnNumpYVC2fuTLh7rxlMlixbvKPVF+bzvPu6Yqmc9yblUzymNN1y9Vpxt6gH4pv+9OMo0l+eQ4+0nSeOxdrysYTR1DUjmme4ELgaJJ/y5DqJz0K0984XpJGlt1h8rHV4abzvHe2wvlAHkcvXE0WERER+eOggPpturKTc2XnOj/3xhFF09F2r3K6rrhY1pdzPh+vpkTWMhkkTAcJGCjqjqJxOOfpHpu/macRcWRoXfiWxyBc7vvxeZn9nE1Plibc3BuAgeN5SVW7j2yXJ/2+ATrnqVu3mf9pHpsL+uHHedJX+T5uu0cfP3k83+n+Pm7fj/b3+Hvy5Ou0LrBsO+JN5Xgvz15o34R+aGsWWTrvGcT9j0rpPHjHybrlaJQxrx1l4zhbVRR1x4Pj+7z5yg0meUIUBWoX2B9ndLWn6Tz31iVV5fB+E9gtNJ0jS2KGsWF/kjIvHKfrmlESUzeOawf9HNVhGkOAD87WlFXg4argaJyz7vr5pBdFTZpY5mXDNE+w1lC7gAl9CCvrjg5DHkPwEXEU0TSe6SiBAHkckccRVeep2r6ZVtM5kijqA/Vm2LJpDVls8AHWZYsJgSy2fSXYBxaVY5BuAmVkiSJL5xxF1VFWjjyLSJO++ntWNhxfFDj6ucDnK3gQF1jTH5/zHmsgSSL2hyl7pLiuo3CB987W1J3ncJzw/qomjiLGlcW1sKhavvjOBVd3coq66+eFjzIs/c2j02XN/VnBqvGYEMgTy+euTbk3L/md9y9Yl46i7Su2i7JjkFpe2x8wL1t2h331tek8Z6uK26dLHi5b5kXN/jglYDgapxSN4+GyYpBYyrYPot84b3n1wZxRlrAz6Of43puVhBCo2v5mRJZG3JsVrKqOB4uSq9MBeWK5sTv8luHPdy4KTlc1Z6uG1w6Gl2szP29+cOs886JlkMY0rp/Xuzt88bnA3w2q6IqIiMgjCqjfoXH+4dy0m/vDT/loPvTZq9NP+xC2xrd78ftxXXwf3YB41EiodZ5/9z+d8+f+zOcYZvHl1x41F+rnZ/ZDRwmbwGk+bFi0M0jBcLld6zw7eUryWPffURYzWzfcX5Qk1vSv03la34/nLtqOk3lFZCOS2HB1mlO0HV0XaJ3jvbOCxnny2JLFlr1ByihPCARmRUsaGY5XNQ9nFSFYhrllmidUrePhouLhomZRNZyuaorO88okI88ivIeqcxzkKY337I0yVlVL0TpiDGdFQ+YCcWLIY8O68pR4Ou9JYkvkDZHpRwdUrq/cZiHgg4NgqVpPcJ7GObo24Ag0zhFbw8mypmk9OyPL2bwjSywXVcvFumFeNZRdBw6OpjnOezoCFsNOnlK1jpt7AxZ1xxffP8MYy+m8Yd22FG3H0SCjbRw2gv+hPObazpCLVc2t/SEXRcvFquQ33zoDAmUbmBUNk0FKoB+h8GBekVpDlsZEFn7nfss3/se3+OHrE948HGMNfO3BimXdMMkSpnlC5z3L2pFFEfcuSurWU3eequmH7heN64NqgEXdMspizlb9v8vOICGJ+nD67smKedmSxZbPXJlc/p6qW8esaLhYVwQMzc6Qdd1xc6///fXdDIpPm9PbdJ4PztebmQ8vPj/4e5kCuYiIyLMpoMr3vTT+9i4Cn/a8j9vX1XHMlZ1nr4+7N3r6vMeX0Vfvn/0az7v4/dPPWWro8fB9GZAHfWXt0ZzMe7OSsum4f7HmrOiYZil16NgfpLyyP2KYRrx3uqZzgcp1/PDVKWXruHdecLxquH1aULuOqnYMsoh3Hq7YncYcDoaMsoiTRc1pVRNFAdcFkjhmGFtmZUNsI1znORxnAKySiP1RRtN1LBvHMO2HzlvgZF72NwPqth8GH1n8strM7za0ruNuKGhazx89mDNIE65MM9aNgwBZYtkbpowHCXXbV7gXZYvza946XnBzf8iibMkTy6poMZGh3DSzurE/5HxV88FFifMwSiMOrOGidtQOYue5N6vpvIdg6BxUtWcQe+ZVTX3SD/dNDLQByq5jse5YVA1vnzQsqo79ccobhyPKpqPpHPujlKs7OUeTnHXdd0n+g7tz1nVH6wPrpuNP3NoD4DffOqWqW+7Na/bHCS7Aq/t9SL0oaloXSCPL0SS/vMkSW8M4Tz52OafnWVUd//HdU8rGM8kj/tRrB4zz/kbOxbohT2OGaXRZAf5u2IZg2HSeuxfF5XH8IARyERGRl6GAKvJ95nkhun/s6SH58ec92bE2jVP2RullaAmvHXBvVlA0DgMfqdC9ejD+SJWs6TxXpwPWTcdP3mjIU8uX3p9xuqoZxDE//sqE1/cnHE0zbp+seOtkTdsF2tDx2t6IB4ua22dr4sgyK2vGacokT0gT+OzRBEPg4bLlndM1e8OELI64fb7G+8DpumYnSwl4qnYz99gE2tbT+sB0EOMDBDwfnBUksWFnkLA/yNkdRyQ2AgN123FRdjRdRxRFNK3neFUySROKzhE6GMQRPgTunBecLWs6Hyg7T930jbdc8KxboGx4t2x4OI/BBPIkZlU7iqbl1t6As7oiSyLqrh/mvyih9YH3zgs+OKso2463TwLn64advL/JMEz7oPfl98+5Ny+xGD64KMjjiEXZ8FbcV3QPxylV2zIZpCTLhnXlKKo166rBu0DpHKM0wXvH1+7P+eBszcNlzc4w4Uc3HZer1vEHd2ccjvOPnfP6eCAsmo5V6dgZJbTec7qqKJqYxFqC6QN+uum0nUSffEhtOs/v3Z2RRZZxFn9iwfBlQ2/rPGXrNt3H3Xc1kIuIiHwvUkAVkRf2eIh9/XD81Avzx4e9P3rOzb1hv+3VfttX98f84b0ZRe2YDGIOJxmvHYx57WDMn3i1oWg61pUjjg2vlS0/emPKunTkKcRRxNWdweVyQdenA+ZVw52Lgmme8PV7MzoPdeuZDlJu7A04XfTry9adIzIG52DWNHSbJlR5HFG0HfvDjCQy7IxiPnt1jPOe9049s1XDoqo4XwViazld1uRZRGIdoyzqh2znCcuyY1a0dD6wbjyd75hmOa3vK9NxBFenGZHtuyM/WDTEJjBIIrLEcmde4h3sDFM+OF9jN12VD0YZzgeyGM6XDYumY1U02MhydZozSGNun60JwINFzeEooekcy6qhajwn7zd8cFb06/9aSxrVtJ1nb5RSdZ53TwpOVy1t5/ihq1OuTvqh4QHDsmppnGNddZyvS37/3pLIBG6na17dG5PG9rJy+/hQdID3z9cE38/x3h2kmAhm64Y4gq98MGOQxgxiy9XdfkRA03qOlyVZ8Z01b2o6T9M50ji63EfrPMEHTGw+Mlf3O9F0nm8eLzEGBsmLHXOyWee6avuh2o+adYmIiEhPAVVEvi0vM3T6yW2v7OTsjq48dbjxlekmrDyqTB32cyrfP1v3S+1shp9+7cGib2ZVNhxN8sulnH781h6H0wGn64a9QUKWRJRNh3P9vNVV5UgTy/1Zid+sHQyGd0/X5LEliS1XdzKq1nMyq7EGbu6PSBPL+aoljQ3LumWc5BACV3cHNM73HaAHMeu6ZZon7I4N68oxSCyLoqULgdhCFsecFzWNC5jgGQ5iOufZGWScrmpWTcOqari3qLm1O+B8XbMsHHECi1XDsu7nouZZQtN6zlct01EfvldVy4N5xemiH7J8MM65c7FmVlasq4jaO27s9MOpXz0Ycl501I1jWXUUbUfjPMMsYW+Y0HWei7LmwaJinCechIbaey6WNVFkuHdvzgfna94/X3N9836VrefqNOfGbs4gifijBwsGacyVcc6VqeVP3Nrtg2zn+MN7880STR3DMmaQRpwXNdM8xaSGWdH0leGuX415nPfds1vnaVpP0fZzt3c3w+YfX6rnnZMVy6pld5jw2sF4s06xhU2ldpRGHxsMX6Qy+mi++CCNL0Mv8Mxh9I9+Fm7sDlhULePNXPVttA3DoUVE5AeTAqqIfCqeN9z4w8ft5cc/dHVyecHcOg8hsDtKCfQNqy6rtJHltYPxZaA1m/0F+nVir+8MKNqOB7OK8SDmwbxkUXT86PUJ750XHIwyHsxrTlcNJgKcxRpDHkXkicNtOh7HEQT6jsQxESEG0wWKKqJxnsPhgDePEr52f0G1WSd3EMMgi7iZjvChH+q5N0hpnKdsO85WDa3rGCdx30QL8ATS2NB0jmEWMxokeAexMfgYjOmX51lVHWXTVysXdcOo6oPQII2ITN+N+nzdMEoTFkWDtYZl2XG6LDkvWiJr2R+mpNcN3zxZMkoiytoziCOa1tE4x1v3W7DQtY7Ww639EfcvClZlR+c9Nu6HS5+ta85WNd84XjLNY948HHNlkjFMY3wIfHC24r2zNe1xQRwHmtYxrx0RECUlu2nM4STng/M1s7I/th+5OmWUxzRdx5fvzAg+EIzlp2/t0oXAJI+xGIZpxP15wThP6Hy4rJQmkeH6Tk7RdOwO028JYI9/XreOt09WAOwMnj6Mua/S+n4o9yb0hgC/f2fG+brGWMOre0NePxx/y3OTTWC+c15cvv6TSxD9cXrae/HBed8l+0Urw9vgZUK1AriIyPZSQBWR7wlPVmEPx1kfTvlwndBnBVr4aHfaUdfPjW2cZ121NCEQGssruyN+6OqYiBl3F4GjScogjri1PySxlvdP13z1wYIHi4rIGrwPEAy7k5RhYnmwqLm+l3K2rtjJYwY2ZpQm7B5mzIqGqoXXDkcAvPdwzShL2R0ngOH9k4JXDkYs1zUYw4005sok42xRM1s3PFjVjNKY69Oc148GWNMvR3W8rFhW/ZJUmH4poKbzzKuOr99fYCxExjCrGyJjSK2haB3H84Y0NkwGGaM0Jc8jus7z1oMV2MDeIOP9swLvHfO6wwBdF9jNYhofyGPLV+8uiY1hWRecLkvK1jPKYiZptJlHDE3rGCQJ75yuuL474L3TNcerimmesq5bqhb+4N4cMMTWMh5E3C5bru0N+P3bM2xkaJxhXjTsT1JMgHdOlozTlP1xwrunS2JreWAtVdtyumxpfWCYWn761b3LOa3OB4qmvwngXODORcnRJCONLNd3BjxcVqzqjsjCvXlJ0TomWcpg07gJPqyMEuDdsxXew7JqORhlmyp+35DKBwMu0LqwGfr84frWj6+/+/55v3ZubKHqvr2A+p0Grbp1/N7dOeMsulwXt3WeRdmRxJZu0yjruxniPomw+GhN4qrxTAfPn2PcdJ53TlZEBmI1qhIR2ToKqCLyPecj81qfcVH7ZKB98uObe0NmRcNrhxMa53HOE0cWYwyfuTrh+t6QJLIM0+iyCvbqwZjdccZX78/7eYdxxCiPyZOYcRoxzlMuyopJEvNzrx/03Y87xzSPmVctq2B5ZW+INXC2argyzfDOcGMn5f6sIuC5Nh3gjedoZ8AosbxxZcTxRcV0mGBM4GiS8/rBkLdOCgyBEALLqqZznukgwRrLyMfsDjO889zcG7CeOCbLkrLusJHh+mhECIFF0TGvGiaDGBtMXz1zHa4B31V03nE4HTB/uKJsO9I4oguBnXHKn3xlh/dPSu4v1jQuYI1lmFjWtWNdd5SN42CcUbWOddtyPCtxIXB/XrCqHHcuCgIBawzzVcPeJGXRNJyvDS54zouGonWMogQbAl0InC1qLoqW2arlLDQcL+LL4bKJNdTOMy9qbu2OMMTMy5rjZYmbBcrWc76uOFk33JgOOCsqirZjd5CwqFoeLmpC6OfjTkf9etVN69kfJqyHHV9/sGBRthBgVbc03lN3gc73a8lGFiZ5gnOeVd1SOcc06+cC50nEg0VJnkQcjjP2R9nmTOznosbWfKSC+uRyVk+7yfJou/fOVjgfLsPlywatsnUQ+vncgQ9DeOs9XeOxWXQZ8h8/hqcdz7ej6Ty3z1a4EMiib79a2zrPsmzJN1X654Xq1nmWVb8WcBR9MvORRUTkk6OAKiLfk77d5YMef/7uMGVddwyIMMDRJH9uIBjnMT98dcoH5wUOTxQst3ZHZLHl1YO+w+1vv3vKKunngx6MM37m1X06H4gs/IfjmEHSr2H62aMJe6OEuxcVgyzlT795wP4opW4D4BhkCXlseTArWZYdV2zOTp6wN0yZVR1l48kiS9fBouqrhl0Ht/YHzKqWZdFiDBSNY5DE7OQpbxz2XY+HaUTrQx/qWsfVScYHs4rzVc3ZqsV5xyCJOJpkHA4zVuOWsrG0Hqw1RBaOFzX35iXLusP7PtC0HpzzWNuvabqXx1RpzJVxytcfLrlSNbzzcE2eWLzvOwsnMZyXLXXwlI1nmsXEsaWqW0IwRMYwyC3Gw2lZc7FuSeN+GaAstSSxYZwlnK8rztYNrXPcmZVcnebMipbEFjzcBNvWey6KhrNFhTGGK1Nok6hvmlXWrMqOB4uC2qfkSYQZJJwuK5wLfPn2BWlicb6vOuaJ5f68v3EwzSN+6+1TdgcphXPs5TEmwL15v+7wtd2cedGR70Q0XR+ehmnMJI+5fVYwHcSs6r7yHULgG8dLkshwPO/n/1obNucFTPOEVw9Gl0v0nK5qkqgfhv6oWvu0Ycvw9Lmx6WZu7rrucN5ytqqJjMFvhk0TDCfLipMldC4QgCQyOA/Gfmeh8tHP2P15icGwN8yeGharpu92nCXRM1/n8TnGgyR97hzjZ81H1rBfEZHtoIAqIj+wPq4S+9TKbGL58RtT0iSiah1XJjm7w3SzbeBwkvNGHuM83NgdMNo0wmk6z++nfZDtqzdR39122C/fMysa4qgfUgn9sFxrDa8dReyNM8ZZzP6wr7z9wb0LdvJ+uOtkGPFgVtKFvpPsT96a9pWw1hNtusu+ejDi3qwkMnC2bum8Z145bk5zpnnKaBCzW3s61wfPNIqIrOWVgxHXRjlpajhfN9St5/XDIaPEcl45XjsM3D23rJuaMQlZaumC6SuTecJnr08JPpCmlpOHa6yBZdkQ2YzruwOOFyVpFHFlkpHElitjS9l0WGsgSrmxmzPMYt44GHK2mRMcgmFR1lSdJ09jLtYtkYWi6tehvTIaULYdZdPylTsz8B8GM+8DwyzhcJQxyiO8D9yfl/2yPs5zZ1aybh3LkzWHkwFvHI5Y1B2YwLLuMI1nXnqmecQwzzhf17gQOFs1/WtPPOuqo93pg/2DRcWsaHj3dMUoS2ic49YeeDx+E3TXTUsAZqW/7IS8rFqyKOK8aGg6x9eOlwTvcVhu7vYh7uo0x4XAg3lFcP1Nh4t1Q+McgzjmaJJxVjQEHxhmEWerhrNVQxM8P351h89em/Rzc2PL4ThlWbWcrRsuVg15FrM7jCkbR9l05GlMZAyrumV/lFM0jofLmr1hgkv79zeNv3Ve75NWVb828TBLLjt9N63nG8crvAvsDVNu7g5Ioo8G69/94ILOO65OBk+d0/voZ/X6bt4vazXJnxsw09hyfadfMuvm7vAyzH/z4YLg+6r1jd3hU5tcvahPMux2vh8qruAsuokiPygUUEXkB9rLVmKTyF52bR2l8WPhFNI4Yn+UYkzfnOnRBW7fpMeTRoaqdaSR5eaV4WW19lE198nhkyHA148XhNBv96hjbZ7281YPRhk/8/o+v3P7jJNVSxZZbuyOSWLD2arhZFkzSBOu7wwZZQmrqiWJI6o2sJs7Dsc5O6MYYwyJtXTecW13iAmBB6uSde04NTU/fGXCeJBwf1b2nZNrt1mzNDDMI3aGI/bHGQmGRdPywXnBIIs5WzUcTVLePytZVy3BByaDhEFiGWQRR+MMa/umUoH+sXEW0wVP7QLWWurO82BZc/+i4v68wGK5uTdib5SyM4i5c1H2x2UgixKwgWwTvJalw3tP3XliY1i0Ledly+mqJEsSDsYxWZzy5tGAnTxiOkiZDmLePylYVh3/8e1zrk5z1sOYLLZENmJV1SyrjnnZkEUR16c5d2drVlVH2zpGeULTeL62WHKyrMAHiOD1/YjTVc3xvMJvhu5OBjGjTfhrNs18V3XLOw9X+Chw/7xknCXM1y3Xdgc8nFeUTcJ7pyv+6HiB94Gyc4yyhNOi5q2TNdYEHi5qstQyzRNu7A05GPXzfRdVS+cCf/RgwSiP2MlT7i8K/uj+kjiGRdUxSGKWdb/80dm6YVG0eGP4iRtTkqhfV/d0WXK8avjmseezV6a0znGyjDieVxxN+1EIR+P8svPyuu4omo7/37tnRBimecznrk1JYsvpqmacR+wNMmILRdvxcFURW4M1hlEa83BRkyeW96p+SZ/DUU6a2G+5SE+jiNj2//bP8mgJIoBh+mEAXVUtbx2v+vnMi5IfudZyNMm4sdv/nD4+3PpFGjB9cL5mVTkmg4grk8ELP/fR8x//XXC89rx1vKBz8ObRmL3xs5vLfTf8cYSi74Xg9WkfY9N53jpeYiO+45ELP+i24d9y28/3T5sCqojIS3he1TWNLa/sj5752OHAcHWaP7Na++T267pjJ4uJnhhyHBlDHBkGacTRNOfP/+h15mXzkWVXjiYNXZiRJbbvVFx1pLHFdf1w4zyP2R0l7Az77fvgHfHafscf3pszK1pia/BAEkfsDFLSyPLBeUEae45GAw5HKeM0wVi7WdPWsqo69kcZ++OMedHSesds0TCvW+7NKl45GHJrf8T+MGGYRMyqljcOx9SdI40tD+YVV3cGHM9KdkcJ7z1c8WBW9VXF2NCFwP443QyVDVyb5jgfCMbQ1J5BGpFnlgezirJ1nK37cOMDDJOIJDKkUcyyafngvCGyNeuq4dbBkOkg4uGyJY4Nu6OIqvNEsaFqOrLEUraOazs5ddvROGg6x52LNRerljSKuHNRcmvf4L3HGHDBYwjY0FdDV7OWuunnF6fWEltDGhk6C8PIs65b/v1bp9xdVlR1hyWQjjNc8JSNY9V0VK3jg9pxNE6Yly1nqxpjKnYGCafrklnR0rSBNIJq7HHB03YD0shyXtSsK8fJouK8rFmWjpNVycW669faDY7XDkaMspgbOwOyyvHGUcbdiwJj4GCcsihb7i1q8IaLdUvVNvxP31xwXtSUtedHbk6IjeHVgxFXJjnnRcPd84J52fJwWfPqwYCHq4rzoub+vCaJ+67JADtpwqJsudgsHWWAxjkgUDWO98/60Fc5z5tXRuwPM3aH6UfW3oU+TD7SdJ66dRhj+uHEs5Isthyvag7HKXubG0xV66g7R0XHyaJmkhfMi4a7s4I8jljVHVcm/c/uoyHWj79G1fZzjdO47zK+qh1RZPqbGcWC2FqyxD6zAvz4vu5crGmcJ4sidocpZ4Xj9+7MqRrPeVHzP//M0Ude/0XVraPu/OVxPvm6T/u91XSeP7g7w/vAwTh74VD0vC7ZT3vt985WtF0/EuDm3hD4ZOY4f1Ie3XQIoZ++8Pj78DLf67f72o9G4czKhiSKMAOjudMv4cmbPl+5MyO19mMbqn23juW9sxUR/Sgp3Wh4OgVUEZGX9Lyq6/Mei635yIX0x0kiSxzZj3Qrbp3ncJSSpwPqzR/dcR5/ywXrKIu5Nh0wzmOWZddXbXcHjLOYcdYvYzPKYi6KhnnZD/9MrGF3mPLZKyN2ByldcByNM3746pQ06S+8DGzmenpiG3N9f8DhMOf+vCSKDDsDT2QtwXvKugMMSRoxDBE2gsNRRgiesg3sDlKGWcLBKKHzj9akpR9+2XgssDvK6FzgeFFSO8s4iRlnES4BFwACeRxTtx3WwrW9jL08IYTAziBhb5RwOM45X9UM0ohF01KUHQSomkAWO87W/ZzUaR6zO8hYFC1l61mULXW7oukcr+wPqVtPUXVcFC3WwN4oYZLHjLMEay0ez+sHI2ZVQ9NCajxnZcVu3s/tfNQReL6ocRaGiaFsDKNBxFsXjn/3zRP+6MGSRdlyf16SRIZg+jm2s6LicJLiAlg8b5+sWZUdi6ql6TpqF0gsNC5wNB5wWtTEseubbvlAMI552eK9oSVwf1b2w6atIbKBe/OCQWS4ay239gZ8cLbCBcOqbrh9XhAMDM5Lru/mWAyjQczZquKbx2veO13TOcfJuuHhquLGzoA3jsa0LnC+qpiXHXfOS77xYM6dizVJZPjs0YT3ztZMBwmjOCaE/mbIWw+XzMqWoul4dW/E4WjA0TTndFkRJxYMPJyX7A4TzpY1rQsMUsutvb6K73xgVbeM6au3X70353xdUXeew9GAe/OSNw5HnCwqVmXNsuh4ZX/I2bohspZ53TBKo83NgI64injjcMyiaClad7l27aubIA/wjQcLFlW//vGjAGoMlE0/P/uD2ZrEWsZ5zNEkf+ryWo+HkGXZh9vY9Em7doF52bA3ymm7vgP0oyD8ZNOspwWjunUUjeOd0xXrqmWQxP2NiDy+vGD/xoMFteuXlXo8gLeu74D9+Fq/zwphTffhMPWLdXNZNT6a5NybFX0TOviWC/LWeeZFe/ka67rjfF1v1kwOvHk0ZrSZc/0ylexP0qPvKU/jfg7/Y+/DOycrYmuIrGGaJ7x9smKcx5/IEk1N53n/fM26dkQ29B3kMazrhJu7w0/ke3vyBsu3u48/rhsKL/taTed5++GK2HI5rYYQiCLzkXP6j0vrPKfLmmEaX/5NeJnX/0GpviqgiohsqWdVa6011Juw+KxmMElkGSRR34goMiRRTNH0w4uPNnP0ms5zPK+YlQ0X65YfuT4B4ObemOt7gbrxHxlSmESeQRpzJeqru+u6Y75uiU1fVXp04fjawZii6fgTAd7frKf59kO4kkcM081xJI5l7dkbZhyOM2brlvEgZidPuLeoSNKIg2FKFkcs646DccrOIGU6SImNxUSB3TSmajy7+zGH05xJllB0jp1BQpZGrMqOuxcRWRbhgufqdIAzgbLoOCsaHs4qiq4PnC4Ymi7lxl7E9Z2czvcdkp2HZdlyvKhpas90HDHMYgwQG0sWR331N7EMkpjTVcWs6EiTiGFuGeRDjiY5H5yt8S4QAnTBM05iMJZl2VJ0HXdX8KXb58zLlsQYRllEnsSsK8/VaYy1MBkkPJxV/UWVD8yKmtI58AYbYH80YFHURFHgcJSwN4yoW8dJ1+GDpe0866Kl9p669Vhj6LwnjgzjNGJ/knG6qFnVLTt5yu4o6SvSScRs3bGMWgyeSRbjTWB3nFLWjjgKnKwdnYdhbCkax/vna37kaoxzgXnRUHrH3jjn5mae6NcfLLko+3nNfgyd66has3n/WhZlQ2b7194ZxAzTIau65ev3F8xrByzJ84i29WRxxLrubyJUm6WOxlnM7iDlg/OCpnM8mFVE12DdtHzzeMa8djSd5w/vLdkb9kOh3zwac7IoOYtqVrVjOkjwQNls3rPaQwC3WV/3YJwxSmPOi5pV1bFuHdYYfujqhBu7A5ZVi/Oer9xtySNL0XS0nf+Wn9VHVdOuv+PSV40dGNNfvFoCxhhmRc0gMRDgrU0jrQDsDlOKpuPOeQEmsDfMeO1gzDiPuVg3fPn9c6yx3J33P4vr0vGl98/4sRu7XJ3mHE1yFlW76UZtWNctP3R1enlDrPOBRdlgSQmhP97fef8c7+FonPL64RiA33rnlLvnJV1wpFHEm1fG5JFlUTXMy5ZX9kZPDQRPNq0CuHtRcPusoPWB989X3NwbkcWGe7OSW3t9gH680voouD75/8+6iH/Ri/zHO2qHpzTWejRnfJQlRCHwh/fnXKwa9scpV6b5C4ePZx1P3526X/KpdZ6dQcJ0kDCIY4x5zg6fo24dnQ+X38M7JyuWVct0kPDGYxX+l3mP3jtbwWYqyrcbyl/k9arG8dX7c6Z58i2V7GdpN53Vh4917a5aR+s9o3Twkb+hlzeKWk/r+2Z2385ohed9f3XrOFk14Gv2RylXxjnwYnPMH1XyvQ/PXSLr0Wt9mutrf6cUUEVEttjThv5+3BI7T9sOvnXIXOs8ozRimA4g9F1ah2nE0W7+1Iu7x/dZt463T1bsP1qP1nBZUUpje/lHfXeUcn13wLsnK6yxuOA4XzccTTJWlePHb+6wO0wJh/0+roxz3j5ZMMwSQoBhFjFKY37s6pR51RJHhkmWsG46itoxGiQkEUzylEke0xb9cSfWkqeWPI8xeI6mObuDlJv7Ay5WNbOy5c2jEffOK+7PSoaDiKLsh9Xe2h+zbJp+OF9k+8A2SFlHLdbEOOfoQmB3mHN1JwcCe8Ocs70hlXMsipZZ2Xcbng4S1o0jT2OSKOJoahjEhsEgIcLycFniO0hiaLpAYiyR7UPH3ijDEKhbz7LuqBpH2Xqy2FA0LZXrGCYJs6qha/qKb5ZEhBAYpjG3zyoMBQFDFltGeUyaWSZxSlH3w2rzJOLGXkZkLefrhrbzRNb0yyI1LcHBZBTzoKw5GMfsjRKGcUQXzGYN3JZxnnHkDGXX4gO0neOtByv2B/3NBhcCi1VD5zzHi4rpsF8GyWKY1zV5HPHOw4JhFnH3vK8cXxQtdy9Kru8MuLqT88r+iDcPJ9Rt4DOZ5d6s4q3jJVXj8L6vup+sagJwfTIgiQxXd3JmZU1ROR7MS87LvgEYQNU6LtY1xlo+ODPcmRX8iZt7VF3/Hr9/UXK2qvtuzngMhs7BWVHjl/3FoQues1XdL4lUB17dy5mVDSfLirrrz/MksoxTS7BggXnVkMQfDkl+ND/3Yt1ircE/qsy5fkjzxbqlCZafu7nD26d9CPjmwxV11zJMU+7NCtZNP2/ZGMswi/jc0YTOBXaGCV9877RfbzdLOV/V2AjA8u5Zwaru+KGjCemrlnnRcPuiYBBHvHu2pmw7FmXH/jhjVbVg+hEgbz9ccjjOeDivyJMY53w/f75zvHe24nhWc7qqiWPDompJI0sUw3zt8D5wZTL4yBDsR78vDscpZdPfXAJYN47FumHVOe7PA8fLmiuTjNNlQ+cDr+yPGGUNs7LB+8DtsxUeQ+cCN3dzHi4qDsf5U4clr6qObxwvGKYR2RPLMz05FPSthwvqzjNKYw7HfZ+A3eGHDcH8ZkL5um4ZpTGp7TuLz8q+ov60G4hPBrGycfzhvTnTTUX78eNJog+r8YO0v8FHgDQ2z+1U/SxN5/m9u3MiY9gbJv3NjboDA0XtPlIZvnOxpmo9sTUcTfLLngpPHv/jFfD4iWWbms6zqlqMMR9pOva0YdHvnKxw3tO5wPWdAbuj9Fv+vi2qFudDfxOCF6t+xtaAMRRN30jv/qwkEOi8ubxZC/T/DvfnJAb+8N6CSZqwM4r5U68dfMch9VGw7Je0C+wNY+LIYjHcPluRxBaD+ZbpA09qnWdW9L/nH4XtR19//ObMu6cr4shwWvbd4b8XK60KqCIi32NetLHT08Lt45LIYm1fhbm2k3/kIuTj9plEloNNZfV5ldxHofZokl/+Eb03K2hd4HD80QuER8fUN02CB4uKJLLc2B1yY3d42Y03soYHi5K3j5cs65Ysjkkiw/4ow4XA7dM1Z0U/x7JoWvaGOVcnGcZCYi3L2hFFlqt5zDiJGKYWbwyjo5ifuL5Dllq+9mDJ+ChmXTu8H/La/pB785pAYJhmxBHsDTMuio6DUd8F1xs2wzMtP3Vrl6Lr54LO6pbdYcL7pwV52ndIjo0liuDnX93jYdHy4B6cripu7Q45mubsDBNc6LseN51jvGledDwv+87CoR/4Pc0t18YDrkxzTtY1uTGsW09nPcM0IrYRdd2xqh1Z3L920/Vrt3o8r+6PGGcJiQXnE37o2oS3HiwxxjEeJLTBkZmUNvFYa5mtW8qkX95mmMSUtWeSx/zY9SmDPOLBvMaHvmr6r3//Ac4ERknEzijlh66OuXtRkGE5bRqSBIY+5adf3+V82TIeWJZFQ0OgqfuL/9O4X0ZnlPY3KlZVxwdnNe+erwk+EBnDunEMs4hB2YfPO873DZ9OIsrKsWpaoshwf14ySvvKUxZHnK4ahllMGllmheH2+bpfvqfzxBi+ebygdYHfettzOMl4dW/Ur4HrAu+dlXjnOJxmxFGEd46vPWhJoogvvntG+agZU+jD8KMuzqdFyyvTATf2Bn1VtGxYlR21c3Tec7JqGCX995mlhkFSU/vN2rptoPMNt89LwLM/SXlwXm/W4m0BaFzfZOwbxwtunxXcOS9pfb8m784ooe48ddtyUTQQYFF5FmVHFvfBLRtZQoCzZcWX7iw4HCV0DqajhAe2IbKe/VHGouxvFvVV6BmRtTycN6ybhrJzHOYpVeuIrCU3McF3vH9WsDfsh/s+GXhOVw1t199IGGUxy6rhtKy5c1FgTcSybkjjiDi23JmXNG1fBSuavvHXf3jrnL1RyrJ2nC0rZlXD/qrmR67sbBrZ9b+rms7zpQ8uuFhuqlc7+beEsqJxuM3Num8cLxkkCXncsjNKWJYdJ8vqslJ/NM4ug+tOnrCo++kB1hhi+61zbdd1x72LkjQ2lxWwdd0RQsBYQ+M8J8v+996j/3aGCUXT9d3HNzc2jsYf/t581OX5yarxo9eDD28ezop+Ka5Bnl4GvHvzkrpxWEvfUX1TOZ+t28sbbsuqY3/Ud5zvm9V9OH/y8Qp4ujknHq2b/M3jBW89XJHGEZ85Gl1W2+9c9NX8R+s3t87zcFFysuynCby+P+IzV8bfMmf70WvNy5bIGMZp/LGVxyTqu3YXTceDeXXZqXyS85Eq9HLTyK8JUDSeg4ml8YGi6b7jgNpX2jviqJ+m4FzgYl3Tdn1IT2JDmkQfmT7wtO8piSy186ybjhASVlXErOhHatybV1zfGWCAZd0yzfuhzN+rc5UVUEVEfkC9aDX2Wc+9tff0hlDP2v7RNq8fjp+7tM+ji7bjRcW8aLlrC27uDdnbNIBalC3zsiOyhqoLvHKQMkj7P2cPFjWLoqXrIPi+QmgBYww3dz9cEmiU9SHgaJJxuq5ou8B00M/rmhUNaRSRxobffvecg6tT9kcZu6OUedX/4b+5O2B3mPJwWTHJEo4XJZ+Z5Ny7KHAh8ObRhKLuMBbcyZqq7jgaJ3z22g4nq4qmc2RxzO44J8tibIA6BMrOU7nAzSym8Z4QUs7WFVXTEdm+QhaAznuuToe8cTTCOU+S9o1+XtkfMVs3rJqW1rXUjaMJMMoiru8NuDZJOV913NofcrwoabtA0TjK1hE2offa7oCzdUVkDAFDE2BZdDTO896p44eOxpwsKoZpzLzquL6bszNOeGVvSOcD756sOF7WzKuGcRJjhglpHLMqO5wLlDjqxnFtd8CNHUPrfX9h3xlmlSOPbX9h5T1N47l7UbKoGnbHKbOqYTJIuLk76MNd1VJ3HedFvaluJVRtoOn6Cuk4S1mVHcM8wTlPUXacVTWjJKZxHc26I0sissRQtx5CS+vAhYALhjwx1F2/3M5X7swZZ5Y8iUkTz7LoSBPD8azmym5O5zyHk5x3Hi4w9GvFnq1rDP36tXvjlLNFzapoKNuOYhNgI2NZNi2LoqFqA6fUNJ3ns1dHOAfDyHBlOuB4WXM8qyidxwYYZDEn64q7F/16uzvDhDyJyGPDB7MaQmCcx8xXHavG8frhgGAs33iwJPjARVlTth3ed6RRjAsO7zzGGM4WNQ/nVR9aI8O66dgfpyyKjvdOC6wF5/spAMuqZd04jO2XT5rkMdd3BqSJpWoc50VHlkXEkeF8XbE67+fEDtOIo0ne/+CHQOcD3zxeUjSOQKBqPXkSsztM6FwgjQz7o4SLqgPbz5t9uKwoG8/xsmSQ9aMd3j/3VJ3n3nnNqugYZfHlBX/rPLGBQRaxahzTxl0Gqj4k1dydFZwtGtLEUNSOm/uwqjwPlxUmMiQ2wnvPdJCwqFrqtn/uunKX5+DBKCONP2xktKo6vvZgQescbz1c89krI65OB7Sub6hWtY6zRc2ybrgoWh6uGm5N+07YaWK4KDoWk5Y4ijgcpzTO0XR9g7G3LxyHx0vO1jU3dwdkcf++vnOy4r3TNWway7kQyOKIk2XT/66NMqrWMUwtdet5sKz4j++e8cr+kINxynvna7rWc140ZJv37tFc81GaEFvDrGgYZR9Wl89WDS70N47yJOJ42VfaoW9q96jiVzaexvXN33aHKSEE3j8vOFvXuNA3mGtduFzb+VH4Bjgcp9w5L3i4qnm4qnltf3jZbftpo4X6tZv7sP9oR8uqIbX9OuhN199seBR+m66jC45ZUXMwyvuq9XcoiSwuBNqmYzpIuLozIIoqBknMuycr5lXgxo7leFkRbXpVPGv4rnOeVdX/fTxZ9c3p9scp66qjGHYMkoh11d+YwHx7lfZtoIAqIvID7EWrsZ/kcz/uef2FpGWYxGSJfepQrrbr51E1TceycBwM+wuf2MDBJOs77eYJN3YHxJG5HDIGfSXD+b5CsjtM6TaNdR5dPO0O+wuXedmyahw7w4Sq6zia5Pz4zT1WVXe5xm0I0DhPZPoLoPGgX/7myiSHCZysKq5NM/JkxEXR4LzHBdgf5sS2HyI7SGJ2R4bd0YAsiZlmEcM0InERJ8sVkbHc2h/RdlC2nkneH/PrhyNu7g6JrWVnGHF9khOC4eo05WLdksaGbzxcETw4PDd2M5yHKDEsi46dPCVPIw6HGfdmBZNhSvCGLLFMhzHOwZXIMNwMcz3aGfDV9+fYENgZpBxMMjxwMOmHDDcucGWc8XBRc2USmJd9M6nzsuEzhylxZEgjS905sk2zq5u7OWkcY/YNs6Ll2jTvS/IhgIFBHDOrahoHd88rFuuGue2HySY24uok4tp0SOsCsaWf0+pbFpVjVtY8NBXew2lRk8WW6SgFGxgPUm7YEVEE58sG1wa+dm/Bzd0B1/ZyPjMZUrUd66KhC5Y8jhhYw8404eG8wZqI1nuK2uFN4Ob+kK++P2NVrShb31dly5a68+yOUk7WfSV0mnrGA0vtBzgfWFeOLIl472TNsmooWkdsDIM05nTVcms3YieDQKDu+rBwvmpIYsOi7IdND5KY3AQ+ezTlzaMhZeuZrTseLitcG3BAZAL3ZjWjNOa1wyHTQcL752uC7xtrNa7hYJSRJxFtF+i85+ZexvFFTTpMSGLL3YuS4APLpmNv2M/vHWWWbx4XLKqGQRJzbWewWW/Y0nSBo1HGou7YG8Tcmdd0YU3Vws4w5uHcsar66tSqavna/QV/eGfOqmm4Mh1iTYAAFos1Hmuh85CYwLsnS0wwXJQNwzQhNpazZcMo66vZe4OYLiQcTDJa348Y6INQP4c4thCbvtr3zeMlWWy5vjPg3qzg9knB8brm1d2M1vc3vYqyI9g+MCexIbURF6uKB/OaKDasa89Pv7pLHkcQ+kZWnfMM0nozYqHm/vmSbz4sOF83HC9KPv/ZK2Sx5b2zNa1zrOuOzjvmZUfTdByvaprN0l9s5j1frBtOFgX7owGTPOLhsuas9Nw+L2i926wB3L/+bDP64u3TJemZJbMRP/X6HpMsIo360Heyqrg7rwnBEVvLe+drvvZgxm6esjtOORpl3J2V/O77M8rWcWM/J48irDWY0FdvP3d1QjCGnUFMbA3OB+7M1jjnOVm3ZLG9HOUSNu/N+ari4bphkiZYs+q7j3d95dkFt6n8Bm6frcliy/1FP/qhqB1xbPoRAAQshqJx/ZDtLMK50C9dlSbsj1P2Rxl9OO5/t1+UHV3naHw/x/Tfv3VCFFnePBxxc2/I4Tjlg7OWNw+GdMHww9cmGNP/zYCPDqX1m/md2aa51ONNwh4F4scr2keTjIt1wzCN6HxgkCaUdUvl+27vi6ojmP7mWNG4y/WlH7csW87XLRC4WLdcm2YsvadsWrCGZdlwsQrEkaHzsJs9fT337wUKqCIisnWSqF8C4PEOxo/kScTuIKVuHZ+7NmE6SC+HDF7fGVzO83lUxXmyUvvk3NwH85IPzov+lUJf4e3vXle8ujfoLyhcIImiy2D7qCLzaF83d/v/P183FLVjVjSM85jTVcO6dnQucDBKuVg3ZJHlomioOsfeMCU2jkFkONwfMspifvTalFnV0bkWi+XKTszNnQHTPOGnbu30S/9Ehlf3RySxZVY0xJHlcJxfrqd7+3zNKI2xxnC8agje473hxu6AH84THsyKft3ZNOHORUGSRewNUgKBvZBQOo/HgeuHER8va+6erekINCEwyftlgm7t9kFrOuirRga4tTdgEEWsqpYs7pf2OZrkvL4/4veLFh/672l3mDKvHE3bcn9R9c2HvKduPIPUcjBKWdQN1ljqtq+aD9OEYWapG8ebhyOyyJAklrdPS6q6n59WtYEktqQ2IhiYDPt/ryyKccEzTgd4Azt5RiBg6OdL0rWs6paLtSWNMz7/uUPWjaNu+3m5H5yXtMGxO0wYZ/0860mWEEWWk3lNFMdUbcswjcjjhKuTjONFwzCL8MFxMMyoW0fr+jVx94YpeWpZlS2GwCRPaboaa/s1ljNLvw7wKvDb71xQd31l0eMZZRl17WAzzHJV95WZZe25WBe0znBrb0ga9yFsZ5jRtB5P4GxVEYLhcJSRpxFV45gXHYM44v6sIU9hXrREWCofGNHfKBhNLVUXODmumZcdeM/5ugbg/qJmkvUX3tNBzOKkw4VAtwsQuKhaBoklYOhcyx/cnWOMoew8r+4PNz96hmEec1G1PJyX/NjNHV7bjyjawEVR88FZwSCNmVcN56sWfKBuHcusw3WOIniOphll3X8/B5O+0vje6YqLomZ3kGJNfxF/vGrYyyN+984FWdw343rzcERsoGw7lkXNSWK4ujPoRyjElvfPSlyoiKzlc9fHrBtPlvadxR/Olty/KLm+O2AyiHn/bM3DZUP9ruPKpA/pf3BnzvG8ZJBG1M7xpffO+MqdvsnaIIsZx5Zl0w9bdsGwrjqSCKq6Zd603LlT0DlYNw0/fLX//RUIWANl2xI8nCwr9n3G3qAPNncuCs5WLZMsZh5afuedMw42S3N1LnC2rmmco6g6lkXHO6crXPDsDzMOypai7of2e+8Bz7rqqCOPCVA7z6rsyNI+3ENM6wPnFyVvPVyRxYZl4TjcSXh1f8QojfnNbxyzarpNE6LAq7sDbp+tuX26pnawrhuuTPvfY7vDlHuzktZ5PrgoWFcdF6uGonH4EEhiw43dEWm0Q+s7qjbh/ryg6jrGWf+zfeeiYGeYbuYRRwwSS5zHdM7ROc950eJcIDKwN+qX0bp9VjId9D/LX3+wYCdPyWK7Cc+B905XDFLLvHRc28l4ZW/E0STn3ZMVX75zzv1ZxbXdnKvjnFcPRuRJxP4ou1yz23nP3jDlxt6A2bqfipEYS2z697R/zb7q/ORQ337ySyC2lhD6qQx7o5RrO/1IhNunaxZNxyRLyS2bv57fmxRQRURk6zxv+HEa912DAaJNVe7JwPhxDaQePbau+yV4jqYZJnA5DG2U9UuCvLlZLiWJzEeGkT16/pP7Gqbx5TDcEGB3kPDK3oBF2bE7TBimMYeTnGXV0nnHjd0RzgcWRwn/s595hWESk8SW22drFnXL7jDntcMBznNZtX3y+3sUSh+fe1Z3jtb1lc5hmjBI+47DLoR+SYzIXt75f/1wzI/f3KF1gd1hwrxo8QQeLmumecyiavlTr+zw3lnJZ65MGGYRJsCVac6tvRHvnCxxIfTz7qK+4U0XAp8JY/IsZneQ9MHEw5tXxiRR3wV1mFpOFzV2aLDW0riOOIowwOmq7Idv1p4r04za981OpnlC6x1ZYvvGV8OkX3Ypibh9Vl7Ofx0kBmv6inDnHAZDGhvKFq7uDpgOIj53dYf7s4KvP1hxtqiwoQ9MnQssyo43j1L2Rv2csdNlxf4oYdFYXtkf0jkwGC7KmqtJTlF3XN/NuD8LvHY4JIkNB8OUVw775kBX1xmDNGZZefbGMRermiYN3J+X4A0msjRVP9wxjgyVc1yULY3vbwjkSQTGM86TTXOzmKbum2+N05jruzn7w5Rl03JROJz3rGq4uZsTWcv9Wckkj3j9cMRrB/28weNFSdU43jsv2BmleGBZN/gQMS8dewPDNO2r+Q+XBXEUYQ1EGPYGMQ5YrPp5ijYEqi5Qto40jhinhnnpuHO2puw6YmsZpQnn65okNiTWcjjJeOdkxd1ZgXeBh6uaZdVxbZpTVC3jLGI8TKHsWNSWs3XDPhBby9VpzumyIQqepnZMhylF1TIrWj53bUrtHKm1VJ3ja/fm3NgdEQKM85jOBe7NSt4NjotVQ9V5ru3kLIuGZd3x3skaawMns5pJ1g8Nb7xjXTvWXceVScrZomF/kLIsHW89XNE5w91Zubkh4Pji+3NiC/NiM1VhXeNdIIos3gSWRcdXH8wZxAkm6oeA+2nGMI2o24hgW+ou4EPMvXnV/+x6z/WdnN+/U/Lu2brvWt70w7ebrl9+7GQz7LVoOpq2H/3xYFlx+2xJ0wWO84QfjwyBwLJquHNeUnaeCMhSy8Ek7TuLty3DLmZW9sNcT9c1F+uOJIkJtetHjUSWVdNxsmhIjcEAD2YlddcP3V1UgWXZMq8bqi7wxffOaDqHwzDNI6rW8/X7C4KFNI2g9SzrwMFmnWtj+m7l9y8KTmY1F2W/VNeqbgDLjo2Z1zXzouSds4rdUcSi9MQWZssVq7Jmb5T3N6DKmjud29ysM4zzfl77w02fAzuHnZMV33y45MGs5v7CkFjDomxpPFwZpgQLg9jwR8dLrDGsKkfdjchiy7Jq+coHF/zenTlnq5r785JX94cMs4hJnnK+rjlZlFysay7WLXfjgtm6IksSTOhvkq67jsRG1F1HfmXK6bLmxm5fRX3UVCq2/e+9qunYG2WXFetHN1tHWX9Tt6w7honle7R4CiigiojIlnreUOBxHvNDVyffEtZedtjxo+d26/7OdL8kz4f7et582afta5p/WPUdpjGDpKPdNFzZGfQXGnFkyTb7ssYQRYbDYXTZQKTp+uUk8thy4muMMQzTD0P40+btPhngH80PvjL+cI3azgWGWQShP57poG/u5FzflTaJ4GjSN8ta1x3LqqNoHF0Hr+yPuT+vWVT9kjuvHgzwgb7rbdd3pazavtEPwN4g4+de2+f+vGJ30FcgDsdZ39yDvknI4TjjD8OMe4uKLgTePByzO8w4L+p+eNowZ1HWDNKYUYj54Stj1nWg6Frm636Id2QM13YGLKqORem4t5mPOc5TEu/4uTf3uSg7bu3lxNbyYNHPN94dZFzfGdA5cN5wMEo5Wzc0re8bTq1KPjjvl35JY8O8cEwHEVMDdRtYVS15GrMsOl6/MqRqA7vjhIt1Q906rEkAw59+84B2E9xuny4pu5IHFxUnq5J3Tws63/9b39zJMaOM/WnK2bzmwaqhaAJ113C69EyXJbujlHESM9mPWdeeVw9H/Q0U77DG4n1gtu6o64YkifEEVo3jtYNBP2w7j4mtZbJZoqMbZZz4isQY4tRyvqxJbN/8JssMh5OUovEM44gosgw2Nz/2xv0Ngabzm7mQ/RzT/WHK0aQfRntedLTOMS9a9kcpAYP3AR88V8YjFmXH+6drnIOAp3b9usKjLGKaR+yNUvbHGZGFh7WjafphsoM0Zi9PaAlE1vDGwZjjZcnBOKfOE4zpu5Yuir7x0L15ycNlybLqOwTPqv48tNZQNYE4tizWNfGiny86zhLyNCKNoe0CPkDROqy1XN1NOV0F8jSi9p6TdU1koWs9RzsZ9y4KFmVLmkQs1w0Yw53NGrjBBEZJzCA2lK2HqGO+DNRdxTixHE5yru3kBGBWVnTOUDtHnkZcrBtu7o04WTYsq350RuMDbRvofEcXAs4HytYTGcODZcXvvn9O20LtWmJrSaOYYRpovOd4XlM1jpt7Oafrfn5p4w1J7GnbfvSJB+qm437Vsi4dRzs5P3R1zDCNebio+jmpXcCEwMmyYF40nL11St05DkYDAg7f9e9VEvfDr09XFcMkovGepk05HPeh72xREzBUrWOa9+cW9Df40s3v4+v7GfM7DfcvSorW9cfsIoqLhq+HvrHVMMuJbd8gaJjEnC1bbp+XrOq+idA4T5kkEUliiUx/k+1wnJImljvnBWXrOF3UXN/NmK1a4ryfQ3u2aLifJ0zziLOi4cGiJrWBJE4o2463H6753JURb5+saLu+4mxN3/Dqmw9XFHXfETs2hlXb0rSBLLWcLGuuTgb9PFFraLpA2bXMipobu0MgcH9e8HBR8v55wZXJgEHWN3wqG9ffyGo958uaVdlwNO1Hs4yzmDyNubU34PY9VVBFRET+WH0n82cf38frh+PLZi1PDql6mdd4WgU3jZ/9OXzYzOO2NU/dz6390bdUbV/0WPpmUDDK+6pr0/XdH/vOjoGq3ayL+5RlhVpn2RslRMZwf1bhPLx2OGJ3kPJwXvHm1THJJmi/tj9kf5xyf1ZyuuznObZdR+fg9YPRc5eoGGaHHC8K3jlZszdMmQ4SXj0YcmWSc76uee/MEFnIIkueJdzYS4G+AuQdfZOeuu8o+ydf2WV3kFD5jqruq16RiXj9oO/WSgAPOAc3dwfsDFJWo5a67S8+f/T6FIPBA/NiwNVpzt15Sdf1lYuq8RSNZ5BFxMYyziJOYFOdBdcFXj8akVjLZ45G/Vyx0M/Ne21/RNM5ZmXH2bLGBvDBM44j1nXHw2XLrb2ccRKR7uXsjVPeOl6DgSbArGjZHaZc38t5ZW/EybzgnfOC41XFJE24Oo25fV7x4KKkaD05jnEeYwksCk/Z9iH2bLXm1f2cPI0pu34+tDGGV/YHl92S66afQ/u56xPmpcMYmK1b1k3HIE144ygnMvBwWXPrYAzeE0wgSyIGScwgizgcGzB9R+3IQtX26/aGDr52f0HduL6ZTmxJbT/ndpwl7O2lHE5SImu4WDVEcX8D58ZuTp5GHI0SjqYp67avlK8bxzhPuDLJWDUdgyxiuW5JI8N52Xc2HiUxbecIJCS2X8P55n7Og1nABMs0c0yGMct1CyGwN0wp2o7IeuIIDqcDyqZlZ5Djg+HaTt43NVrWdA7eOy94sCiZDjNa5/AEksTSdn2Fe3+Q0rQdMXC4k+GDZVbWtF1LawyHk5ydQcKDeUnTBRZVy8E4Y151zNaWs1VNZPoA43zNrb0h92cVby+XeO9pVoHpecm73Yo4inhwXlAFhw/9z87uIMbTXXYXXlUd00GMtRG7o5TFuuFi2VB5xziN2RmmlHXDouo4GGUcTlN2B9GmYugYpBHXdnLOVw0npm/A9t7pmnnZ0rnNfN8843CSMkgtzvVDj09XNZh+fvNnr1iqLmJnkBAM7IwS3NIz2cwzPRxFfPXejIermvvnNVjPrcMhO2VM7QKLouWiqEljy52zgmD7Gxad9ySRwUWBs7KmcZ6Hs4raO2JT9425pjnXdwaUXd9lela0hGCII8O67vjg3G861BuWVUdHYFE3YPomZFfGKceLiv1RjLWBdeN4uKyJrGE6iJkM+1EjN3ZzLtaO2+cFVeMv1/I+C/3Ni84Fbu3n1E0/9H5R9tNCmtDfhDhdBb70/gWdD8QGfub1Az5zNCGJYDpIOF03LIua87Ll4bJk97zgaJrzw9emxJHROqgiIiLfy/owl378hi+8r2cH3I9b+udZ230Sx9R0DffnFYMkIo0Me8P0ucsZZFF0uQTR7jC9nBP8aGmEdDPvNWzW0H00dHg6iBkk0f+/vXONsSy77vpvnffrvqpuVXVX9WN6pmdiO37FiUICKESEPIiQDFIgwwewUKQgSAR84EPCBwiBSAkSSCAhokAsORHBtgIRFgoYo1jiA8JxEhzGdhhnZtwzPf2q7nre53luPuxTNdXdVdXV457p6un1k67q3H3PvXefe5f2rbXXWv/Fcje8r43QvdeVRR4iCXllHYNZadNclzoR29OCJPCYFjYCYYyNSt+r2PnNjTFbk5LIrxmkPmXjsSMl5/pd4tBjlJdMcxtB+9Ba3zpi7XVfGnZY7sTAW604JrltJWKAbOrSBC494+F7LrOyIfYdvnp9B88RpBVmEoTYd1npRMwqq+baiX2WMqv+XNYNnchnIQ7YiAIchPFoRmnsZ7ncC0giG11aSGw7kVs7BbOywHNgpRux1AkIPI9ZUTKpYTkLGXZihmlIRYNhSj8LuNzJeOPOhMZAGvj0YpckcrizaxVWb+zkXBq6DGIfDyv2cn6Qcn4hoZ8EfOPWmA/FXZY7Mf3E56XrOwwinxs7c6LApRN5pL4LCKNZSRJaZ+TiIOXmzpwscnhze05ZNZztJ0yrkk4tlI1htR/y/66PmBSlff68oKrBFcF34PwgYpCGtia4E/Jtyx2ubU+5OZoxLmqC0OfmqGzT+gMGicswDXluOWN9lLM7LbiZzxExNLV1hoZZyLSocdr2QnHocr5n6zAbMyOaCbO8YVxVVFN4bhgReEJeN1zbmDPsBpzpRjy33OGZhZidmW1Xcmt3jjgOncDDcYXVXozrODhOwyC2it8wY1bWrfKzg+d5ts2Ua8V9XM+hqWzKcRI4iLj4HhhjcIxVnXVwqA0kgTCvHJv6XVXUpqYTB7y+Add2Z3jAtGzwxNA0UDZ2Y6WfBDy3nBF4Lrd2pjSN2F7Dk4K6bEh8lzx28WqrEl5UDd04ZF43zIuK9d2czXFOEvlEvkfoC55j+wynkcvutKauGlwxzGtDUTYsZNBLbHSyqmompU+nbNjNC4oCNncLdqc2Yps3DWFuVXt7mW3lszu3IlOjecW8qNr2Qz5R4DBIAp5dztidlUSew1fe2MZxhWsbE+LIpRv6bNcFjmlwMEyrmllRkwa2RMB+T8I4b+hFDq+t5+S1Vf0+04vpJR7PLWVc2ZjgALHv0Yk8OqHH5qSkn4R4jl3nNicFo1nB7+9MWUhCPrDW5WI/ZlQaTF1xezynaRrbA3maM8x8Li6k+K7w5uaUeWkF25Y6PtPCKjonUcJiEjKZl1xZHzMuKpraEAc7+I7D6iBmMQs50w3ZmeTU0xqMLUW4tWP7Tgee3UQbTxrtg6ooiqIoyunFd6VNtQ2YFvWxTvBh0eC9Wte1e2px96LCy22ktG7YF0Y6yT9GvusQ+y6zst4XxNprBbQ2SKiahuubU84tpjgiiNztTPZin+5aj7xsWOqEAFzftmI0tp+lbVWSt+0k9p67d533bk4EXrBf67snfrU3z1u7M7ZnBY4IxgjnBykfONvj5u6MNLIqptIq7qahw6So6KfBfrR5t7BCTknk8b3DIYHvUlQ1gkNeVJwbZJwbxGShx0fmJa/dnrJphWHJgpBB6nN7Msd3wfOsA5BFHp3Y49X1EYEjzOc1zw4zFjshge9yphNR1oaq2mFzWrR9MxuSULg9tm17VroRFxZSyqYhDXwi30HE2kvsu4xnFb00YDH1ubiY4TlCXm3jOcLl5dSqUnvCeF6y2ao3T/KaNHYIvZA4cLiyPuG19QnigOdahe6qMkSBtS3Hcwh8d7+tSew5+J7DpWHGs0sZL4U7VGXDbFYSRDY1t6pculHAmV7CqGjYGOXEoccw88mrmktuSuQJf3Rjl41xwZ1JQS/0mcwrQk94/5kuPi5Xt8d0QxvB72U+y15IFvhc357TmAbXcdiclCSBQ9kYHHGoakNdllxYyIgih6VuQOy5hL7D+ihnfdywkIVsjnPb5qkXs5CGzMuGWVWxmIXEoctLV7dxGqtJ67lC4nv0Ip+lbkRRGbqJz6yoiQOPoe9hqoYN3+H1zSmuWEXZwLFpsdOioJ+EDAKoDJzpxKwtxFxYTGwGgRhubc+5vmUd7KVuTFnWdGPrQKehFewxYvDEZTTPmVU1u5OaadkgkpP4Lhgb1R5mIZFbUJzp0tSGnby0olKu8NrNEbVYQR/BpiwLDlniWAXseU4WCMNuDNIQuEJVNbz0+hY1hqI2CFacrjKQVzW+51glchEiz8P3YaUb43tWEKqaN4jYVktl3dAYiDyXhSiwkf1OyNl+RCfy2ZyWFJWN/C/1Qq5vzIk9YTGLbPmA5/DCmS7ru3OyyMNzHVY8B88Rzg4iiqphlNd0Y8FxKs4PYoyB17dnlKXh+mhOXVWMZiUNgjGGa9tzIs/l0lLKcysZzw0zPFd4Y3NKGjjc2M1Z7YdsTgum84o705zJvCby4fU7Y0Qcvnp9l8vDmGE3Jg1c2wJpZ8rXb5SEns3YWOunPLOUUjX3K+A/KaiDqiiKoihPAYHn0gk9psVbjuDx5x8f/T1s/Nml++uCHzyvw8WtfNfZ7zjjOg51Y3Ddu/v6+a6t9zKw3zbooDNtjE0vzKvmRNf84GtN6Ux9HBzSyOXq5pTAt1GdQRZQ14aNxjCbVQSebUG0J7q12k/48GqfbjRhVtUMQp809hFjiIO2PjT2CD2XXhRQGWG5E7CTefyFD68y7EQ0rdJxNwrIy4p+FHBhMcUYw/PLXT50rs9oVhEHNupT12ZfUAyB2ztzotDjbC9s1UUD+kmI64h1UFyPSV7t11H34oDVXkzTwELqUzb2Ozrbjbk1mmOMdRD2otqDOOD6zozAdXj5xi4rvYitWYGLQ/ecz+1pzkoWcGdckgYO13cKxDEsZSGjWY3jCEno0Yt9Ut9l0Kq5gq3N/sM3t63idBTwvtUuF4cJw8xujJwfxNRVzUIW0DSGUV5xZ1LwzfU5O7OKyHfbnq0VzW5bhx2FrAwCtuYBk6Iir2q6kUcvCtiaFezMC3qR16qB25rKyHG4uTvDwSEKPT56YcD2PKdqrFpx1nh8eG3ALLepm9bmbA/ShSygG/lsTwumVYM0Dc8MMwTh9Y0JLlgHKrb/nr+yMybxXJZ7kW2B1Iu5sTUl2nWpaljIAqrdW3QTj3lR0U9CLi+lzKqahThguR+z3AlZ7kaUdcMbmxOoDLOyom5sNsAg8VntBzyzmBEHPme6Ef/rtTvMyxrXiegD38xH7E5yyhrOD236+mhesdwNuLGTs9QJEYFvyzqM5hWbs8JucPiuFRqqDN8+7FNWht3ZnGs7c7qRz6wy7MzL/Q2t65tzGmOwrmzDvIY4gKaxzuu8qumntqZ9KQuYlDXnFu1G1UIaEfs2Gr47rZlVFXUtLKQhF5dSPAfESg3wjVtjZmVFnjcY07A5zoljF3Edu7lRNMxyW3/vObY/6p2xVRgGQ8f3cAJhmle8vj5mWlZ8Y31ELwpY7oVEnstonDPohAw6DvO84kw/RowwrxpbJ+o7bE1zDHBnVHCmG/Lq+oRX1yuaRkAanlnM2JzkuGI/gzsjK0K1Oc55/2rNM8OMxdQncLtE22Py3LYO2p1WjPOK86Y58Zp32lAHVVEURVGeAk6qcvytvsej6o17bxufw2pxj7qmg693bx3wt8Ke87vnxJ0fJCx1ov2eh0XVII7tD+liqOu7FTazyKefhCyITZ0eJAFJ8FZUd2+eZd3wzGJC4DmUt1xWB7YFUVE1B1opWec0i+z4SteK7PQimyJ97+f1obU+b0STfeXrpY4VjjrYyumwz/O55Q6+69BgePX2BM8toI3O9drI9F5U23dt1LioG4adkMVOSC/xbZ9aV3A3HKZV3bZE8plWDdc3ZzSVYZCGDNOI8bykG/kErrO/4QCw2k+4tTvj/CAh9BwueRnd2EeAwHVIAo8XznT308Jv7sxJtid44uB7todwXlpBrNIYktAjCRw+eG6B2Pds39VG+JMvLNOLA27sTFntRdzYydkcFWyM4H1nXXCwTmzik/kevu+wFia8emfCNK/ZHOWcHyRcXkoZFRW9pGG5E9rWIv2YJPC4uTvDdx22JgVR4OEKpJHDMI0YdkJevrVL1TQ0je2x22DohD5rvYS1XsKrt0dc7Nv3dBEuDFOaxhD5DnlhGHrC+bZ+3XNsq62qNlxYSBjNK+6Mc8Rx2JkVdJOUD64ucHmls5858MG1Hp4jfPP22Dqg/RgxIWUNrsC8rNic2JrUzUlBFnoMUp9Lw4zaGHZnJa/4I3amFWVlcJyaswsRg8CnkyzyyvqI3XnNldtTPEcIPIfVXkQv9KhMwxsbUwLXZaUXcXkpZZxXjPKSxU7EUhZijFX5nc7rVtTKZTELqI2wOy3Ymub4uQsCzy+nPDvMuLI5oTaGq9tz5mVJFvo0tSHyPUZlw2RWsi3Cy7fGrHRizvVTHIE3t6c0xlBWFdNZiThixai258yriiTyWeiEzMqaxTSkrA13RjP7mlsTBlGEJy53dnLGRUXme7x0dZtuEuCI1QjYnefUpmFWNWShSxR6VE3DUhbSGMMsr9nNC0aFYZCEJL7DzrTk5Vu75FVD1TSYRpiVBZnvcn4xYyELCHa/peXusaIOqqIoiqI8JTzK2tZ3g5PM90HnPOprPs7R992GTugTLrr70cuDTvO97ZEOq9Hdoxv5GGgVlp1j3/skmw+HKV8f5rwfVid8eaXD9rSgaaCX+MzLmro290Wmj9pUAPZVpa9uTlnqhozmFWuDhPed6VE11nnqJ1YF+LDNCN91GOUNNDULacDzy939Psee04rC1G+lcU/yijTwGSQN5wYLFG0ErjZCWdWs9VJWBzEr3ZiPXoCqvjs1/aKXUVY2mtdPA0bzgoXU1sjWDcyKhn7iMUxDxnnJje3ZvvBY1cB3XRpSNs1+uvne9w027dQAwyzk8lKHsmm4nJdc35rhiFXh9l0IXY9u6PPV6ztgBGPgg2t9zvRidlqhq63U5VwvIfYFW9dbMilti6nQs3Wlt8cFZVXjOy4fOd+nagwfy3wCET5yYZHLy527PutOaG3v+ZUu377mcLNtdfON9V0mRcX6jo0YGyPMypobOzOapv2OpuV+ZP07L9j67q1ZwZluTNUYljsRC2nI63fGzPIS3/MY5wXzomGQhXz/IOblWyPe3JxSN1A0DR9a6xOHHlHgUNeGsjLcHs8YZMF+Wr2IkEY+8zzk+q7PxqRgOi9Z7aV0k4DV2jCvGq7emXBzd04nquhEAZ3YJwoayqpmqReSVw3ro7lt65PFTPOKq5tTxvOaooL3ne1YxXWx/YpH04rXNkZUlSFta7QX0oCPnO/ytesjznZDhp0Q0xi25hXPDW19q+cJr96akG9MiHyfJDA8O0zZnBZsT+aIIwwSn/MLCWVjaCrbgzUNbTmEmeYEhcfaICJ3G8LFhM3YtqEKXWGSl5A3XNuatn29n5x1H9RBVRRFURRFeSiOS3c+zlE8qj3Sca8zjOWhU61POu+TOu8HI8d1Y45Uf37Q/PYirFVjWzqd6yd4bSr3g2qWReDiQkJjjBW58p39yHNe1tzYmSPGpoSvDZJ9de69GuLr2zOmRYUj4Dkundjb76F8KTy8ZdXehkKDYWPkEvm2n/CfeWGZN7amRJ5LGrosd0Ju7cy5M5nTi0Kydn6DMLivTzFwn40UVcOt3TmuIzgCq90YzxOaxl5PL/F5ZjGlahrKpmFtkNBPKnxX+OOXHcbzCt/1yQKPudfglxVNYzdItmcFGMNSJ6Ixtg3JILa12qFvVcwflJWw2tZiZ7HLxqjg8rCiqAxX7kzpRj5n+xH9xMMR8B1hK6+Y5TUgXFhMiVvRscCVdgPBvuetUc7OtOBsP2FtkHJxaDMFlrKIL5Q3yQIP0xjOD5P9Ofiu/bz8deHm7oyrmzPS0GWQhHRCH89x+HNrfXuu49BLAjBQVtv4RYnrCi+sdHFdh8gXticVC6nPJK/sd10ZupmHaaCb+rzgZ0Sug+tZOz3Ti8DYyOsgDsgChzDqMggDcGz/481JwbyouTTMWMoisthlMbXjxhi6kY/nCJFnlZQHcUAv9fjA2R7XNib8wdVturHHaF7z/jMpN0cFuWlY6Uf004ArtycY49g+yhgi1+WDF/qYxpDnFVe2poDh1lyY5PUTWYeqDqqiKIqiKMoj4lFFdPfO8w60IHqcPIoU8XtfAzjx6/muY9uSwH31xFVjo1f9JGBe1vt1v3sCWJO8ohv7LHVshGzQpg7f65Dey8ENhWcWs7sc8j115r15vH+1y7Vt22M2Dd27osqHpa8fHCvrhiRw9x3uvfmt9ROmRdVeY4ODkAR7/ZCtmNdLmTDIfFLfY2dmHaOVfkQWe0hrOvOyJi8b0tDd73V83Od+2EYGwM6sojHGttvpR219sOA6wqyq97+bqoIs9nFdK2p2mN0spiGX2j7TlTGk4Vs15L04YDEJcF2HpjYk/t09oAPP4f2rPZtGnob0k4A3NiZc2RgTuB5nus19jvd3XFzgznjOYhYzKyqKuib0PC4twkov4s44Z31nxpZXsj2teGElY9gJmRce43nT1l8Ly92IzUnBuX5ELwru6+98YTFla5qzvpvznW1rGWA/RX9aVPiOw7SsCH2HUV4jGPqRFVPrJSGLWURR1dR1TeS5JL7D1rigrm3LrSwMCFxhd26FnxaT0KZ4ew6xHzGsawLXZeO6YV7VT2QdqjqoiqIoiqIoygN5FOnSRzk/J3ne0anVVm14Xh4uALYnuLWXknxUe6WTzPeo8edXuqz2E3jI19+b357Y173zyyKPP/HskGlRkQQeWXS3CnU3dOjHoe1NOsmZlw2dyEaHrUiY7dFZNw1LnbtTzh+Gsm6IPJfIc/EcWGgdQ7BCZG9sThnnNZ4rrA6i/bTmt9LJ736/0HdZ7e/VU9+dDt9PAz64OmBWVsS+Rz+9vw1Y4NmU6aJqMMBiFtKLA4adkNqY+6KGWeQReAnGQFFbAbGzvZjtWdGqZPt0ln0QuLo5pRcHpIHHM4sZFxZTdqYFN3fnTHPbNubigq237acBS93ors2K3daRnxUNq33vrs987/sbYJ3XnVmB7zgkod1QWM4itiY566M5y52QXhIRBC6h57KTlwyiAAZ2vnFg1aHP9mLODRL7/m2P5bJuyHyH5+5x1J8U1EFVFEVRFEVRTj1vN7X6UUR/Tza3t9dP+SSp4Qcd07ue69paxXlZtw7VW6nXZd3QCa2i816LpreL7zqkgUtjDI7IXQJWncjHdxwW04C8alg9JEL9MNcceA7Pn3m4VPjlTsTtke0DepRi92HvmUbeXYrfBnhmMWWpbZu151SmoUdeG7qRx5U7E8Z5RRK49zngk7wi8h3SMKQxx7d5Oex7TUP4nueWePXOmMhzCTwhqR1MA4PEZzGL+N5eTNk0vLk1Ja9qQte96/v42MVFpkVFed1lkD2aHt/vNuqgKoqiKIqiKE80jyq1+nHxdufnOXKsc+s6cmha9NuZ31HvE/lWRfdhI9THXfPDpsLb4wdvQhxXh33c8/cc9KoxnOlFdzmw957niNCYt/+ZD7KAD0f9u6Kyq32b6r33npO8Iq+a/Rrmg47wnuMbnWJ7fxAnclBF5EeAfwm4wL8zxvziPY+HwK8B3wlsAD9ujLnSPvazwE8ANfB3jDGff2SzVxRFURRFUZSnmLcbWT6t7/N2+VY3IR7kMJ/kGh/VZ3G/I313JHQvbb14yB7PTwoPdFBFxAX+NfCDwJvAl0Xkc8aYrx847SeALWPMZRF5Efgl4MdF5APAi8C3A6vA/xCRF4wx9aO+EEVRFEVRFEVR3uLdihyf9gj1o+DtRHXfybmchk2Bd4qTXM13A68YY14zxhTAp4GP33POx4FPtce/CfyAiEg7/mljTG6M+SbwSvt6iqIoiqIoiqIoytsg8JyHFuR6UjjJFa0BVw/cf7MdO/QcY0wF7ACLJ3yuoiiKoiiKoiiKopwOkSQR+UngJ9u7YxF5+XHO5wQMgTuPexLKqURtQzkOtQ/lKNQ2lKNQ21COQ+1DOYrTbhsXj3rgJA7qNeD8gfvn2rHDznlTRDyghxVLOslzMcb8CvArJ5jLqUBEfs8Y812Pex7K6UNtQzkOtQ/lKNQ2lKNQ21COQ+1DOYon2TZOkuL7ZeB5EbkkIgFW9Ohz95zzOeAT7fGPAb9jjDHt+IsiEorIJeB54HcfzdQVRVEURVEURVGU9xIPjKAaYyoR+Wng89g2M580xnxNRH4e+D1jzOeAXwV+XUReATaxTizteZ8Fvg5UwE+pgq+iKIqiKIqiKIpyGCeqQTXG/Dbw2/eM/cMDx3PgLx/x3F8AfuFbmONp5IlJR1beddQ2lONQ+1COQm1DOQq1DeU41D6Uo3hibUNsJq6iKIqiKIqiKIqiPF7ee41zFEVRFEVRFEVRlCcSdVAfEhH5ERF5WUReEZGfedzzUd59ROSKiLwkIl8Rkd9rxxZE5Asi8sft30E7LiLyr1p7+b8i8rHHO3vlUSIinxSRdRH56oGxh7YFEflEe/4fi8gnDnsv5cniCNv4ORG51q4dXxGRHz3w2M+2tvGyiPzwgXH9zXmPISLnReSLIvJ1EfmaiPzddlzXDuU4+9D14ylHRCIR+V0R+cPWNv5xO35JRL7Ufs+faUVtaUVqP9OOf0lEnjnwWofazKnBGKO3E96wIlGvAs8CAfCHwAce97z09q7bwRVgeM/YPwN+pj3+GeCX2uMfBf4rIMD3AF963PPX2yO1he8DPgZ89e3aArAAvNb+HbTHg8d9bXp7R2zj54C/f8i5H2h/T0LgUvs74+pvznvzBpwFPtYed4BvtDaga4fejrMPXT+e8lu7BmTtsQ98qV0TPgu82I7/MvC32uO/Dfxye/wi8JnjbOZxX9/Bm0ZQH47vBl4xxrxmjCmATwMff8xzUk4HHwc+1R5/CviLB8Z/zVj+N9AXkbOPYX7KO4Ax5n9ilcsP8rC28MPAF4wxm8aYLeALwI+845NX3lGOsI2j+DjwaWNMboz5JvAK9vdGf3Pegxhjbhhj/qA9HgF/BKyha4fCsfZxFLp+PCW0a8C4veu3NwP8WeA32/F71469NeU3gR8QEeFomzk1qIP6cKwBVw/cf5PjFw3lvYkB/ruI/L6I/GQ7tmKMudEe3wRW2mO1maePh7UFtZGni59u0zQ/uZfCidrGU0ubcvcd2EiIrh3KXdxjH6Drx1OPiLgi8hVgHbsp9SqwbYyp2lMOfs/7NtA+vgMs8gTYhjqoivLw/GljzMeAPw/8lIh838EHjc2fUHlsRW1BuZd/AzwHfBS4Afzzxzob5bEiIhnwH4G/Z4zZPfiYrh3KIfah64eCMaY2xnwUOIeNer7v8c7onUEd1IfjGnD+wP1z7ZjyFGGMudb+XQd+C7tA3NpL3W3/rrenq808fTysLaiNPCUYY261/1w0wL/lrZQqtY2nDBHxsc7HvzfG/Kd2WNcOBTjcPnT9UA5ijNkGvgh8Lzbt32sfOvg979tA+3gP2OAJsA11UB+OLwPPt2pZAbbg+HOPeU7Ku4iIpCLS2TsGfgj4KtYO9hQUPwH85/b4c8Bfb1UYvwfYOZDCpbw3eVhb+DzwQyIyaFO2fqgdU95j3FN//pewawdY23ixVVy8BDwP/C76m/OepK0B+1Xgj4wx/+LAQ7p2KEfah64fiogsiUi/PY6BH8TWKH8R+LH2tHvXjr015ceA32mzM46ymVOD9+BTlD2MMZWI/DT2B8AFPmmM+dpjnpby7rIC/Jb9/cADfsMY899E5MvAZ0XkJ4DXgb/Snv/bWAXGV4Ap8Dfe/Skr7xQi8h+A7weGIvIm8I+AX+QhbMEYsyki/wT7zwTAzxtjTiquo5xSjrCN7xeRj2JTN68AfxPAGPM1Efks8HWgAn7KGFO3r6O/Oe89/hTw14CX2loygH+Arh2K5Sj7+Ku6fjz1nAU+JSIuNsj4WWPMfxGRrwOfFpF/Cvwf7AYH7d9fF5FXsKJ9L8LxNnNaEOtIK4qiKIqiKIqiKMrjRVN8FUVRFEVRFEVRlFOBOqiKoiiKoiiKoijKqUAdVEVRFEVRFEVRFOVUoA6qoiiKoiiKoiiKcipQB1VRFEVRFEVRFEU5FaiDqiiKoiiKoiiKopwK1EFVFEVRFEVRFEVRTgXqoCqKoiiKoiiKoiingv8PLCZk1qXIexIAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot classifier error\n",
    "\n",
    "C.plot_progress()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Sv7R988-4pCv"
   },
   "source": [
    "## Classification Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Tfi78c5XTLD3"
   },
   "outputs": [],
   "source": [
    "# load MNIST test data\n",
    "\n",
    "mnist_test_dataset = MnistDataset( './dataset/mnist/mnist_test.csv')\n",
    "\n",
    "test_dataset = ds.GeneratorDataset(mnist_test_dataset, ['image', 'target'], shuffle=False).batch(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 281
    },
    "colab_type": "code",
    "id": "cdQCvzor-QWx",
    "outputId": "7bd211b9-c41e-4415-d728-8879f9567f84"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# pick a record\n",
    "record = 19\n",
    "\n",
    "# plot image and correct label\n",
    "mnist_test_dataset.plot_image(record)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 283
    },
    "colab_type": "code",
    "id": "CqZmjoqs-iJH",
    "outputId": "228cf568-36cd-4b71-ed67-3a1129982886"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD5CAYAAAA3Os7hAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8+yak3AAAACXBIWXMAAAsTAAALEwEAmpwYAAAO5klEQVR4nO3df6zdd13H8edr7TbZBsNkVzL6g87QOaoojJtCnIYZfthtpI1ItCUIEqB/SGUGQqyODDKjGWo0GuePxgEBhboR0WtWKQpDIrLRjo1tXbdxKWNtwa2MsQWGjMLbP863eri7t+d0Pff09MPzkTT3fL/fT7+fV2/vfd3v+ZwfN1WFJOnkd8qJDiBJGg0LXZIaYaFLUiMsdElqhIUuSY2w0CWpEQMLPcl7kjyY5M4FjifJnyeZTXJ7kgtHH1OSNMgwV+jvA9Yd5fglwOruz2bgr44/liTpWA0s9Kr6FPD1owzZALy/em4Cnp7k3FEFlCQNZxRr6MuA/X3bB7p9kqQxWjrOyZJsprcsw5lnnvmCCy64YJzTS9JJ75ZbbvlaVU3Nd2wUhX4QWNG3vbzb9wRVtQ3YBjA9PV27d+8ewfSS9MMjyZcXOjaKJZcZ4LXds11eBDxSVV8dwXklScdg4BV6kg8BFwPnJDkAvBM4FaCq/hrYAVwKzAKPAa9frLCSpIUNLPSq2jTgeAFvHlkiSdKT4itFJakRFrokNcJCl6RGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpERa6JDXCQpekRljoktQIC12SGmGhS1IjLHRJaoSFLkmNsNAlqREWuiQ1wkKXpEZY6JLUCAtdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpERa6JDXCQpekRgxV6EnWJbknyWySrfMcX5nkxiS3Jrk9yaWjjypJOpqBhZ5kCXANcAmwBtiUZM2cYe8Arquq5wMbgb8cdVBJ0tENc4W+Fpitqn1V9TiwHdgwZ0wBT+tunw18ZXQRJUnDGKbQlwH7+7YPdPv6vQt4TZIDwA7gN+c7UZLNSXYn2X3o0KEnEVeStJBRPSi6CXhfVS0HLgU+kOQJ566qbVU1XVXTU1NTI5pakgTDFfpBYEXf9vJuX783ANcBVNVngB8BzhlFQEnScIYp9F3A6iTnJTmN3oOeM3PG3A+8BCDJc+gVumsqkjRGAwu9qg4DW4CdwF56z2bZk+SqJOu7YW8D3pTk88CHgF+vqlqs0JKkJ1o6zKCq2kHvwc7+fVf23b4LuGi00SRJx8JXikpSIyx0SWqEhS5JjbDQJakRFrokNWKoZ7lIk2LV1huO+xz3XX3ZCJJIk8crdElqhIUuSY2w0CWpERa6JDXCQpekRljoktQIC12SGmGhS1IjLHRJaoSFLkmNsNAlqREWuiQ1wkKXpEZY6JLUCAtdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpERa6JDViqEJPsi7JPUlmk2xdYMyvJLkryZ4kHxxtTEnSIEsHDUiyBLgGeBlwANiVZKaq7uobsxr4HeCiqno4yY8tVmBJ0vyGuUJfC8xW1b6qehzYDmyYM+ZNwDVV9TBAVT042piSpEGGKfRlwP6+7QPdvn7nA+cn+XSSm5KsG1VASdJwBi65HMN5VgMXA8uBTyV5blV9o39Qks3AZoCVK1eOaGpJEgx3hX4QWNG3vbzb1+8AMFNV362qLwH30iv4H1BV26pquqqmp6amnmxmSdI8hin0XcDqJOclOQ3YCMzMGfNP9K7OSXIOvSWYfaOLKUkaZGChV9VhYAuwE9gLXFdVe5JclWR9N2wn8FCSu4AbgbdX1UOLFVqS9ERDraFX1Q5gx5x9V/bdLuCt3R9J0gngK0UlqREWuiQ1wkKXpEZY6JLUCAtdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpERa6JDXCQpekRljoktQIC12SGmGhS1IjLHRJaoSFLkmNsNAlqREWuiQ1wkKXpEZY6JLUCAtdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGDFXoSdYluSfJbJKtRxn3y0kqyfToIkqShjGw0JMsAa4BLgHWAJuSrJln3FOBy4GbRx1SkjTYMFfoa4HZqtpXVY8D24EN84z7PeDdwP+MMJ8kaUjDFPoyYH/f9oFu3/9JciGwoqpuONqJkmxOsjvJ7kOHDh1zWEnSwo77QdEkpwB/Arxt0Niq2lZV01U1PTU1dbxTS5L6DFPoB4EVfdvLu31HPBX4KeCTSe4DXgTM+MCoJI3XMIW+C1id5LwkpwEbgZkjB6vqkao6p6pWVdUq4CZgfVXtXpTEkqR5DSz0qjoMbAF2AnuB66pqT5Krkqxf7ICSpOEsHWZQVe0AdszZd+UCYy8+/liSpGPlK0UlqREWuiQ1wkKXpEZY6JLUCAtdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpEUP9xiJJ0mCrtt5w3Oe47+rLnvTf9QpdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpERa6JDXCQpekRljoktSIoQo9ybok9ySZTbJ1nuNvTXJXktuTfDzJs0YfVZJ0NAMLPckS4BrgEmANsCnJmjnDbgWmq+qngQ8DfzjqoJKkoxvmCn0tMFtV+6rqcWA7sKF/QFXdWFWPdZs3ActHG1OSNMgwhb4M2N+3faDbt5A3AP96PKEkScdupL8kOslrgGngxQsc3wxsBli5cuUop5akH3rDXKEfBFb0bS/v9v2AJC8FrgDWV9V35jtRVW2rqumqmp6amnoyeSVJCxim0HcBq5Ocl+Q0YCMw0z8gyfOBv6FX5g+OPqYkaZCBhV5Vh4EtwE5gL3BdVe1JclWS9d2wPwLOAq5PcluSmQVOJ0laJEOtoVfVDmDHnH1X9t1+6YhzSZKOka8UlaRGWOiS1AgLXZIaYaFLUiMsdElqhIUuSY2w0CWpERa6JDXCQpekRljoktQIC12SGmGhS1IjLHRJaoSFLkmNsNAlqREWuiQ1wkKXpEZY6JLUCAtdkhox1O8UlfSDVm294bjPcd/Vl40gifT/vEKXpEZY6JLUCAtdkhphoUtSIyx0SWqEhS5JjbDQJakRFrokNcJCl6RGWOiS1AgLXZIa4Xu5SCcp309Gc3mFLkmN8Apd0nGZhHsKk5BhEniFLkmNGKrQk6xLck+S2SRb5zl+epJ/6I7fnGTVyJNKko5q4JJLkiXANcDLgAPAriQzVXVX37A3AA9X1bOTbATeDfzqYgTWiePdWmmyDbOGvhaYrap9AEm2AxuA/kLfALyru/1h4C+SpKpqhFl/aFmkkoaRQZ2b5FXAuqp6Y7f9a8ALq2pL35g7uzEHuu0vdmO+Nudcm4HN3eZPAPccZ/5zgK8NHLW4JiEDTEaOScgAk5FjEjLAZOSYhAwwGTlGkeFZVTU134GxPsulqrYB20Z1viS7q2p6VOc7WTNMSo5JyDApOSYhw6TkmIQMk5JjsTMM86DoQWBF3/bybt+8Y5IsBc4GHhpFQEnScIYp9F3A6iTnJTkN2AjMzBkzA7yuu/0q4BOun0vSeA1ccqmqw0m2ADuBJcB7qmpPkquA3VU1A1wLfCDJLPB1eqU/DiNbvjkOk5ABJiPHJGSAycgxCRlgMnJMQgaYjByLmmHgg6KSpJODrxSVpEZY6JLUCAtdkhpx0rzbYpIL6L0idVm36yAwU1V7T1yqE6f7fCwDbq6qb/btX1dVHx1ThrVAVdWuJGuAdcDdVbVjHPMvkOn9VfXaEzV/l+Hn6L3C+s6q+tiY5nwhsLeqHk3yFGArcCG9V3T/QVU9MqYcbwE+UlX7xzHfAhmOPBvvK1X170leDfwssBfYVlXfHWOWHwdeSe9p3d8D7gU+WFWPLsp8J8ODokl+G9gEbKf3fjLQez78RmB7VV19orIdkeT1VfXeMc31FuDN9L5AnwdcXlX/3B37XFVdOIYM7wQuoXdR8G/AC4Eb6b3nz86q+v0xZJj79NkAvwB8AqCq1i92hi7HZ6tqbXf7TfT+bz4CvBz4l3F8fSbZA/xM96y0bcBj9N6G4yXd/lcudoYuxyPAt4AvAh8Crq+qQ+OYuy/D39P7ujwD+AZwFvCP9D4XqarXLfy3R5rjLcArgE8BlwK3dnl+CfiNqvrkyCetqon/Q++n2qnz7D8N+MKJztdluX+Mc90BnNXdXgXsplfqALeOMcMSet80jwJP6/Y/Bbh9TBk+B/wdcDHw4u7jV7vbLx7j/8etfbd3AVPd7TOBO8aUYW//52XOsdvG+bmgt5T7cnpPZz4EfJTe61SeOqYMt3cflwIPAEu67Yzra7Ob746+uc8APtndXrlY36cny5LL94FnAl+es//c7thYJLl9oUPAM8aVAzilumWWqrovycXAh5M8q8syDoer6nvAY0m+WN1dyKr6dpJx/Z9MA5cDVwBvr6rbkny7qv5jTPMfcUqSH6VXZKnuirSqvpXk8Jgy3Nl3L/HzSaaraneS84GxLTHQW4L7PvAx4GNJTqV3T24T8MfAvO9BMmKndMsuZ9Ir0rPpvT7mdODUMczfbym9pZbT6d1ToKru7z4vizLZyeC3gI8n+QJwZG1uJfBsYMtCf2kRPAP4ReDhOfsD/NcYczyQ5HlVdRtAVX0zySuA9wDPHVOGx5OcUVWPAS84sjPJ2Yzph2xXHH+a5Pru4wOcmK/ps4Fb6H0dVJJzq+qrSc5ifD9g3wj8WZJ30Hvzp88k2U/v++WNY8oAc/691VuvngFmkpwxpgzXAnfTuwd5BXB9kn3Ai+gt247L39J7u/GbgZ+n97biJJmi9wNm5E6KNXSAJKfQe6Cp/0HRXd1V4rgyXAu8t6r+c55jH6yqV48px3J6V8j/Pc+xi6rq02PIcHpVfWee/ecA51bVHYudYZ65LwMuqqrfHffc8+kK7BlV9aUxzvk04Dx6P9gOVNUD45q7m//8qrp3nHMukOOZAFX1lSRPB15Kb1n0s2PO8ZPAc+g9QH73os93shS6JOnofB66JDXCQpekRljoktQIC12SGmGhS1Ij/hdHWovsb9WkAwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# visualise the answer given by the neural network\n",
    "\n",
    "image_data = mindspore.ops.ExpandDims()(mindspore.Tensor(mnist_test_dataset[record][0]), 0)\n",
    "\n",
    "# query from trained network\n",
    "output = C.construct(image_data)[0]\n",
    "\n",
    "# plot output tensor\n",
    "pandas.DataFrame(output.asnumpy()).plot(kind='bar', legend=False, ylim=(0,1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "yYobxR3rMne7"
   },
   "source": [
    "## Classifier Performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "dkcHXgffGcKn",
    "outputId": "9e5511b4-8a69-4eba-c2cb-0331433ce9e1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8754 10000 0.8754\n"
     ]
    }
   ],
   "source": [
    "# test trained neural network on training data\n",
    "\n",
    "score = 0\n",
    "items = 0\n",
    "\n",
    "for image_data_tensor, target_tensor in test_dataset:\n",
    "    target = target_tensor[0].asnumpy()\n",
    "    answer = C.construct(image_data_tensor)[0].asnumpy()\n",
    "    if (answer.argmax() == target.argmax()):\n",
    "        score += 1\n",
    "    items += 1\n",
    "\n",
    "print(score, items, score/items)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "V-wOSRixTbpb"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "03_mnist_classifier.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}