neural_network_project.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "79CcW_AZZcdk"
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import pandas as pd\n",
    "import keras\n",
    "from matplotlib import pyplot\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "l27BZzHHaI6g",
    "outputId": "108858b1-aedf-479c-a439-5c7acb52de73"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train: X=(60000, 28, 28), y=(60000,)\n",
      "Test: X=(5000, 28, 28), y=(5000,)\n"
     ]
    }
   ],
   "source": [
    "# load dataset\n",
    "fashion_mnist = keras.datasets.fashion_mnist\n",
    "(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()\n",
    "x_train = x_train.astype(\"float32\") / 255\n",
    "x_test = x_test.astype(\"float32\") / 255\n",
    "(test1X, test1y) = (x_test[0:5000], y_test[0:5000])\n",
    "print('Train: X=%s, y=%s' % (x_train.shape, y_train.shape))\n",
    "print('Test: X=%s, y=%s' % (test1X.shape, test1y.shape))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dD1_fFZMePp9"
   },
   "source": [
    "\n",
    "\n",
    "1.   hidden layer: relu\n",
    "2.   optimizer: SGD\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "z9hcb598V_1t",
    "outputId": "c6708671-a221-4f52-f66f-e77dc6354da2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1500/1500 - 5s - loss: 0.7426 - sparse_categorical_accuracy: 0.7564 - val_loss: 0.5484 - val_sparse_categorical_accuracy: 0.8073\n",
      "Epoch 2/30\n",
      "1500/1500 - 4s - loss: 0.5000 - sparse_categorical_accuracy: 0.8258 - val_loss: 0.4755 - val_sparse_categorical_accuracy: 0.8322\n",
      "Epoch 3/30\n",
      "1500/1500 - 4s - loss: 0.4542 - sparse_categorical_accuracy: 0.8408 - val_loss: 0.4457 - val_sparse_categorical_accuracy: 0.8444\n",
      "Epoch 4/30\n",
      "1500/1500 - 4s - loss: 0.4276 - sparse_categorical_accuracy: 0.8502 - val_loss: 0.4342 - val_sparse_categorical_accuracy: 0.8461\n",
      "Epoch 5/30\n",
      "1500/1500 - 4s - loss: 0.4052 - sparse_categorical_accuracy: 0.8583 - val_loss: 0.4090 - val_sparse_categorical_accuracy: 0.8553\n",
      "Epoch 6/30\n",
      "1500/1500 - 4s - loss: 0.3897 - sparse_categorical_accuracy: 0.8633 - val_loss: 0.3949 - val_sparse_categorical_accuracy: 0.8584\n",
      "Epoch 7/30\n",
      "1500/1500 - 4s - loss: 0.3756 - sparse_categorical_accuracy: 0.8686 - val_loss: 0.3888 - val_sparse_categorical_accuracy: 0.8632\n",
      "Epoch 8/30\n",
      "1500/1500 - 5s - loss: 0.3645 - sparse_categorical_accuracy: 0.8717 - val_loss: 0.3826 - val_sparse_categorical_accuracy: 0.8652\n",
      "Epoch 9/30\n",
      "1500/1500 - 4s - loss: 0.3539 - sparse_categorical_accuracy: 0.8742 - val_loss: 0.3784 - val_sparse_categorical_accuracy: 0.8679\n",
      "Epoch 10/30\n",
      "1500/1500 - 5s - loss: 0.3440 - sparse_categorical_accuracy: 0.8786 - val_loss: 0.3667 - val_sparse_categorical_accuracy: 0.8723\n",
      "Epoch 11/30\n",
      "1500/1500 - 5s - loss: 0.3347 - sparse_categorical_accuracy: 0.8819 - val_loss: 0.3746 - val_sparse_categorical_accuracy: 0.8652\n",
      "Epoch 12/30\n",
      "1500/1500 - 5s - loss: 0.3268 - sparse_categorical_accuracy: 0.8840 - val_loss: 0.3567 - val_sparse_categorical_accuracy: 0.8741\n",
      "Epoch 13/30\n",
      "1500/1500 - 5s - loss: 0.3198 - sparse_categorical_accuracy: 0.8862 - val_loss: 0.3665 - val_sparse_categorical_accuracy: 0.8698\n",
      "Epoch 14/30\n",
      "1500/1500 - 5s - loss: 0.3126 - sparse_categorical_accuracy: 0.8881 - val_loss: 0.3590 - val_sparse_categorical_accuracy: 0.8718\n",
      "Epoch 15/30\n",
      "1500/1500 - 4s - loss: 0.3063 - sparse_categorical_accuracy: 0.8903 - val_loss: 0.3599 - val_sparse_categorical_accuracy: 0.8749\n",
      "Epoch 16/30\n",
      "1500/1500 - 4s - loss: 0.3007 - sparse_categorical_accuracy: 0.8916 - val_loss: 0.3416 - val_sparse_categorical_accuracy: 0.8779\n",
      "Epoch 17/30\n",
      "1500/1500 - 4s - loss: 0.2940 - sparse_categorical_accuracy: 0.8954 - val_loss: 0.3397 - val_sparse_categorical_accuracy: 0.8781\n",
      "Epoch 18/30\n",
      "1500/1500 - 4s - loss: 0.2887 - sparse_categorical_accuracy: 0.8955 - val_loss: 0.3305 - val_sparse_categorical_accuracy: 0.8836\n",
      "Epoch 19/30\n",
      "1500/1500 - 4s - loss: 0.2832 - sparse_categorical_accuracy: 0.8991 - val_loss: 0.3364 - val_sparse_categorical_accuracy: 0.8811\n",
      "Epoch 20/30\n",
      "1500/1500 - 4s - loss: 0.2783 - sparse_categorical_accuracy: 0.9002 - val_loss: 0.3262 - val_sparse_categorical_accuracy: 0.8841\n",
      "Epoch 21/30\n",
      "1500/1500 - 4s - loss: 0.2734 - sparse_categorical_accuracy: 0.9019 - val_loss: 0.3402 - val_sparse_categorical_accuracy: 0.8777\n",
      "Epoch 22/30\n",
      "1500/1500 - 4s - loss: 0.2686 - sparse_categorical_accuracy: 0.9034 - val_loss: 0.3257 - val_sparse_categorical_accuracy: 0.8857\n",
      "Epoch 23/30\n",
      "1500/1500 - 4s - loss: 0.2638 - sparse_categorical_accuracy: 0.9051 - val_loss: 0.3274 - val_sparse_categorical_accuracy: 0.8842\n",
      "Epoch 24/30\n",
      "1500/1500 - 4s - loss: 0.2592 - sparse_categorical_accuracy: 0.9080 - val_loss: 0.3282 - val_sparse_categorical_accuracy: 0.8832\n",
      "Epoch 25/30\n",
      "1500/1500 - 5s - loss: 0.2548 - sparse_categorical_accuracy: 0.9098 - val_loss: 0.3376 - val_sparse_categorical_accuracy: 0.8788\n",
      "Epoch 26/30\n",
      "1500/1500 - 5s - loss: 0.2516 - sparse_categorical_accuracy: 0.9098 - val_loss: 0.3278 - val_sparse_categorical_accuracy: 0.8848\n",
      "Epoch 27/30\n",
      "1500/1500 - 5s - loss: 0.2485 - sparse_categorical_accuracy: 0.9109 - val_loss: 0.3184 - val_sparse_categorical_accuracy: 0.8849\n",
      "Epoch 28/30\n",
      "1500/1500 - 4s - loss: 0.2438 - sparse_categorical_accuracy: 0.9128 - val_loss: 0.3189 - val_sparse_categorical_accuracy: 0.8875\n",
      "Epoch 29/30\n",
      "1500/1500 - 4s - loss: 0.2398 - sparse_categorical_accuracy: 0.9145 - val_loss: 0.3193 - val_sparse_categorical_accuracy: 0.8871\n",
      "Epoch 30/30\n",
      "1500/1500 - 4s - loss: 0.2364 - sparse_categorical_accuracy: 0.9159 - val_loss: 0.3104 - val_sparse_categorical_accuracy: 0.8879\n",
      "157/157 [==============================] - 0s 2ms/step - loss: 0.3372 - sparse_categorical_accuracy: 0.8792\n",
      "accuracy is:  0.8791999816894531 loss is: 0.337249219417572\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='relu'),\n",
    "  tf.keras.layers.Dense(100,activation='relu'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.SGD(),\n",
    "    loss=tf.keras.losses.sparse_categorical_crossentropy,\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    x_train,y_train,\n",
    "    epochs=30, validation_split=0.2, verbose=2\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print('accuracy is: ',accuracy, 'loss is:' ,loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 673
    },
    "id": "rTdQRLoTdU2X",
    "outputId": "0bd05fa1-cc39-4124-e2c6-ec487803df88"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[4.9191121e-06 1.3104421e-08 7.2219115e-07 ... 9.1676461e-03\n",
      "  8.8479574e-05 9.8707610e-01]\n",
      " [6.7780293e-06 1.3455959e-09 9.9701452e-01 ... 1.5424485e-12\n",
      "  1.4248023e-08 1.5841239e-13]\n",
      " [2.0166528e-06 9.9999750e-01 7.9313011e-08 ... 2.5363684e-09\n",
      "  3.2868360e-08 1.7759850e-12]\n",
      " ...\n",
      " [5.9410676e-09 4.1605851e-12 3.5403624e-12 ... 3.7692053e-07\n",
      "  5.9965010e-12 9.9996817e-01]\n",
      " [6.0870534e-01 5.8680815e-03 2.1483861e-03 ... 4.2008054e-05\n",
      "  1.0713136e-03 1.2008744e-06]\n",
      " [1.0322727e-06 1.3454620e-06 1.5999080e-07 ... 9.9904865e-01\n",
      "  7.5610536e-05 9.6093507e-05]]\n",
      "[[413   1   7  14   1   0  65   1   5   0]\n",
      " [  0 469   0  12   0   0   0   0   0   0]\n",
      " [ 11   1 398  11  56   0  43   1   0   0]\n",
      " [  8   7   3 448  13   0  18   0   3   0]\n",
      " [  0   0  46  21 420   0  34   0   0   0]\n",
      " [  0   0   0   0   0 465   0  13   1   6]\n",
      " [ 51   1  33  12  30   0 353   0   2   0]\n",
      " [  0   0   0   0   0   8   0 477   0  15]\n",
      " [  3   1   3   4   3   1  11   3 497   0]\n",
      " [  0   0   0   0   0   3   0  18   0 456]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f779cbe3c88>"
      ]
     },
     "execution_count": 36,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "predict = model.predict(test1X)\n",
    "print(predict)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Jnmlk4u-Umhy"
   },
   "source": [
    "1. Hidden layer : relu\n",
    "2. optimizer : adam\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "OLLOmF1Js27S",
    "outputId": "868ab39d-c01f-4f9f-b33b-8568a8cfeef2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1500/1500 - 5s - loss: 0.4921 - sparse_categorical_accuracy: 0.8226 - val_loss: 0.4112 - val_sparse_categorical_accuracy: 0.8480\n",
      "Epoch 2/30\n",
      "1500/1500 - 5s - loss: 0.3684 - sparse_categorical_accuracy: 0.8660 - val_loss: 0.3570 - val_sparse_categorical_accuracy: 0.8687\n",
      "Epoch 3/30\n",
      "1500/1500 - 5s - loss: 0.3286 - sparse_categorical_accuracy: 0.8780 - val_loss: 0.3549 - val_sparse_categorical_accuracy: 0.8733\n",
      "Epoch 4/30\n",
      "1500/1500 - 5s - loss: 0.3065 - sparse_categorical_accuracy: 0.8854 - val_loss: 0.3262 - val_sparse_categorical_accuracy: 0.8811\n",
      "Epoch 5/30\n",
      "1500/1500 - 5s - loss: 0.2864 - sparse_categorical_accuracy: 0.8929 - val_loss: 0.3196 - val_sparse_categorical_accuracy: 0.8838\n",
      "Epoch 6/30\n",
      "1500/1500 - 5s - loss: 0.2732 - sparse_categorical_accuracy: 0.8973 - val_loss: 0.3190 - val_sparse_categorical_accuracy: 0.8852\n",
      "Epoch 7/30\n",
      "1500/1500 - 5s - loss: 0.2600 - sparse_categorical_accuracy: 0.9018 - val_loss: 0.3259 - val_sparse_categorical_accuracy: 0.8843\n",
      "Epoch 8/30\n",
      "1500/1500 - 5s - loss: 0.2483 - sparse_categorical_accuracy: 0.9057 - val_loss: 0.3178 - val_sparse_categorical_accuracy: 0.8858\n",
      "Epoch 9/30\n",
      "1500/1500 - 5s - loss: 0.2380 - sparse_categorical_accuracy: 0.9104 - val_loss: 0.3102 - val_sparse_categorical_accuracy: 0.8871\n",
      "Epoch 10/30\n",
      "1500/1500 - 5s - loss: 0.2321 - sparse_categorical_accuracy: 0.9123 - val_loss: 0.3330 - val_sparse_categorical_accuracy: 0.8869\n",
      "Epoch 11/30\n",
      "1500/1500 - 5s - loss: 0.2213 - sparse_categorical_accuracy: 0.9154 - val_loss: 0.3184 - val_sparse_categorical_accuracy: 0.8914\n",
      "Epoch 12/30\n",
      "1500/1500 - 5s - loss: 0.2126 - sparse_categorical_accuracy: 0.9181 - val_loss: 0.3328 - val_sparse_categorical_accuracy: 0.8870\n",
      "Epoch 13/30\n",
      "1500/1500 - 5s - loss: 0.2029 - sparse_categorical_accuracy: 0.9215 - val_loss: 0.3227 - val_sparse_categorical_accuracy: 0.8898\n",
      "Epoch 14/30\n",
      "1500/1500 - 5s - loss: 0.1982 - sparse_categorical_accuracy: 0.9250 - val_loss: 0.3291 - val_sparse_categorical_accuracy: 0.8932\n",
      "Epoch 15/30\n",
      "1500/1500 - 5s - loss: 0.1907 - sparse_categorical_accuracy: 0.9284 - val_loss: 0.3642 - val_sparse_categorical_accuracy: 0.8863\n",
      "Epoch 16/30\n",
      "1500/1500 - 5s - loss: 0.1846 - sparse_categorical_accuracy: 0.9291 - val_loss: 0.3349 - val_sparse_categorical_accuracy: 0.8930\n",
      "Epoch 17/30\n",
      "1500/1500 - 5s - loss: 0.1796 - sparse_categorical_accuracy: 0.9320 - val_loss: 0.3352 - val_sparse_categorical_accuracy: 0.8952\n",
      "Epoch 18/30\n",
      "1500/1500 - 5s - loss: 0.1717 - sparse_categorical_accuracy: 0.9342 - val_loss: 0.3632 - val_sparse_categorical_accuracy: 0.8924\n",
      "Epoch 19/30\n",
      "1500/1500 - 5s - loss: 0.1672 - sparse_categorical_accuracy: 0.9359 - val_loss: 0.3663 - val_sparse_categorical_accuracy: 0.8923\n",
      "Epoch 20/30\n",
      "1500/1500 - 5s - loss: 0.1637 - sparse_categorical_accuracy: 0.9379 - val_loss: 0.3503 - val_sparse_categorical_accuracy: 0.8913\n",
      "Epoch 21/30\n",
      "1500/1500 - 5s - loss: 0.1562 - sparse_categorical_accuracy: 0.9403 - val_loss: 0.3823 - val_sparse_categorical_accuracy: 0.8923\n",
      "Epoch 22/30\n",
      "1500/1500 - 5s - loss: 0.1528 - sparse_categorical_accuracy: 0.9405 - val_loss: 0.4495 - val_sparse_categorical_accuracy: 0.8833\n",
      "Epoch 23/30\n",
      "1500/1500 - 5s - loss: 0.1507 - sparse_categorical_accuracy: 0.9426 - val_loss: 0.4052 - val_sparse_categorical_accuracy: 0.8851\n",
      "Epoch 24/30\n",
      "1500/1500 - 5s - loss: 0.1470 - sparse_categorical_accuracy: 0.9444 - val_loss: 0.3843 - val_sparse_categorical_accuracy: 0.8933\n",
      "Epoch 25/30\n",
      "1500/1500 - 5s - loss: 0.1400 - sparse_categorical_accuracy: 0.9467 - val_loss: 0.3906 - val_sparse_categorical_accuracy: 0.8958\n",
      "Epoch 26/30\n",
      "1500/1500 - 5s - loss: 0.1374 - sparse_categorical_accuracy: 0.9476 - val_loss: 0.4095 - val_sparse_categorical_accuracy: 0.8901\n",
      "Epoch 27/30\n",
      "1500/1500 - 5s - loss: 0.1326 - sparse_categorical_accuracy: 0.9490 - val_loss: 0.4063 - val_sparse_categorical_accuracy: 0.8938\n",
      "Epoch 28/30\n",
      "1500/1500 - 5s - loss: 0.1298 - sparse_categorical_accuracy: 0.9506 - val_loss: 0.4215 - val_sparse_categorical_accuracy: 0.8924\n",
      "Epoch 29/30\n",
      "1500/1500 - 5s - loss: 0.1279 - sparse_categorical_accuracy: 0.9515 - val_loss: 0.4332 - val_sparse_categorical_accuracy: 0.8946\n",
      "Epoch 30/30\n",
      "1500/1500 - 5s - loss: 0.1236 - sparse_categorical_accuracy: 0.9525 - val_loss: 0.4637 - val_sparse_categorical_accuracy: 0.8888\n",
      "157/157 [==============================] - 0s 2ms/step - loss: 0.5132 - sparse_categorical_accuracy: 0.8782\n",
      "0.8781999945640564 0.5131762623786926 (5000, 28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='relu'),\n",
    "  tf.keras.layers.Dense(100,activation='relu'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.Adam(0.001),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    x_train,y_train,\n",
    "    epochs=30,\n",
    "    validation_split=0.2, verbose=2\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 455
    },
    "id": "7F-8SnejdBvq",
    "outputId": "2749f826-4e3d-4a6e-a49e-5da1d6bbd9ff"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[432   0   7   7   1   1  57   0   2   0]\n",
      " [  1 473   1   6   0   0   0   0   0   0]\n",
      " [  6   1 421   6  40   0  47   0   0   0]\n",
      " [ 21   3   6 431  18   1  16   0   4   0]\n",
      " [  0   0  77  19 389   0  35   0   1   0]\n",
      " [  0   0   0   0   0 476   0   5   0   4]\n",
      " [ 61   1  35  11  18   0 353   0   3   0]\n",
      " [  0   0   0   0   0  19   0 473   0   8]\n",
      " [  4   0   3   3   2   3   6   1 504   0]\n",
      " [  0   0   0   0   0  12   0  26   0 439]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f779d2970f0>"
      ]
     },
     "execution_count": 38,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "predict = model.predict(test1X)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vb9pQNCfgwMi"
   },
   "source": [
    "\n",
    "\n",
    "1.   Hidden layer: relu\n",
    "2.   optimizer: RMSprop\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "mYO7U3PMrk3z",
    "outputId": "f369c329-8602-467d-c910-d6a76cbbbbf4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1500/1500 - 7s - loss: 0.5134 - sparse_categorical_accuracy: 0.8123 - val_loss: 0.4741 - val_sparse_categorical_accuracy: 0.8169\n",
      "Epoch 2/30\n",
      "1500/1500 - 7s - loss: 0.3940 - sparse_categorical_accuracy: 0.8586 - val_loss: 0.4026 - val_sparse_categorical_accuracy: 0.8583\n",
      "Epoch 3/30\n",
      "1500/1500 - 6s - loss: 0.3677 - sparse_categorical_accuracy: 0.8710 - val_loss: 0.3698 - val_sparse_categorical_accuracy: 0.8720\n",
      "Epoch 4/30\n",
      "1500/1500 - 7s - loss: 0.3530 - sparse_categorical_accuracy: 0.8771 - val_loss: 0.4070 - val_sparse_categorical_accuracy: 0.8690\n",
      "Epoch 5/30\n",
      "1500/1500 - 7s - loss: 0.3411 - sparse_categorical_accuracy: 0.8811 - val_loss: 0.3746 - val_sparse_categorical_accuracy: 0.8768\n",
      "Epoch 6/30\n",
      "1500/1500 - 7s - loss: 0.3385 - sparse_categorical_accuracy: 0.8846 - val_loss: 0.4138 - val_sparse_categorical_accuracy: 0.8738\n",
      "Epoch 7/30\n",
      "1500/1500 - 7s - loss: 0.3240 - sparse_categorical_accuracy: 0.8891 - val_loss: 0.4573 - val_sparse_categorical_accuracy: 0.8603\n",
      "Epoch 8/30\n",
      "1500/1500 - 7s - loss: 0.3220 - sparse_categorical_accuracy: 0.8903 - val_loss: 0.4864 - val_sparse_categorical_accuracy: 0.8731\n",
      "Epoch 9/30\n",
      "1500/1500 - 7s - loss: 0.3217 - sparse_categorical_accuracy: 0.8917 - val_loss: 0.4440 - val_sparse_categorical_accuracy: 0.8716\n",
      "Epoch 10/30\n",
      "1500/1500 - 7s - loss: 0.3156 - sparse_categorical_accuracy: 0.8936 - val_loss: 0.4511 - val_sparse_categorical_accuracy: 0.8765\n",
      "Epoch 11/30\n",
      "1500/1500 - 7s - loss: 0.3150 - sparse_categorical_accuracy: 0.8971 - val_loss: 0.4390 - val_sparse_categorical_accuracy: 0.8752\n",
      "Epoch 12/30\n",
      "1500/1500 - 7s - loss: 0.3042 - sparse_categorical_accuracy: 0.8993 - val_loss: 0.4656 - val_sparse_categorical_accuracy: 0.8785\n",
      "Epoch 13/30\n",
      "1500/1500 - 7s - loss: 0.2995 - sparse_categorical_accuracy: 0.8994 - val_loss: 0.4883 - val_sparse_categorical_accuracy: 0.8794\n",
      "Epoch 14/30\n",
      "1500/1500 - 7s - loss: 0.2984 - sparse_categorical_accuracy: 0.9016 - val_loss: 0.5546 - val_sparse_categorical_accuracy: 0.8753\n",
      "Epoch 15/30\n",
      "1500/1500 - 7s - loss: 0.2943 - sparse_categorical_accuracy: 0.9031 - val_loss: 0.4922 - val_sparse_categorical_accuracy: 0.8797\n",
      "Epoch 16/30\n",
      "1500/1500 - 6s - loss: 0.2933 - sparse_categorical_accuracy: 0.9038 - val_loss: 0.5059 - val_sparse_categorical_accuracy: 0.8763\n",
      "Epoch 17/30\n",
      "1500/1500 - 6s - loss: 0.2842 - sparse_categorical_accuracy: 0.9060 - val_loss: 0.5255 - val_sparse_categorical_accuracy: 0.8760\n",
      "Epoch 18/30\n",
      "1500/1500 - 6s - loss: 0.2819 - sparse_categorical_accuracy: 0.9076 - val_loss: 0.5319 - val_sparse_categorical_accuracy: 0.8773\n",
      "Epoch 19/30\n",
      "1500/1500 - 7s - loss: 0.2836 - sparse_categorical_accuracy: 0.9071 - val_loss: 0.7057 - val_sparse_categorical_accuracy: 0.8765\n",
      "Epoch 20/30\n",
      "1500/1500 - 7s - loss: 0.2775 - sparse_categorical_accuracy: 0.9097 - val_loss: 0.5309 - val_sparse_categorical_accuracy: 0.8816\n",
      "Epoch 21/30\n",
      "1500/1500 - 7s - loss: 0.2732 - sparse_categorical_accuracy: 0.9114 - val_loss: 0.5849 - val_sparse_categorical_accuracy: 0.8799\n",
      "Epoch 22/30\n",
      "1500/1500 - 7s - loss: 0.2792 - sparse_categorical_accuracy: 0.9126 - val_loss: 0.6467 - val_sparse_categorical_accuracy: 0.8754\n",
      "Epoch 23/30\n",
      "1500/1500 - 7s - loss: 0.2686 - sparse_categorical_accuracy: 0.9137 - val_loss: 0.5708 - val_sparse_categorical_accuracy: 0.8789\n",
      "Epoch 24/30\n",
      "1500/1500 - 7s - loss: 0.2660 - sparse_categorical_accuracy: 0.9129 - val_loss: 0.7330 - val_sparse_categorical_accuracy: 0.8537\n",
      "Epoch 25/30\n",
      "1500/1500 - 7s - loss: 0.2587 - sparse_categorical_accuracy: 0.9151 - val_loss: 0.6461 - val_sparse_categorical_accuracy: 0.8802\n",
      "Epoch 26/30\n",
      "1500/1500 - 7s - loss: 0.2591 - sparse_categorical_accuracy: 0.9161 - val_loss: 0.6645 - val_sparse_categorical_accuracy: 0.8842\n",
      "Epoch 27/30\n",
      "1500/1500 - 7s - loss: 0.2584 - sparse_categorical_accuracy: 0.9169 - val_loss: 0.7841 - val_sparse_categorical_accuracy: 0.8810\n",
      "Epoch 28/30\n",
      "1500/1500 - 7s - loss: 0.2581 - sparse_categorical_accuracy: 0.9179 - val_loss: 0.8070 - val_sparse_categorical_accuracy: 0.8790\n",
      "Epoch 29/30\n",
      "1500/1500 - 7s - loss: 0.2493 - sparse_categorical_accuracy: 0.9195 - val_loss: 0.8239 - val_sparse_categorical_accuracy: 0.8822\n",
      "Epoch 30/30\n",
      "1500/1500 - 7s - loss: 0.2493 - sparse_categorical_accuracy: 0.9212 - val_loss: 0.7427 - val_sparse_categorical_accuracy: 0.8842\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f779b83cac8>"
      ]
     },
     "execution_count": 81,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='relu'),\n",
    "  tf.keras.layers.Dense(100,activation='relu'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.RMSprop(0.001),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    x_train,y_train,\n",
    "    epochs=30,\n",
    "    validation_split=0.2, verbose=2\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 707
    },
    "id": "gDSwyMe_SdD-",
    "outputId": "3ee8ddd6-a3e8-48e4-d694-7b4b296de5f1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "157/157 [==============================] - 0s 2ms/step - loss: 2.7717 - sparse_categorical_accuracy: 0.5580\n",
      "0.5580000281333923 2.771714448928833 (5000, 28, 28)\n",
      "[[0.0000000e+00 0.0000000e+00 0.0000000e+00 ... 4.6999546e-04\n",
      "  0.0000000e+00 9.9869710e-01]\n",
      " [1.2564442e-01 1.6285967e-02 1.9412297e-01 ... 5.1031090e-02\n",
      "  7.8216381e-02 2.6164798e-02]\n",
      " [0.0000000e+00 1.0000000e+00 0.0000000e+00 ... 0.0000000e+00\n",
      "  0.0000000e+00 0.0000000e+00]\n",
      " ...\n",
      " [0.0000000e+00 0.0000000e+00 0.0000000e+00 ... 0.0000000e+00\n",
      "  0.0000000e+00 1.0000000e+00]\n",
      " [1.2564442e-01 1.6285967e-02 1.9412297e-01 ... 5.1031090e-02\n",
      "  7.8216381e-02 2.6164798e-02]\n",
      " [0.0000000e+00 0.0000000e+00 0.0000000e+00 ... 1.0000000e+00\n",
      "  0.0000000e+00 0.0000000e+00]]\n",
      "[[ 74   0 424   5   0   3   1   0   0   0]\n",
      " [  0 444  30   6   0   0   1   0   0   0]\n",
      " [  0   1 520   0   0   0   0   0   0   0]\n",
      " [  0   9 200 289   0   0   1   0   1   0]\n",
      " [  0   0 519   0   0   1   0   0   1   0]\n",
      " [  0   0  59   1   0 412   0   9   1   3]\n",
      " [  4   0 478   0   0   0   0   0   0   0]\n",
      " [  0   0 141   0   0   1   0 355   0   3]\n",
      " [  0   1 215   3   0   1   0   0 306   0]\n",
      " [  0   0  82   0   0   1   0   4   0 390]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fb52038f4e0>"
      ]
     },
     "execution_count": 16,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)\n",
    "predict = model.predict(test1X)\n",
    "print(predict)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bOATwBHMh2vR"
   },
   "source": [
    "\n",
    "\n",
    "1.   Hidden layer: relu\n",
    "2.   optimizer: Nadam\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "rqZWPaIvT6Gn",
    "outputId": "a3682fcf-5ad8-4a96-cb3c-2a36a2df40dc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 7.2198 - sparse_categorical_accuracy: 0.7068 - val_loss: 0.5997 - val_sparse_categorical_accuracy: 0.7980\n",
      "Epoch 2/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.5066 - sparse_categorical_accuracy: 0.8261 - val_loss: 0.4799 - val_sparse_categorical_accuracy: 0.8310\n",
      "Epoch 3/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.4489 - sparse_categorical_accuracy: 0.8440 - val_loss: 0.5084 - val_sparse_categorical_accuracy: 0.8308\n",
      "Epoch 4/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.4413 - sparse_categorical_accuracy: 0.8491 - val_loss: 0.4368 - val_sparse_categorical_accuracy: 0.8444\n",
      "Epoch 5/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.4021 - sparse_categorical_accuracy: 0.8583 - val_loss: 0.4234 - val_sparse_categorical_accuracy: 0.8484\n",
      "Epoch 6/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.3850 - sparse_categorical_accuracy: 0.8623 - val_loss: 0.4355 - val_sparse_categorical_accuracy: 0.8500\n",
      "Epoch 7/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.3656 - sparse_categorical_accuracy: 0.8679 - val_loss: 0.4451 - val_sparse_categorical_accuracy: 0.8522\n",
      "Epoch 8/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.3551 - sparse_categorical_accuracy: 0.8711 - val_loss: 0.4183 - val_sparse_categorical_accuracy: 0.8554\n",
      "Epoch 9/30\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3513 - sparse_categorical_accuracy: 0.8747 - val_loss: 0.4300 - val_sparse_categorical_accuracy: 0.8528\n",
      "Epoch 10/30\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3356 - sparse_categorical_accuracy: 0.8770 - val_loss: 0.4036 - val_sparse_categorical_accuracy: 0.8628\n",
      "Epoch 11/30\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3254 - sparse_categorical_accuracy: 0.8802 - val_loss: 0.4111 - val_sparse_categorical_accuracy: 0.8658\n",
      "Epoch 12/30\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3231 - sparse_categorical_accuracy: 0.8833 - val_loss: 0.4239 - val_sparse_categorical_accuracy: 0.8602\n",
      "Epoch 13/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.3147 - sparse_categorical_accuracy: 0.8882 - val_loss: 0.4157 - val_sparse_categorical_accuracy: 0.8708\n",
      "Epoch 14/30\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3104 - sparse_categorical_accuracy: 0.8889 - val_loss: 0.4012 - val_sparse_categorical_accuracy: 0.8706\n",
      "Epoch 15/30\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3034 - sparse_categorical_accuracy: 0.8915 - val_loss: 0.4148 - val_sparse_categorical_accuracy: 0.8720\n",
      "Epoch 16/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.3110 - sparse_categorical_accuracy: 0.8864 - val_loss: 0.4503 - val_sparse_categorical_accuracy: 0.8660\n",
      "Epoch 17/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.3049 - sparse_categorical_accuracy: 0.8898 - val_loss: 0.4413 - val_sparse_categorical_accuracy: 0.8626\n",
      "Epoch 18/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2992 - sparse_categorical_accuracy: 0.8916 - val_loss: 0.4197 - val_sparse_categorical_accuracy: 0.8738\n",
      "Epoch 19/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.3045 - sparse_categorical_accuracy: 0.8903 - val_loss: 0.4443 - val_sparse_categorical_accuracy: 0.8650\n",
      "Epoch 20/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2874 - sparse_categorical_accuracy: 0.8950 - val_loss: 0.3932 - val_sparse_categorical_accuracy: 0.8782\n",
      "Epoch 21/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2860 - sparse_categorical_accuracy: 0.8956 - val_loss: 0.4495 - val_sparse_categorical_accuracy: 0.8712\n",
      "Epoch 22/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2809 - sparse_categorical_accuracy: 0.8959 - val_loss: 0.4403 - val_sparse_categorical_accuracy: 0.8808\n",
      "Epoch 23/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2713 - sparse_categorical_accuracy: 0.8993 - val_loss: 0.4613 - val_sparse_categorical_accuracy: 0.8770\n",
      "Epoch 24/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2729 - sparse_categorical_accuracy: 0.8975 - val_loss: 0.5390 - val_sparse_categorical_accuracy: 0.8726\n",
      "Epoch 25/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2821 - sparse_categorical_accuracy: 0.8970 - val_loss: 0.5248 - val_sparse_categorical_accuracy: 0.8716\n",
      "Epoch 26/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2723 - sparse_categorical_accuracy: 0.9009 - val_loss: 0.5103 - val_sparse_categorical_accuracy: 0.8762\n",
      "Epoch 27/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.2699 - sparse_categorical_accuracy: 0.9013 - val_loss: 0.5813 - val_sparse_categorical_accuracy: 0.8668\n",
      "Epoch 28/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.2676 - sparse_categorical_accuracy: 0.9022 - val_loss: 0.6254 - val_sparse_categorical_accuracy: 0.8592\n",
      "Epoch 29/30\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.2708 - sparse_categorical_accuracy: 0.9011 - val_loss: 0.6120 - val_sparse_categorical_accuracy: 0.8744\n",
      "Epoch 30/30\n",
      "1875/1875 [==============================] - 11s 6ms/step - loss: 0.2571 - sparse_categorical_accuracy: 0.9052 - val_loss: 0.7349 - val_sparse_categorical_accuracy: 0.8534\n",
      "157/157 [==============================] - 0s 1ms/step - loss: 0.5273 - sparse_categorical_accuracy: 0.8448\n",
      "0.8447999954223633 0.5272934436798096 (5000, 28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD6CAYAAACiefy7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd3xUVfr48c+ZlkmZIT0EEoqKIKRQA4pS1oLdRVf5YkFUYFUsq67ruhYs6LqL61p3Ff1hwwK2VVdWQQURRSVgNICISE0I6WXSJlOe3x8zGRJISIAJCZPzfr3ua26/584kzz1z5t7nKBFB0zRNC32Gzi6ApmmadmTogK9pmtZN6ICvaZrWTeiAr2ma1k3ogK9pmtZN6ICvaZrWTbQZ8JVSqUqp5UqpjUqpDUqpm1tY5zKl1I9KqVyl1NdKqcwmy7b75+copbKDfQKapmla+5jasY4buE1E1imlbMBapdQyEdnYZJ1twHgRKVdKnQXMB0Y3WT5RREraW6j4+Hjp169fe1fXNE3r9tauXVsiIgkHWqfNgC8iBUCBf9yhlPoJ6A1sbLLO1002+QZIOaQS+/Xr14/sbP1lQNM0rb2UUjvaWueg2vCVUv2AYcC3B1jtGuB/TaYFWKqUWquUmnUwx9M0TdOCpz1NOgAopaKAd4A/iEhVK+tMxBfwT24y+2QRyVdKJQLLlFKbRGRlC9vOAmYB9OnT5yBOQdM0TWuPdtXwlVJmfMH+NRF5t5V1MoAXgAtEpLRxvojk+1+LgPeArJa2F5H5IjJSREYmJBywGUrTNE07BG3W8JVSCvh/wE8i8lgr6/QB3gWuEJHNTeZHAgZ/238kcAbwQFBKroU8l8tFXl4e9fX1nV0UTesyrFYrKSkpmM3mg962PU06Y4ErgFylVI5/3l+APgAi8ixwLxAH/Mt3fcAtIiOBJOA9/zwT8LqIfHzQpdS6pby8PGw2G/369cP/N6Rp3ZqIUFpaSl5eHv379z/o7dtzl84q4ID/bSIyA5jRwvytQOb+W2ha2+rr63Ww17QmlFLExcVRXFx8SNvrJ221Lk0He01r7nD+J0Iq4D/52S98sfnQrnyapmmhLqQC/vyVW/niZx3wteCJiorq7CJoWtCEVMC3W01U1bs6uxiapmldUkgFfJvVjEMHfK0DiAi33347aWlppKens2jRIgAKCgoYN24cQ4cOJS0tjS+//BKPx8P06dMD6/7zn//s5NJrmk+7n7Q9GtjDTTjq3Z1dDK0D3P/hBjbubvEB70M2uJedOecNade67777Ljk5Ofzwww+UlJQwatQoxo0bx+uvv86kSZO466678Hg81NbWkpOTQ35+PuvXrwegoqIiqOXWtEMVcjV83aSjdYRVq1YxdepUjEYjSUlJjB8/njVr1jBq1ChefPFF7rvvPnJzc7HZbBxzzDFs3bqVG2+8kY8//hi73d7Zxdc0IMRq+DariV+LdQ0/FLW3Jn6kjRs3jpUrV/LRRx8xffp0br31VqZNm8YPP/zAJ598wrPPPsvixYtZsGBBZxdV00Krhm+3mqmq0zV8LfhOOeUUFi1ahMfjobi4mJUrV5KVlcWOHTtISkpi5syZzJgxg3Xr1lFSUoLX6+Wiiy5i7ty5rFu3rrOLr2lACNbwHfVuREQ/sKMF1eTJk1m9ejWZmZkopfj73/9Oz549efnll5k3bx5ms5moqCheeeUV8vPzueqqq/B6vQD89a9/7eTSa5pPSAV8e7gZt1eoc3mIsITUqWmdpLq6GvA93Thv3jzmzZvXbPmVV17JlVdeud92ulavdUUh1aRjs/qCvL5TR9M0bX8hFfDtVl+6UN2Or2matr+QCviNNfwqXcPXNE3bT0gFfHu4v4av78XXNE3bT2gFfN2Gr2ma1qoQC/i6DV/TNK01bQZ8pVSqUmq5UmqjUmqDUurmFtZRSqknlVJblFI/KqWGN1l2pVLqF/+w//1rQWTzB3xdw9c0Tdtfe2r4buA2ERkMjAFmK6UG77POWcAA/zAL+DeAUioWmAOMBrKAOUqpmCCVfT9WswGzUek2fE3zy8nJYcmSJUfkWDNmzGDjxo0Hvd2KFSs499xzO6BE2r7aDPgiUiAi6/zjDuAnoPc+q10AvCI+3wDRSqlkYBKwTETKRKQcWAacGdQzaEIppVMka12W233kv3keqYDv8Xh44YUXGDx437rg0cXj8XR2ETrUQbXhK6X6AcOAb/dZ1BvY1WQ6zz+vtfkdxm41UVWnm3RCzv/+DC+eE9zhf38+4CFramo455xzyMzMJC0tjUWLFtGvXz/+9Kc/kZ6eTlZWFlu2bAHgww8/ZPTo0QwbNozTTjuNwsJCAO677z6uuOIKxo4dyxVXXMGGDRvIyspi6NChZGRk8MsvvwCwcOHCwPzf//73Bww8H3/8McOHDyczM5NTTz0VgO+++44TTzyRYcOGcdJJJ/Hzzz/T0NDAvffey6JFixg6dCiLFi2ipqaGq6++mqysLIYNG8b7778PQG1tLZdccgmDBw9m8uTJjB49muzsbADeeOMN0tPTSUtL44477giUIyoqittuu43MzExWr17NhAkTAtu0t4zt0dp2Ho+HP/7xj6SlpZGRkcFTTz0FwJo1azjppJPIzMwkKysLh8PBSy+9xA033BDY57nnnsuKFStaPI8HHniAUaNGkZaWxqxZsxARALZs2cJpp51GZmYmw4cP59dff2XatGn85z//Cez3sssuC7ynXZKItGsAooC1wIUtLPsvcHKT6c+AkcAfgbubzL8H+GMr+58FZAPZffr0kUN17pNfyvQF3x7y9lrXsXHjxr0TS+4QWXB2cIcldxzw+G+//bbMmDEjMF1RUSF9+/aVuXPniojIyy+/LOecc46IiJSVlYnX6xURkeeff15uvfVWERGZM2eODB8+XGpra0VE5IYbbpCFCxeKiIjT6ZTa2lrZuHGjnHvuudLQ0CAiItddd528/PLLLZapqKhIUlJSZOvWrSIiUlpaKiIilZWV4nK5RERk2bJlcuGFF4qIyIsvviizZ88ObH/nnXfKq6++KiIi5eXlMmDAAKmurpZ58+bJrFmzREQkNzdXjEajrFmzRvLz8yU1NVWKiorE5XLJxIkT5b333hMREUAWLVoU2Pf48eNlzZo1B13G5cuXB97HlrS23b/+9S+56KKLAstKS0vF6XRK//795bvvvmu27b7vwznnnCPLly9v8Twayysicvnll8sHH3wgIiJZWVny7rvviohIXV2d1NTUyIoVK+SCCy4QEd/fR79+/QLl6UjN/jf8gGxpI463K+GMUsoMvAO8JiLvtrBKPpDaZDrFPy8fmLDP/BWtXHjmA/MBRo4cKe0pV0vs4Sb94FUoOuuRI37I9PR0brvtNu644w7OPfdcTjnlFACmTp0aeL3lllsAyMvLY8qUKRQUFNDQ0ED//v0D+zn//PMJDw8H4MQTT+Shhx4iLy+PCy+8kAEDBvDZZ5+xdu1aRo0aBUBdXR2JiYktlumbb75h3Lhxgf3HxsYCUFlZyZVXXskvv/yCUgqXq+VmzaVLl/LBBx/w6KOPAlBfX8/OnTtZtWoVN9/sux+jscYMvtryhAkTSEhIAHw12JUrV/Lb3/4Wo9HIRRddFPQy7qu17T799FOuvfZaTCZT4Di5ubkkJycH3sv29EWw73ksX76cv//979TW1lJWVsaQIUOYMGEC+fn5TJ48GQCr1QrA+PHjuf766ykuLuadd97hoosuCpSnK2rPXToK+H/ATyLyWCurfQBM89+tMwaoFJEC4BPgDKVUjP/H2jP88zqMLUy34WvBcfzxx7Nu3TrS09O5++67eeCBBwCaZWJtHL/xxhu54YYbyM3N5bnnnqO+vj6wTmRkZGD80ksv5YMPPiA8PJyzzz6bzz//HBHhyiuvJCcnh5ycHH7++Wfuu+++gyrrPffcw8SJE1m/fj0ffvhhs+M3JSK88847gWPt3LmTE0444aCO1chqtWI0GoNexmBt15TJZApkLwWa7aPpedTX13P99dfz9ttvk5uby8yZM9s83rRp01i4cCEvvvgiV1999UGX7UhqTxv+WOAK4DdKqRz/cLZS6lql1LX+dZYAW4EtwPPA9QAiUgY8CKzxDw/453UYe7huw9eCY/fu3URERHD55Zdz++23BzJgNvZnu2jRIk488UTAVwvt3dv389TLL7/c6j63bt3KMcccw0033cQFF1zAjz/+yKmnnsrbb79NUVERAGVlZezYsaPF7ceMGcPKlSvZtm1bYN19j//SSy8F1rfZbDgcjsD0pEmTeOqppwLt0t9//z0AY8eOZfHixQBs3LiR3NxcALKysvjiiy8oKSnB4/HwxhtvMH78+AO+bwdbxra0tt3pp5/Oc889F/gxvKysjIEDB1JQUMCaNWsAcDgcuN1u+vXrR05ODl6vl127dvHdd9+1eKzG4B4fH091dTVvv/024HsfU1JSAu31TqeT2tpaAKZPn87jjz8O0OV/tG7PXTqrRESJSIaIDPUPS0TkWRF51r+OiMhsETlWRNJFJLvJ9gtE5Dj/8GJHngzojsy14MnNzQ38kHr//fdz9913A1BeXk5GRgZPPPFEoIPy++67j4svvpgRI0YQHx/f6j4XL15MWloaQ4cOZf369UybNo3Bgwczd+5czjjjDDIyMjj99NMpKChocfuEhATmz5/PhRdeSGZmJlOmTAHgT3/6E3feeSfDhg1rdjfQxIkT2bhxY+BH23vuuQeXy0VGRgZDhgzhnnvuAQg0SwwePJi7776bIUOG0KNHD5KTk3nkkUeYOHEimZmZjBgxggsuuOCA79vBlrEtrW03Y8YM+vTpQ0ZGBpmZmbz++utYLBYWLVrEjTfeSGZmJqeffjr19fWMHTuW/v37M3jwYG666SaGDx/e4rGio6OZOXMmaWlpTJo0KdA0BPDqq6/y5JNPkpGRwUknncSePXsASEpK4oQTTuCqq65q9zl1FtV4pe9KRo4cKY2/9h+sJz79hX9+upktD52FyRhSDxJ3Oz/99NMhNzd0lH79+pGdnX3AoH408ng8uFwurFYrv/76K6eddho///wzFouls4vW5dXW1pKens66devo0aPHETlmS/8bSqm1IjLyQNt13V8XDlFjxsxqp5voCP3HqmntUVtby8SJE3G5XIgI//rXv3Swb4dPP/2Ua665hltuueWIBfvDEbIB31GvA74WfNu3bz9ixxo9ejROp7PZvFdffZX09PSgH8tms3Go36qD5cUXX+SJJ55oNm/s2LE888wznVSitp122mmt/t7SFYVcwG9MkVxZ52p2n6imHW2+/Xbf5xtD21VXXXVUtIMfzUKukVt3c6hpmtaykAv4gRTJ+k4dTdO0ZkI24OsavqZpWnOhF/DD/f3a6k5QNE3Tmgm5gB8Vptvwtc4RFRXV6rLt27eTlpZ2BEujafsLuYBvMhqItBh1G76mado+Qu62TNDpFULR3777G5vKNgV1n4NiB3FH1h2tLv/zn/9Mamoqs2fPBnzpE0wmE8uXL6e8vByXy8XcuXPbTDWwr/r6eq677jqys7MxmUw89thjTJw4kQ0bNnDVVVfR0NCA1+vlnXfeoVevXlxyySXk5eXh8Xi45557AqkKNO1ghWTA1wnUtGCYMmUKf/jDHwIBf/HixXzyySfcdNNN2O12SkpKGDNmDOeff36zDJpteeaZZ1BKkZuby6ZNmzjjjDPYvHkzzz77LDfffDOXXXYZDQ0NeDwelixZQq9evfjoo48AXyIxTTtUIRnwbVYzDqeu4YeSA9XEO8qwYcMoKipi9+7dFBcXExMTQ8+ePbnllltYuXIlBoOB/Px8CgsL6dmzZ7v3u2rVKm688UYABg0aRN++fdm8eXOLufJby8mvaYci5NrwQXdzqAXPxRdfzNtvv82iRYuYMmUKr732GsXFxaxdu5acnBySkpIOKT97S1rKld9aTn5NOxQhW8PfVlLT2cXQQsCUKVOYOXMmJSUlfPHFFyxevJjExETMZjPLly8/pDwqp5xyCq+99hq/+c1v2Lx5Mzt37mTgwIHNcuXv3LmTH3/8kUGDBhEbG8vll19OdHQ0L7zwQgecpdZdhGTA190casEyZMgQHA4HvXv3Jjk5mcsuu4zzzjuP9PR0Ro4cyaBBgw56n9dffz3XXXcd6enpmEwmXnrpJcLCwli8eDGvvvoqZrOZnj178pe//IU1a9Zw++23YzAYMJvN/Pvf/+6As9S6i5DLhw/wt4838cKXW9k896yD+jFN61q6Yj58TesKDjUffnv6tF2glCpSSq1vZfntTbo+XK+U8iilYv3Ltiulcv3LjljuVbvVjMsj1Lu8ba+saZrWTbSnSecl4GnglZYWisg8YB6AUuo84JZ9+q2dKCIlh1nOg7I3Y6aLcEv7O1nWtMOVm5vLFVdc0WxeWFhYt0t1rHVNbQZ8EVmplOrXzv1NBd44nAIFQ2NO/Kp6F4l2ayeXRutO0tPTycnJ6exiaFqLgnZbplIqAjgTeKfJbAGWKqXWKqVmtbH9LKVUtlIqu7i4+LDK0ljD1z/capqm7RXM+/DPA77apznnZBEZDpwFzFZKjWttYxGZLyIjRWRkQkLCYRVEp0jWNE3bXzAD/v+xT3OOiOT7X4uA94CsIB6vVXarTpGsaZq2r6AEfKVUD2A88H6TeZFKKVvjOHAG0OKdPsFm0zV8TdO0/bTntsw3gNXAQKVUnlLqGqXUtUqpa5usNhlYKiJNH29NAlYppX4AvgM+EpGPg1n41gQ6QdEZM7Uj6ED58EPFihUr+Prrr4/Isc4++2wqKioOeruXXnqJG264oQNKdPRrz106U9uxzkv4bt9sOm8rkHmoBTsc4WYjRoPSKZK1kOZ2uzGZjuzD8itWrCAqKoqTTjqpw44hIogIS5Ys6bBjHAmN52EwdJ2UZSGZWkEppROohZg9Dz+M86fg5sMPO2EQPf/yl1aXBzMffkFBAVOmTKGqqgq3282///1vTjnlFKKiopg5cyZLly6lZ8+evPnmmyQkJPD8888zf/58GhoaOO6443j11VeJiIhg+vTpWK1Wvv/+e8aOHcsFF1zAzTffDPj+7leuXInNZmPevHksXrwYp9PJ5MmTuf/++1st2yuvvMKjjz6KUoqMjAxeffVVPvzwQ+bOnUtDQwNxcXG89tpr1NXV8eyzz2I0Glm4cCFPPfUUgwYN4tprr2Xnzp0APP7444wdO5bi4mIuvfRSdu/ezYknnsiyZctYu3Yt8fHxPPbYYyxYsACAGTNm8Ic//IHt27czadIkRo8ezdq1a1myZAnjx48nOzub+Pj4dpcxKSmpzc+ite2qq6u58cYbyc7ORinFnDlzuOiii/j444/5y1/+gsfjIT4+ns8++4z77ruPqKgo/vjHPwKQlpbGf//7X4D9zuORRx5hzZo11NXV8bvf/S7wWaxZs4abb76ZmpoawsLC+OyzzzjnnHN48sknGTp0KAAnn3wyzzzzDJmZQao7N16FutIwYsQIOVyn/O1zufmNdYe9H63zbNy4MTBe8NBDsv3yK4I6FDz00AGPv27dOhk3blxg+oQTTpCdO3dKZWWliIgUFxfLscceK16vV0REIiMjW93Xo48+KnPnzhUREbfbLVVVVSIiAsjChQtFROT++++X2bNni4hISUlJYNu77rpLnnzySRERufLKK+Wcc84Rt9stIiLnnnuurFq1SkREHA6HuFwu+eSTT2TmzJni9XrF4/HIOeecI1988UWL5Vq/fr0MGDBAiouLRUSktLRURETKysoC5/X888/LrbfeKiIic+bMkXnz5gW2nzp1qnz55ZciIrJjxw4ZNGiQiIjMnj1bHn74YRER+d///ieAFBcXS3Z2tqSlpUl1dbU4HA4ZPHiwrFu3TrZt2yZKKVm9enVg33379pXi4uKDLuOLL74YeB9b0tp2f/rTn+Tmm29utl5RUZGkpKTI1q1bmx173/dhyJAhsm3bthbPo3Ebt9st48ePlx9++EGcTqf0799fvvvuOxERqaysFJfLJS+99FKgDD///LO0Fgub/m80ArKljdgakjV80AnUQs2BauIdJZj58EeNGsXVV1+Ny+Xit7/9baAGZzAYAj1YXX755Vx44YUArF+/nrvvvpuKigqqq6uZNGlSYF8XX3wxRqPvCfKxY8dy6623ctlll3HhhReSkpLC0qVLWbp0KcOGDQOgurqaX375hXHj9r8r+vPPP+fiiy8mPj4egNjYWADy8vKYMmUKBQUFNDQ00L9//xbP69NPP2Xjxo2B6aqqKqqrq1m1ahXvvfceAGeeeSYxMTGAry+AyZMnExkZCcCFF17Il19+yfnnn0/fvn0ZM2ZM0Mu4r9a2+/TTT3nzzTcD68XExPDhhx8ybty4wDqNxz6Qfc9j8eLFzJ8/H7fbTUFBARs3bkQpRXJyMqNGjQLAbrcDvs/2wQcfZN68eSxYsIDp06e365zaq+s0LgWZLUx3c6gdvmDlwx83bhwrV66kd+/eTJ8+nVdeaTFTSSDZ3/Tp03n66afJzc1lzpw5zY7RGCzB1+z0wgsvUFdXx9ixY9m0aRMiwp133klOTg45OTls2bKFa6655qDO+8Ybb+SGG24gNzeX5557rtVz9Hq9fPPNN4Fj5efnH/KP103PK5hlDNZ2TZlMJrzevbm6Wvt8tm3bxqOPPspnn33Gjz/+yDnnnHPA40VERHD66afz/vvvs3jxYi677LKDLtuBhGzA190casEwZcoU3nzzTd5++20uvvhiKisrDykf/o4dO0hKSmLmzJnMmDGDdevWAb6A+fbbbwPw+uuvc/LJJwPgcDhITk7G5XLx2muvtbrfX3/9lfT0dO644w5GjRrFpk2bmDRpEgsWLKC6uhqA/Px8ioqKWtz+N7/5DW+99RalpaUAlJX5npusrKykd+/eALz88suB9W02Gw6HIzB9xhln8NRTTwWmG9NKjB07lsWLFwOwdOlSysvLAV9fAP/5z3+ora2lpqaG9957r81evA62jG1pbbvTTz+dZ555JjBdXl7OmDFjWLlyJdu2bWt27H79+gU+w3Xr1gWW76uqqorIyEh69OhBYWEh//vf/wAYOHAgBQUFrFmzBvB93m63L17NmDGDm266iVGjRgW+GQVLyAZ83ZG5Fgwt5cPPzs4mPT2dV155pd358FesWEFmZibDhg1j0aJFgR9aIyMj+e6770hLS+Pzzz/n3nvvBeDBBx9k9OjRjB079oDHePzxx0lLSyMjIwOz2cxZZ53FGWecwaWXXsqJJ55Ieno6v/vd75oF6X3P76677mL8+PFkZmZy6623Ar4fqC+++GJGjBgRaEoBOO+883jvvfcYOnQoX375JU8++STZ2dlkZGQwePBgnn32WQDmzJnD0qVLSUtL46233qJnz57YbDaGDx/O9OnTycrKYvTo0cyYMSPQ9HSgz+BgytiW1ra7++67KS8vJy0tjczMTJYvX05CQgLz58/nwgsvJDMzM9D8dtFFF1FWVsaQIUN4+umnOf7441s8VuNnPmjQIC699FLGjh0LgMViYdGiRdx4441kZmZy+umnB2r+I0aMwG63c9VVV7X7nNorJPPhAzzw4UYWZ+9i/f2T2l5Z65K6Qz78qKioQE08lDidToxGIyaTidWrV3PdddfppHLttHv3biZMmMCmTZtavaXzUPPhh+yPtjariWqnG49XMBp0JyiadiTt3LmTSy65BK/Xi8Vi4fnnn+/sIh0VXnnlFe666y4ee+yxDrl/P2QDfmOK5Op6Nz0izJ1cGq27ONh8+Eeqdl9aWsqpp5663/zPPvuMuLi4oB9vwIABfP/990Hf78F46KGHeOutt5rNu/jii7nrrrs6qURtmzZtGtOmTeuw/YdswN+bItmlA/5RTESOqm4qu2o+/Li4uC5Zro501113dengfqgOpxk+ZH+0bUyRrPPpHL2sViulpaWH9QeuaaFERCgtLcVqPbSOnUK2hm8PdHOob808WqWkpJCXl8fhdoijaaHEarWSkpJySNuGbsBv7OZQ58Q/apnN5nY/PalpWttCtknHpmv4mqZpzYRswNdt+Jqmac2FbMCP0jV8TdO0ZkI24JuNBsLNRp1eQdM0za89XRwuUEoVKaVa7I9WKTVBKVWplMrxD/c2WXamUupnpdQWpdSfg1nw9tAJ1DRN0/ZqTw3/JeDMNtb5UkSG+ocHAJRSRuAZ4CxgMDBVKTX4cAp7sGxWMw6nruFrmqZBOwK+iKwEyg5h31nAFhHZKiINwJtA233BBZHu5lDTNG2vYLXhn6iU+kEp9T+l1BD/vN7Aribr5PnntUgpNUspla2Uyg7WgzY6RbKmadpewQj464C+IpIJPAX851B2IiLzRWSkiIxMSEgIQrF8D1/pbg41TdN8Djvgi0iViFT7x5cAZqVUPJAPpDZZNcU/74ixWU26hq9pmuZ32AFfKdVT+dMZKqWy/PssBdYAA5RS/ZVSFuD/gA8O93gHw2416zZ8TdM0vzZz6Sil3gAmAPFKqTxgDmAGEJFngd8B1yml3EAd8H/iS2/oVkrdAHwCGIEFIrKhQ86iFTariQaPl3qXB6vZeCQPrWma1uW0GfBFZGoby58Gnm5l2RJgyaEV7fAFEqjVu3TA1zSt2wvZJ21Bp0jWNE1rKsQDvk6RrGma1iikA75OkaxpmrZXSAf8pm34mqZp3V1IB3xdw9c0TdsrpAO+bsPXNE3bK6QDfoTFiNGgdA1f0zSNEA/4SilsVpNuw9c0TSPEAz405tPRNXxN07SQD/i+fDq6hq9pmhbyAV/X8DVN03y6QcA36zZ8TdM0ukHAt1vNuoavaZpGNwj4+i4dTdM0n5AP+PZwM9VON16vdHZRNE3TOlXoB3yrCRGobtDNOpqmdW/dIODr9AqapmnQjoCvlFqglCpSSq1vZfllSqkflVK5SqmvlVKZTZZt98/PUUplB7Pg7aUTqGmapvm0p4b/EnDmAZZvA8aLSDrwIDB/n+UTRWSoiIw8tCIenkCKZF3D1zStm2tPn7YrlVL9DrD86yaT3wAph1+s4NE1fE3TNJ9gt+FfA/yvybQAS5VSa5VSsw60oVJqllIqW/qZHewAACAASURBVCmVXVxcHLQCBdrw9a2ZmqZ1c23W8NtLKTURX8A/ucnsk0UkXymVCCxTSm0SkZUtbS8i8/E3B40cOTJo91DqGr6maZpPUGr4SqkM4AXgAhEpbZwvIvn+1yLgPSArGMc7GDZ9l46maRoQhICvlOoDvAtcISKbm8yPVErZGseBM4AW7/TpSBaTAavZgMOpa/iapnVvbTbpKKXeACYA8UqpPGAOYAYQkWeBe4E44F9KKQC3/46cJOA9/zwT8LqIfNwB59AmnSJZ0zStfXfpTG1j+QxgRgvztwKZ+29x5OkUyZqmad3gSVvw3Yuv79LRNK276xYB35cTX9fwNU3r3rpFwLdbTTh0G76mad1ctwj4uoavaZrWTQK+PVx3gqJpmtY9Ar7VTIPbS73L09lF0TRN6zTdIuDr9AqapmndJOA3JlBz6GYdTdO6sW4R8HUNX9M0rZsE/EAnKLqGr2laN9YtAr6u4WuapnWTgK87Mtc0TesmAV/X8DVN07pJwI+0mDAo3YavaVr31i0CvsGgiArTKZI1TeveukXAB3+KZN2Gr2laN9ZtAr5OoKZpWnfXroCvlFqglCpSSrXYJ63yeVIptUUp9aNSaniTZVcqpX7xD1cGq+AHy27VCdQ0Teve2lvDfwk48wDLzwIG+IdZwL8BlFKx+PrAHQ1kAXOUUjGHWtjDYbOadRu+pmndWrsCvoisBMoOsMoFwCvi8w0QrZRKBiYBy0SkTETKgWUc+MLRYezhJt2Gr2latxasNvzewK4m03n+ea3N349SapZSKlsplV1cXBykYu1lt5p18jRN07q1LvOjrYjMF5GRIjIyISEh6Pu3W004nG68Xgn6vjVN044GwQr4+UBqk+kU/7zW5h9xNqsZEahp0O34mqZ1T8EK+B8A0/x364wBKkWkAPgEOEMpFeP/sfYM/7wjzh7uS6+gb83UNK27MrVnJaXUG8AEIF4plYfvzhszgIg8CywBzga2ALXAVf5lZUqpB4E1/l09ICIH+vG3w9iadYIS3hlF0DRN61TtCvgiMrWN5QLMbmXZAmDBwRctuPZmzNQ1fE3Tuqcu86NtR9ubMVPfqaNpWvfU7QK+ftpW07TuqtsE/MZuDvXTtpqmdVftasMPBboTFE0LYV4PuJ3gcYK7ATz+oXGe1wNKAQqUwT80HTfsXSYe3/riAa8bvN4m4y3NP9A8/ysKDEZQxn1e/cdunGcOh2Mndtjb1G0CfpjJSJjJoNMraAcm4vsn9bpAvL5p8QKydxqaLxOv/5/dHxAC4y1MI/sfr+WC+PfN/sdvNu0Fj8sX2Nz1/iBX7wt6zaadvvFmgci7f2BqDFiBANUYFI1NAlPTICn7HMvZvAwe5955BhMYLb7BZAFjWJPXsL3LDMa9AdvjamHctc+xnP4yh4DIRLj9lw7bfbcJ+KBTJHe4QLA8QMBrHFr9h276T+1sEkz8gSMw7BPcvK4mx2urDJ6963tczae9IfT3YbSAydrk1dxCLdOwf63T4A8L4vVdABovDuK7GIjXg6fGg7vGg9clGMIsKGsYhrAwlNWOITICFRaGsjQe2x/MxbtPLXyf14Ya8JT5jmk0+7czgyXSfzEw770oNB03hfkuGoFtmsxrvKAYTOx30dz3winiG5Ta+z4E3pN93p+mNfTW5jV9X5FWLrb7fCNQHdvK3q0Cvj08BFMke9xQXwn1FXuDYKu1ItfeINlQ4xtctdBQvXe62XiNLwDKvv/0svcPNVDD9RyBWpbyBS6TL4CJwYLHHYbbaUKk+T+jMhjAEAYqAoxN/1kNiNEEohBlBDEgGBFRgAHB4J+nMFjMGCKsGCPCMERaMYaHoawWlMHoK4tSe2vABrPv2AbT3nLsN21s8R9aPIK7rAJ3STmuknJcxWV46+oxRffAGNMDk38wRvfAEBXpO7emTRD+ACcGM+5yB649JbgKCmnIy8eVl4crL4+GvDy8DgcGux1jjx5NBt+0oUcPjPbGwYanuhp3cTHukhI8JSW4i3zj7pIS3KXl4G56Yaxp+dMKC8NgtaLCw30Xg4gIDOHhviEiHBUegyE8osl0OIbICAzWMP97plBG33umjP5zNhr88xTKYMDrdOJ11OKtrUXq6vDW1OKtLcNbV4e3tnF+LSKCKS4eU3w8poR4jHFxmOITMCX45hmjo337DQLxenEXFtKwfbtvyMvD3KsX4enphA0ahMFiCcpxDkW3CvhdNkWyx+0L2HXlUFfRfLyufG9Ab1xWX7l3vKH68I5tifLVoBoHcyRYo8He21+zMjf5Cr/367x4wV3txO1w4a5y4qlpCMR8cXsRjyBu3+B1exG3xz/tRVnMGCIjMEZGYIiMxBAViSHKhjEqCoPNhsFmx2Czo0xhvgBWWo67tAJ3UQmuokLchUW4i3yDNNQG5zNoL6MRQ1RUoKxGmw1DZCTKavUFtbCwvbXdMCsGaxjK4punLBY85RW49+zBVViIu6DA91pcDJ72XSyVxeILVrGxGONiMUZH4ymv8AX2/HykoaHJygpTYiLm1BQiR4/GGN0DT5UDT2UlnspKGrZtxVNRiaeiAnG1UhEyGDDFxWH0B8awgQMxJSQEAqchMhJxOvHW1SPO+hZfvfV1SF093vp6vHW1eKoduIuKfEHZP0ht8D5HFRGBofHi4h/H66VuVw7ukhKkvr7F8zTGxfouAjHRGKMbhxjfa9N5Mb55eL2+gL5tG87G4L5tOw07diB1dXv3bTLtvUCazViPPx5rehrh6elY09IJO+7YoF1s2tKtAr7d2oEpkl31UFviD877DHUVTaYrmgfwugpocDTblQhNvn0qxBiFWOxgtiNmG2JKRKzHIlGRiCkCMYQjxgjEq/wBV/C6GwOuB6/LH3Bd/sErqLBwVJj/q7fZjLJY9r5azCizb9pTVekLrsXFvqHI9+opKztA+7OPMpt9+w/zBTuDf9/ehhq8jjy81dXNA1QbVHg45sRETElJhA8diikpEXNSEqbERJQljED7eNNy+celyTxlNPlqjIFXI8o/NI5jMPiCUXU1XocDj6Mar6PK/+rAU+3A6x93FRYiTidSX4+3oQGpr/dNtxJEVXg45p49MfVMInLMGEzJPTEn+abNycmYk5IwRETgLq/AU1aKu6TU91paFph2l5XiKS3DuWULxuhowgYMIGriRCypKZhTUjD3TsHcuxeGsLA231cRQerr8VRV4amoxFtVicFm89V8Y2KOSDBqLIO3ttb33nkbmwe9iNfXtCQe3zfJxmm8Xt+FtklgV1ar7xvQAY7jranFU1KMu7QUd7H/W0tJMR7/tKe8HFf+btwVFXirqtr8OwfAaMSc0htLv35Ejs7C0r8/ln79sPTvjykxEXdBAXW566lfn0td7nqq/vsRFW8uAnx/D9bBgwlPS8Oano797LMOeA6HQ0l7TuYIGzlypGRnZwd9v7NfW8emPVV8dtuE9m9UVw6OPb6hugiqG18LkaoC3IWFuApLaCirx11rxOtRiEfhde999XoU4jXh9ZrwegyIR/mCsyhobCXxCnjF/8fcAZ+JyeQLuGFhYDJCgwuvy+ULSq3V7hr5a3mmxERf7a5xSEzElOgbN8bE+muzlr0XkXb80XobGvDW1PgCq3/wOKrx1lQjbg+mxIRAUDfYbCilgvSGdDzxeJCGBryNFwCnE2N0NAa7/ag6j+5MPB7fhbC8Ak9FBZ6K8sA4SCCoW1JSUAfRVCNeLw3bdwQuAPW5udT/9BPG6GgGfLHikMqqlForIiMPtE73quGHm6iqdyMi/g8MaKiFqnyozENV5kHlLqjc6X/NA2cV7jojDdVGXNUmGmqMuGosNNRacDkU4hZ8uXn8+XkMyv/jVRiGcCuG8AhUeCSG8HCM4VbM1vDA13tlNKFMJpTJ6K9tGsFk9M9vrG2afMHT3PhqRplM4H9VZov/1eRrOgiz+I9v9dWq/bVrZWr9oxavF3G7kYYGxOXyvfrHjTYbxtjYDqvlGfw1f2I6pSO0DqWMRl+7dLjO3XS0UkYjppgYTEH++1QGA2HH9CfsmP70OP98AMTlwlVYFNTj7KvbBHwRoU/hNiZ/9wm/TrgXV2F5O7aK9A97qfBwLH36EDY4laiUVMypKVhS+2BJTcGUnOwLrkdZ7U0ZDL7aSSf+mKRp3Z0ym7GktNg/VNCEdMAXr5e677+n6pNPcCz7lAkFBYgCS896Yk6OQcWlQnisf4jxvVqiQNGs7dcUF+9rG01N9dV2j7KArmmaBiEY8MXtpjY72xfkP/0UT3EJymIhcvgJWI7Np0dCCd5Tbydi4h/BGHKnr2ma1qqQiXjeujoKH34Yx6ef4SkvR4WHEzVuHLZTJxLFaozrnsUR2ZdLyubwxNBr6K+DvaZp3UzIRD1ltVK3fgORJ52EbdIZRJ1yCgbHdnhnJhTmwsirWdPnJn58faNOkaxpWrcUOgFfKfq/+46vfd3rhe/mw7J7IcwGU9+EgWcRtc3X2ZbuBEXTtO6ovV0cngk8ARiBF0TkkX2W/xNoTPEWASSKSLR/mQfI9S/bKSLnB6PgrZQTqgrg/dnw62cwYBJc8DREJQK6ExRN07q3NgO+UsoIPAOcDuQBa5RSH4jIxsZ1ROSWJuvfCAxrsos6ERkavCIfwE8fwgc3gasOznkMRl7tT4nq05gTP+Ty6WiaprVDe57fzQK2iMhWEWkA3gQuOMD6U4E3glG4g1JbBv+ZDdF94NovYdQ1zYI96Jz4mqZ1b+1p0ukN7GoynQeMbmlFpVRfoD/weZPZVqVUNuAGHhGR/7Sy7SxgFkCfPn3aUax9RMTC9P9CwiBfNsUWRFlMKIXOia9pWpfjaHBQ4awg1ZbaYccIdoae/wPeFmmWJ7evP7/DpcDjSqljW9pQROaLyEgRGZmQkHBoR0/OaDXYAxgMiqgwk86Jr2lal+Lxerhj5R1cseQKal0dlwG2PQE/H2h6yUnxz2vJ/7FPc46I5PtftwIraN6+f8TZrWbdhq9pWpfyxLon+DL/S64fej0R5ogOO057Av4aYIBSqr9SyoIvqH+w70pKqUFADLC6ybwYpVSYfzweGAts3HfbI8lmNek2fE3Tuoz3t7zPixteZMrAKVwy8JIOPVabbfgi4lZK3QB8gu+2zAUiskEp9QCQLSKNwf//gDeleb7lE4DnlFJefBeXR5re3dMZ7Fazvi1T07QuIacoh/tX309WzyzuyLqjw4/XrvvwRWQJsGSfeffuM31fC9t9DaQfRvmCzh5uYndFCz3eaJqmHUF7avbwh+V/oGdkT/4x/h+YDeYOP2bIPGnbXjarGYfT0faKmqZpHaTOXcdNn99Evaee/zfp/xFtjT4ix+12Ad/XzaFuw9c0rXOICPd8dQ+byjbx9KlPc2x0izcudoiO6TixC7P52/C7YteOmqaFvud+fI5Ptn/CLSNuYVzKuCN67G4X8O3hJrwCNQ2etlfWNE0LomU7lvFMzjOcd8x5TB8y/Ygfv9sFfJvV98OIvlNH07QjaVPZJu5adRcZCRnMOWlOp/Sc1+0Cvt0f8HU7vqZpR0ppXSk3fX4TdoudJyY+QZgxrFPK0e1+tNUpkjVNOxw7qnbwxqY32Fy+meTIZFJsKaREpZBqSyXFlkKcNa5Z7b3B08AtK26hvL6cl896mfjw+E4re7cL+DpFsqZpB0tE+HbPtyzcuJCVeSsxGowMjh3MNwXfUPRrUbN1rUZr4CKQYkshz5HH90XfM2/8PAbHDe6kM/AJqYC/tXIryZHJhJvCW11Hp0jWNK296t31fLT1Ixb+tJAtFVuItcZybea1XDLwkkBN3elxkl+dT54jL/Ca58gjrzqP7/Z8R527jtlDZ3NmvzM7+WxCKOBX1Fdw+UeXM6bXGB4d/ygG1fLPE3vb8HUNX9O0lhXWFLLo50W8tfktKpwVDIwZyINjH+Ss/mft1/4eZgzjmB7HcEyPY/bbj4hQ567r0IRoByNkAn60NZpZGbP4x9p/8PT3T3PT8JtaXK+xhq9TJGuatq8NJRt4eePLLNu+DI94mJg6kcsHX87IpJGHdFeNUqrLBHsIoYAPcOWQK9letZ3nc5+nX49+nH/s/t3nWs1GLCaDbsPXNC2g3l3PE+ueYOFPC4k0RzL1hKlMHTS1Qzsj6QwhFfCVUtw1+i52OXYx5+s59I7qzYikEfutZ9cpkjVN89tUtok/r/wzv1b+yqWDLuXGYTcSZYnq7GJ1iJC7D99sNPPYhMdIiUrhD8v/wK6qXfutY7eadRu+pnVzHq+HBesXMPWjqVQ1VPHsac9y5+g7QzbYQwgGfIAeYT14+tSnEYTZn8+mqqGq2XLdCYqmdW+7q3dzzdJr+OfafzIxdSLvnv8uY3uP7exidbiQatJpqq+9L/+c8E9mLZvFrStu5d+n/TuQb9qmuznUtC7H4/XwfO7zrMxbSa+oXqTaUulj6+N7tfchITzhsNMRiAj/3fpfHv72YQRh7ti5nH/s+Z2S5qAztCvgK6XOBJ7A1+PVCyLyyD7LpwPz2NvX7dMi8oJ/2ZXA3f75c0Xk5SCUu11G9RzFvWPu5d6v7+Xhbx/m3jH3opTCHm5iT5XuBEXTuoqy+jLuWHkH3xR8Q1pcGj+V/sSnOz7FI3uTHDY+0NT0IpBqS6WvvS89I3u2eit2o0pnJQ+sfoClO5YyLHEYD5/8MCm2lI4+tS6lzYCvlDICzwCnA3nAGqXUBy10VbhIRG7YZ9tYYA4wEhBgrX/b8qCUvh0mD5jM9qrtLFi/gP72/kwbMg1bmG7D17SuIqcoh9u+uM0XkE96gMkDJgPg8rrYU72HXY5d7HTsZKdjJ7scu9hRtYOvdn+F0+MM7MNisAQuAn1sfehj70Nfe1/62vuSGJHItwXfcvequymrL+Pm4Tdz1ZCrMBqMnXXKnaY9NfwsYIuIbAVQSr0JXED7OiOfBCwTkTL/tsuAM4E3Dq24h+bm4Tezo2oHj2Y/Sh97H+zhiW224Vc6K1m9ezXZhdmMSR7DaX1PO0Kl1bTuQUR4deOr/HPtP0mOSmbh2QsZFDsosNxsMJNqTyXVnspJnNRsW694KaotYmfVTnY4drCzaqdvcOzkq/yvaPA2BNYNM4bh9Dg5pscxPHXqU52e3qAztSfg9waa3uqSB4xuYb2LlFLjgM3ALSKyq5Vtex9iWQ+ZQRl4+OSHmf7xdP608k+cHTuXOpeHZ5ZvYWpWH2IjLXi8HjaUbuCr/K9YtXsV60vW4xUvJoOJRT8v4vxjz+fOrND+BV/TjpTqhmru/fpelu1Yxm9Sf8ODJz+I3WJv9/YGZaBnZE96RvYkKzmr2TKveCmsKWSnYyc7qnwXgyhLFNOHTMdqsgb7VI4qqq2en5RSvwPOFJEZ/ukrgNFNm2+UUnFAtYg4lVK/B6aIyG+UUn8ErCIy17/ePUCdiDzawnFmAbMA+vTpM2LHjh3BOcMmimqLmPrRVLxeRYLjj6zZXk6YbQupvXdSrTZS7a5CoUiPT2ds77GM7T2WE2JP4Pnc55n/43ySI5N55JRHGJo4NOhl07TuYnP5Zm5dcSt5jjxuGXEL0wZP6zY/mnYkpdRaERl5wHXaEfBPBO4TkUn+6TsBROSvraxvBMpEpIdSaiowQUR+71/2HLBCRA7YpDNy5EjJzs4+YLkO1cbSjUz/eDomZcLh8nVmLm4b7urjSbEO46rhp/O7YQMJMzVv38spyuHPX/6ZgpoCZqbP5PeZvz8ivcxrWih5f8v7zP1mLjaLjXnj57X4YKR2aIIV8E34mmlOxXcXzhrgUhHZ0GSdZBEp8I9PBu4QkTH+H23XAsP9q64DRjS26bemIwM+wMq8lbyx6Q1GJo3k5N4n0zO8P++ty+eV1TvYWlJDXKSFqVl9uGxMH5J77M28Wd1QzV+/+ysf/PoBGfEZ/PWUv9LH3qfDyqlpocLpcfLXb//KO7+8Q1bPLP427m+dmhc+FAUl4Pt3dDbwOL7bMheIyENKqQeAbBH5QCn1V+B8wA2UAdeJyCb/tlcDf/Hv6iERebGt43V0wG+N1yt89WsJL3+9g882FWJQiklDkjgvoxcnHhtHdIQFgE+2f8IDqx/A5XXx56w/M/m4yforqaY1ISLsdOxkfcl61pesZ1X+KrZXbWdm+kxmD53dLe+Q6WhBC/hHWmcF/KZ2ldWy8NsdLFqzi4paF0rBkF52xh4bz9jj4umT6GLut/fy7Z5vObXPqcw5cQ4x1phOKavb66bCWUF0WDQmQ8g+S9flFNYU8vH2j9lcvplYayyx1ljiwuOIs8YFXqOt0R3a9CciQa1slNeX89bmt3B73cSHxzc7l7jwuFYzPxbWFPqCe6kvwG8o3YCjwddkajVaGRQ7iJkZMxmXMi5oZdWa0wE/CFweLz/squCrLaV89WsJ3+8sx+URLEYDQ/vYsSetZm3V60Rbo/nL6DvJTMgkPjy+zYdA2qPeXU9RbRHFdcWU1JUEhuLaYkrqSyip9U2XO8vxiheb2caonqMY02sMY5LH0M/eT3/zCLJKZyXLdixjybYlZO/JRhASwhOoaqhqdl94U9Fh0YGAeVz0caQnpJMen04fW5+D/nxqXDXkFOWwtnAt2YXZbCjZwIikEVw/9PrDupnA6XHy+k+v8/yPzwd+22pJuCmcOGuc72IQHhe4u624rhgAkzIxIGYAQ+KHkBaXRlp8GsdGH6srIkeADvgdoLbBzXfbyvj611K+2lLCxoIqlGU3ESmLUJZCAMwGC6m2FHpH9W7W1VnjeIQ5AhGh3FlOQU0BBdUFvtd9xsvq9/+pw6RMxIbHkhCeQHx4fGCIscbwS/kvrN69mt01uwFIikhiTPKYwAXgQG2mIkJVQxV7avYEjl9UW0RKVAqjk0d3uycSm6p11fJF3hcs2bqEVbtX4fa66Wfvx9n9z+as/mfRr0c/RIQaVw2l9aWU1pVSVl9GaV1pYLq0vpTiumJ+Kf+FOncdAHaLnfT4dNLi08hIyCAtPo1Ya2yzY1c6K1lbuDYw/FT2E17xYlRGhsQN4fjY4/l85+eU1ZcxttdYrh96PRkJGe0+NxHh4+0f88S6J8ivzueU3qdw64hb6dujL+X15ZTWlVJSVxI4j8bxsroySutLEREGxw32Bfj4NAbGDOz2tz52Fh3wj4DymgZWby1l5S+7+XLXtxTV78ZgLsMSXkFkRCUuVUyD1DXbJjosmnp3PfWe5ukdwk3hJEcmkxyZTM/InvSK6kVSRJIvuEf4Ant0WPQBvz2ICHmOPL7Z8w3f7P6Gb/d8S6WzEoDjoo9jTPIYBsQMoLi2mIKagkCA31Ozh1p3bbN9KRSC7++jd1RvxiSPYXTyaLJ6ZhEXHnfY752IUOmsJL8mn93Vu9ldvZv86nzyq33TVQ1VRIdFE2ONISYsxvdqjSE2LJZoazSx1thm84PxraqRy+ti9e7VfLT1I5bvWk6du47EiETO6ncWZx9zNifEnnBI357cXje/VvzK+pL15JbkkluSy5aKLXjFC/je5/T4dOwWO98Xf88v5b8AvidJMxIyGJE0ghFJI8hMyAw0r9S6annz5zd5cf2LVDgrOKX3KVw/9HrS4tMOWJa1hWv5R/Y/yC3JZWDMQG4beRsn9jrxoM9J6xp0wO8EBZV1fLO1lK+3lLJ6ayl55bVgrCW2h4PjejWQGFuD1VpJks1Or6hevsAe2YvkyGR6hPUIehOMV7xsKtvENwXf8G3Bt6wtXBtoeoi1xja7wPSM7BmYTo5KJiYshu1V2wPbZu/JDnzdHxAzgNE9RzM6eTQjk0YGHkgTERwuB5X1lVQ4KwJDpXPvdGFNYSDI17hqmpXXZrHRO6o3vSJ7YQ+z+7apr6DcWU5ZfVmgXXhfVqOVVHvq3sfqbX0Dj9knRCS0eDFweVy+PkirfX2Q7nLs8r1W+17r3HXYLXbO6HcGZ/c/mxFJI4J6UWlU66plY+lG1pes58eSH1lfsp5KZyVDE4cyImkEI5NGkhafhsVoOeB+alw1vLHpDV7a8BKVzkrGp4znuqHXMSRuSLP1tldu5/F1j/PZzs9IjEjkpmE3ce4x5+ofUo9yOuB3AbvKaln9qy/4f/1rCYVV/mAbaSEzpQcZKdEMTY0mI6UHcVFhbezt8Dk9Topqi0iMSNyvb862uL3uZheP74u+x+lxYlRGUmwpOBocVDormyW8akqhsIfZSYpIoldUL1KiUugV1YteUb18QT6qV5tPW7q8LiqdlZTVl1FRX0GZs4yyujLyq/MDj9bvcuzC5d2bK6npxcBmsbG7eje7HLsorC0M1KzB9wh+SlQKqbZUUmwpjEkew0m9TsJsPLqet6huqOb1Ta/z8oaXqWqoYkLqBK7PvJ6ekT159odnWfzzYixGC9ekX8MVg68g3BTe9k61Lk8H/C5GRNhWUsPqraXk7Kzgx7xKNhc5aPwIUmLCyUyJJjPVdyFI792DyLCu+2OX0+Pkh6If+HbPt+yo2oHdYic6LJoeYT2IDoveb9xmsR2RWqTH62FP7Z5AfpUdjh3sqtrFDscOHA0Oekf1DgT1VFtq4DeWYKTf7UocDQ5e++k1Xtn4Co4GB+GmcJweJ78b8DuuG3qdvg8+xOiAfxSocbpZn1/JD3kV/JBXyQ+7Ksgr97X5GxQMSLSRmdqDoakxDOsTzfFJNoyG0AlKWseraqjitY2vkV+dz1VpV3Fs9LGdXSStA+iAf5QqrXbyY74v+P+wq4KcXRWU1/qaKCIsRtJ792Bon2iGpUYzNDWGnj30XRGa1t21J+B33faCbiwuKoyJAxOZODAR8DUF7SitJccf/L/fVcGCVdtweXwX6+QeVoamRpMSE050hIXYSAsxEWaiIyzERFiIiTQTHW7BYgrJHi01TWsnHfCPAkop+sVH0i8+kt8O82WXrnd52FhQRc5O30Xgx7wKlv9cRL3L2+p+osJMREeYiYu0kGi3kmQPI9Hmf7VbBoVL/QAACqlJREFUSfKPx0RYMOhmI00LOTrgH6WsZiPD+8QwvE/zdA51DR7Kaxsor22gotblG69poNw/XlHroqTaya6yWrK3lwWaipoyGRSJNt9FID4qjPgoC3FRFuIiw4iLshAfFRaYjokwYzLqbw6adjTQAT/EhFuMhFvC6RXdvlvtnG4PxQ4nhVVOiqrqKayqp6hx2lFPfkUdP+ZV8P/bO9cYu6oqAH/rnnvuax6dKUwHpC2CmggxBkiDMRIlRgj6Q9QYlIQEowZjJMHwR6M/RBKTBh/RHwZBJcFERRTE/jFKIkZJFHlbHgGhAW1Tpp2287h37uM8lj/2vjN3hpnpbTvDzDmzvuRk77PP4+6Vlbv22Wuvc9axRockffN6jwjObVQLGayEDJWLDJQDBsshQ5Uig+UiA+Uig5WiP1akVgqohAXKxYBqKaASBlSKBVeGgS1KG8Y6YQZ/i1MuBuwcrbFzdPmPYnVJU2WmFTFZ73Cs3uZYw5WT9Q7HGm1ONCJm2zGNdszR2Tb1dsxsK6LejllmnFiVUlCgHLoBICwIYbFAsSCEQYHSCvXBSpGRaolt1ZCRWsi2asi2WshINfRtJYYrRZuNGFsaM/hGXxQKwkitxEitxDt39J/mUVVpRSmz7Yh6K6bejml2ElpxSrOT0I4TWlFCK0ppRgv1VuSORYkSJylRokRJSpSkxKnSiVMa7Xi+PtuKmWm6QWc1hipFxgYXXFPz25BzUY0NLbTXSkGu4vINwwy+sa6IiHczBewYWv/fi5KUmWbEdDNiqhkxPefrcx2mmhEnGh0mGx0mZ9u8PDHLPw4cY2qZdQzXdyh7V1O3rBS9Oypc7IpyMw432+hupWKBUtDT5vdde+DKYoFSsKT0M5coSUlSJUrUlWlKnCixL5NUiVNluFJkzK+5DNggZayCGXwjV4RBgbMGy6f0mYpOnHK80WGy3vabc1c1OgntaJkZSOxmINNzHSaidH4m0p2BRInSSVI68coRU+tFNQzYMVxmbLDsBoGhbllh+0CJWjlgoOTWWaqlIgOlgFqpuGlCdlM/sC1NMWqsDWbwjS1PqVjgnG2VNX+BTVXnn9C7A0AnSYl82YlT2vFCe2e+nhAnShgUKAZCsSAUCwWCQAgLPW1BgYLAdDPi6GybI7PtnrLFSxOzPPrKJLOt1d1cAGEgVMOAgXKRaikg8LOE7vJL7wuavUsypaDA+HCFc4YrjG9z5TnbyowPVxgfrrB9SYhvmqqLEjvR5OCJOQ6eaPptjkMnmhycatKJUyphgdHawprMSLXkypovffuwX6MZrrg1m6Fy0WY4q9CXwReRa4Af4VIc/kxV9y45fivwRVyKw6PA51X1dX8sAfb7U/+rqh9fo74bxqZGRJxxDqDKxj2xtiIXiXW80WGukzDXiWl0EpqdmEbb7bv2hEbb1bXHrAvSrfS0Ldx7YqbNC4dnmKy3WfrifhgIO4YqjA2VmW5GHDrRpJMsnvmcNVBi52iVi84d5qqLxxmqFJlpxUzNuXDi6bmIA5N1puYipuaiN13fS0FguDsAVBe2Shg4V1jq1oS6A3E87yZzW5KmlIKCc0N6t101DBbv+3o1DBjsiURbFJVWLi4bbaaqNKNkfj2r3o4X1QE+ddn65Z44qcEXkQD4MXAVcBB4XET2qeoLPac9DexR1TkR+TJwB/AZf6ypqqefiscwjDOiEgbs2l5j1/bVI7HOlChJOTrb5o2ZFhPTLd6Yac3Xj9bbnDdS5eqLx9k5WvWRYVXOG61SK/XvaOgazO47JjPNmOlmNL9uM9NyZW/b4ekmrShdNFtyA3HB7wuVsOBmUX7tpNlxv9GMElqdhGaUeJde/266WsnNmAZKgQss8FFsq0WtnTVQ2liDD1wOvKKqBwBE5D7gWmDe4KvqIz3n/xO4YS07aRjG5icMCrxtpP93QE4HEaFWKlIrFdf1d1YiTZV27NZzGu2YRsc9oXeN+dIn90Ynpt5OKBcLDPon/+6soLv1zg4GK+vrZe/n7ucB/+vZPwi8b5XzvwD8sWe/IiJP4Nw9e1X1oVPupWEYxiagUFiIOts+sHpCms3Img4nInIDsAf4UE/z+ap6SEQuBP4iIvtV9dVlrr0JuAlg9+7da9ktwzAMA+gnFusQsKtnf6dvW4SIfAT4JvBxVW1321X1kC8PAH8FLl3uR1T1blXdo6p7xsbG+hbAMAzD6I9+DP7jwLtE5AIRKQGfBfb1niAilwJ34Yz9kZ72UREp+/rZwAfo8f0bhmEYbx0ndemoaiwiNwN/woVl3qOqz4vI7cATqroP+C4wCPzWx8B2wy8vAu4SkRQ3uOxdEt1jGIZhvEVYxivDMIwc0E/Gq83xPrVhGIax7pjBNwzD2CKYwTcMw9gibEofvogcBV4/zcvPBibXsDsbTd7kgfzJlDd5IH8y5U0eeLNM56vqqjHtm9Lgnwki8sTJFi6yRN7kgfzJlDd5IH8y5U0eOD2ZzKVjGIaxRTCDbxiGsUXIo8G/e6M7sMbkTR7In0x5kwfyJ1Pe5IHTkCl3PnzDMAxjefL4hG8YhmEsQ24MvohcIyIvicgrIvL1je7PWiAir4nIfhF5xucUyBwico+IHBGR53ratovIwyLyH1+ObmQfT4UV5LlNRA55PT0jIh/byD6eCiKyS0QeEZEXROR5EbnFt2dZRyvJlEk9iUhFRP4lIs96eb7t2y8Qkce8zfuN/7jl6vfKg0vHp2F8mZ40jMD1Wf9Qm4i8hksdmdn4YRH5IFAHfqGq7/FtdwDHVXWvH5xHVfVrG9nPfllBntuAuqp+byP7djqIyLnAuar6lIgMAU8CnwA+R3Z1tJJM15FBPYn7IuWAqtZFJAQeBW4BbgUeVNX7ROQnwLOqeudq98rLE/58GkZV7QDdNIzGBqOqfwOOL2m+FrjX1+/F/RkzwQryZBZVPayqT/n6LPAiLstdlnW0kkyZRB11vxv6TYEPA7/z7X3pKC8Gf7k0jJlVcA8K/FlEnvQZwfLCuKoe9vU3gPGN7MwacbOI/Nu7fDLj/uhFRN6OS1D0GDnR0RKZIKN6EpFARJ4BjgAPA68CU6oa+1P6snl5Mfh55QpVvQz4KPAV707IFep8iln3K94JvAO4BDgMfH9ju3PqiMgg8ADwVVWd6T2WVR0tI1Nm9aSqiapegss4eDnw7tO5T14Mfl9pGLNGT3rII8DvcYrOAxPez9r1tx45yfmbGlWd8H/IFPgpGdOT9ws/APxSVR/0zZnW0XIyZV1PAKo6BTwCvB8YEZFuEqu+bF5eDP5J0zBmDREZ8AtOiMgAcDXw3OpXZYZ9wI2+fiPwhw3syxnTNYyeT5IhPfkFwZ8DL6rqD3oOZVZHK8mUVT2JyJiIjPh6FRec8iLO8H/an9aXjnIRpQPgQ6x+yEIaxu9scJfOCBG5EPdUDy4V5a+yKJOI/Bq4EvdlvwngW8BDwP3AbtxXUa9T1UwshK4gz5U4N4ECrwFf6vF/b2pE5Arg78B+IPXN38D5vLOqo5Vkup4M6klE3otblA1wD+n3q+rt3kbcB2wHngZuUNX2qvfKi8E3DMMwVicvLh3DMAzjJJjBNwzD2CKYwTcMw9gimME3DMPYIpjBNwzD2CKYwTcMw9gimME3DMPYIpjBNwzD2CL8HwAYUML+G4EUAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='relu'),\n",
    "  tf.keras.layers.Dense(100,activation='relu'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.Nadam(0.001),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    trainX,trainy,\n",
    "    epochs=30,\n",
    "    validation_data=(validX, validy)\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 488
    },
    "id": "qFuhEqc1iJJO",
    "outputId": "b1900480-7119-4bf8-f179-f4e3c9304731"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "157/157 [==============================] - 0s 2ms/step - loss: 0.5273 - sparse_categorical_accuracy: 0.8448\n",
      "0.8447999954223633 0.5272934436798096 (5000, 28, 28)\n",
      "[[411   1   1  12   0   0  80   0   2   0]\n",
      " [  1 470   0   9   0   0   1   0   0   0]\n",
      " [  4   1 325   9  57   0 125   0   0   0]\n",
      " [ 14   7   2 455   1   0  19   0   2   0]\n",
      " [  0   0  25  54 328   0 114   0   0   0]\n",
      " [  0   0   0   0   0 476   3   2   1   3]\n",
      " [ 63   1  25  14   9   0 365   0   5   0]\n",
      " [  0   0   0   0   0  35   1 456   0   8]\n",
      " [  0   1   0   3   0   0  16   1 505   0]\n",
      " [  0   0   0   0   0   9   5  30   0 433]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fb52d56ccf8>"
      ]
     },
     "execution_count": 19,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)\n",
    "predict = model.predict(test1X)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "k-PfmsjOkhGn"
   },
   "source": [
    "\n",
    "--------------------------------------------------------------------------------\n",
    "1.   hidden layer: tanh\n",
    "2.   optimizer: SGD\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "N-QTTOmikNCe",
    "outputId": "fc2ad260-cdb1-4db1-da04-a5c6aadf77c0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1500/1500 - 5s - loss: 0.9935 - sparse_categorical_accuracy: 0.6531 - val_loss: 0.9228 - val_sparse_categorical_accuracy: 0.6929\n",
      "Epoch 2/30\n",
      "1500/1500 - 5s - loss: 0.9381 - sparse_categorical_accuracy: 0.6531 - val_loss: 0.8625 - val_sparse_categorical_accuracy: 0.6787\n",
      "Epoch 3/30\n",
      "1500/1500 - 5s - loss: 0.9593 - sparse_categorical_accuracy: 0.6413 - val_loss: 0.9003 - val_sparse_categorical_accuracy: 0.6654\n",
      "Epoch 4/30\n",
      "1500/1500 - 5s - loss: 0.9534 - sparse_categorical_accuracy: 0.6403 - val_loss: 0.8872 - val_sparse_categorical_accuracy: 0.6361\n",
      "Epoch 5/30\n",
      "1500/1500 - 4s - loss: 0.9000 - sparse_categorical_accuracy: 0.6545 - val_loss: 0.8468 - val_sparse_categorical_accuracy: 0.6838\n",
      "Epoch 6/30\n",
      "1500/1500 - 4s - loss: 0.9019 - sparse_categorical_accuracy: 0.6509 - val_loss: 0.8778 - val_sparse_categorical_accuracy: 0.6718\n",
      "Epoch 7/30\n",
      "1500/1500 - 4s - loss: 0.9202 - sparse_categorical_accuracy: 0.6389 - val_loss: 0.9692 - val_sparse_categorical_accuracy: 0.5832\n",
      "Epoch 8/30\n",
      "1500/1500 - 4s - loss: 0.9155 - sparse_categorical_accuracy: 0.6511 - val_loss: 0.9608 - val_sparse_categorical_accuracy: 0.6590\n",
      "Epoch 9/30\n",
      "1500/1500 - 4s - loss: 0.9175 - sparse_categorical_accuracy: 0.6483 - val_loss: 0.8645 - val_sparse_categorical_accuracy: 0.6693\n",
      "Epoch 10/30\n",
      "1500/1500 - 4s - loss: 0.8835 - sparse_categorical_accuracy: 0.6580 - val_loss: 0.8656 - val_sparse_categorical_accuracy: 0.6789\n",
      "Epoch 11/30\n",
      "1500/1500 - 4s - loss: 0.8416 - sparse_categorical_accuracy: 0.6763 - val_loss: 0.8718 - val_sparse_categorical_accuracy: 0.6312\n",
      "Epoch 12/30\n",
      "1500/1500 - 4s - loss: 0.8662 - sparse_categorical_accuracy: 0.6577 - val_loss: 0.8421 - val_sparse_categorical_accuracy: 0.6853\n",
      "Epoch 13/30\n",
      "1500/1500 - 4s - loss: 0.9071 - sparse_categorical_accuracy: 0.6458 - val_loss: 0.9689 - val_sparse_categorical_accuracy: 0.6262\n",
      "Epoch 14/30\n",
      "1500/1500 - 5s - loss: 0.9296 - sparse_categorical_accuracy: 0.6439 - val_loss: 0.8331 - val_sparse_categorical_accuracy: 0.6751\n",
      "Epoch 15/30\n",
      "1500/1500 - 5s - loss: 0.9347 - sparse_categorical_accuracy: 0.6366 - val_loss: 0.9766 - val_sparse_categorical_accuracy: 0.5914\n",
      "Epoch 16/30\n",
      "1500/1500 - 4s - loss: 0.9196 - sparse_categorical_accuracy: 0.6428 - val_loss: 0.8817 - val_sparse_categorical_accuracy: 0.6810\n",
      "Epoch 17/30\n",
      "1500/1500 - 5s - loss: 0.8782 - sparse_categorical_accuracy: 0.6601 - val_loss: 0.8377 - val_sparse_categorical_accuracy: 0.6817\n",
      "Epoch 18/30\n",
      "1500/1500 - 4s - loss: 0.9279 - sparse_categorical_accuracy: 0.6422 - val_loss: 0.8420 - val_sparse_categorical_accuracy: 0.6868\n",
      "Epoch 19/30\n",
      "1500/1500 - 4s - loss: 0.8747 - sparse_categorical_accuracy: 0.6622 - val_loss: 0.8996 - val_sparse_categorical_accuracy: 0.6368\n",
      "Epoch 20/30\n",
      "1500/1500 - 4s - loss: 0.8935 - sparse_categorical_accuracy: 0.6505 - val_loss: 0.8507 - val_sparse_categorical_accuracy: 0.6729\n",
      "Epoch 21/30\n",
      "1500/1500 - 4s - loss: 0.8670 - sparse_categorical_accuracy: 0.6603 - val_loss: 0.8659 - val_sparse_categorical_accuracy: 0.6688\n",
      "Epoch 22/30\n",
      "1500/1500 - 4s - loss: 0.8501 - sparse_categorical_accuracy: 0.6770 - val_loss: 0.8322 - val_sparse_categorical_accuracy: 0.6685\n",
      "Epoch 23/30\n",
      "1500/1500 - 4s - loss: 0.8980 - sparse_categorical_accuracy: 0.6554 - val_loss: 0.9006 - val_sparse_categorical_accuracy: 0.6338\n",
      "Epoch 24/30\n",
      "1500/1500 - 4s - loss: 0.8448 - sparse_categorical_accuracy: 0.6686 - val_loss: 0.8700 - val_sparse_categorical_accuracy: 0.6625\n",
      "Epoch 25/30\n",
      "1500/1500 - 4s - loss: 0.8871 - sparse_categorical_accuracy: 0.6442 - val_loss: 0.8947 - val_sparse_categorical_accuracy: 0.6722\n",
      "Epoch 26/30\n",
      "1500/1500 - 4s - loss: 0.8803 - sparse_categorical_accuracy: 0.6509 - val_loss: 0.9419 - val_sparse_categorical_accuracy: 0.6158\n",
      "Epoch 27/30\n",
      "1500/1500 - 4s - loss: 0.8908 - sparse_categorical_accuracy: 0.6481 - val_loss: 0.8639 - val_sparse_categorical_accuracy: 0.6613\n",
      "Epoch 28/30\n",
      "1500/1500 - 4s - loss: 0.8661 - sparse_categorical_accuracy: 0.6589 - val_loss: 0.9061 - val_sparse_categorical_accuracy: 0.6108\n",
      "Epoch 29/30\n",
      "1500/1500 - 4s - loss: 0.8751 - sparse_categorical_accuracy: 0.6394 - val_loss: 0.9437 - val_sparse_categorical_accuracy: 0.5954\n",
      "Epoch 30/30\n",
      "1500/1500 - 4s - loss: 0.9192 - sparse_categorical_accuracy: 0.6271 - val_loss: 0.8284 - val_sparse_categorical_accuracy: 0.6683\n",
      "157/157 [==============================] - 0s 2ms/step - loss: 0.8558 - sparse_categorical_accuracy: 0.6576\n",
      "0.6575999855995178 0.8558036088943481 (5000, 28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='tanh'),\n",
    "  tf.keras.layers.Dense(100,activation='tanh'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.SGD(),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    trainX,trainy,\n",
    "    epochs=30,\n",
    "    validation_split=0.2,verbose=2\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 486
    },
    "id": "7TtXJyMok06L",
    "outputId": "7c7418fa-1eb1-4300-bd44-54766ae808c7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "157/157 [==============================] - 0s 2ms/step - loss: 0.8341 - sparse_categorical_accuracy: 0.6562\n",
      "0.6561999917030334 0.8340784907341003 (5000, 28, 28)\n",
      "[[383   1  22  45   0   0  39   0  17   0]\n",
      " [ 20 430   7  18   0   0   5   0   1   0]\n",
      " [  7   0 457   3   1   0  47   0   6   0]\n",
      " [125  11   9 281   0   0  68   0   6   0]\n",
      " [ 25   0 355   8   1   1 127   0   4   0]\n",
      " [  0   0   0   0   0 312   0  60  10 103]\n",
      " [108   1 268  24   2   0  63   0  16   0]\n",
      " [  0   0   0   0   0  12   0 434   5  49]\n",
      " [  1   2  18   3   1   2   6   2 489   2]\n",
      " [  0   0   0   0   0  11   0  35   0 431]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fb530639748>"
      ]
     },
     "execution_count": 21,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)\n",
    "predict = model.predict(test1X)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eqke4RCTlFTq"
   },
   "source": [
    "\n",
    "--------------------------------------------------------------------------------\n",
    "1.   Hidden layer: sigmoid\n",
    "2.   optimizer: SGD\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "VAcBGCbWlaJP",
    "outputId": "441eac76-6af7-4cce-edf0-a38ffb3b65aa"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 1.4741 - sparse_categorical_accuracy: 0.6249 - val_loss: 0.7350 - val_sparse_categorical_accuracy: 0.7618\n",
      "Epoch 2/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.6780 - sparse_categorical_accuracy: 0.7826 - val_loss: 0.6156 - val_sparse_categorical_accuracy: 0.7914\n",
      "Epoch 3/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5910 - sparse_categorical_accuracy: 0.8012 - val_loss: 0.5859 - val_sparse_categorical_accuracy: 0.7920\n",
      "Epoch 4/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5571 - sparse_categorical_accuracy: 0.8101 - val_loss: 0.5899 - val_sparse_categorical_accuracy: 0.7940\n",
      "Epoch 5/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5537 - sparse_categorical_accuracy: 0.8095 - val_loss: 0.5835 - val_sparse_categorical_accuracy: 0.7882\n",
      "Epoch 6/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5436 - sparse_categorical_accuracy: 0.8105 - val_loss: 0.5723 - val_sparse_categorical_accuracy: 0.7960\n",
      "Epoch 7/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5465 - sparse_categorical_accuracy: 0.8105 - val_loss: 0.5702 - val_sparse_categorical_accuracy: 0.7952\n",
      "Epoch 8/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5402 - sparse_categorical_accuracy: 0.8138 - val_loss: 0.5698 - val_sparse_categorical_accuracy: 0.7974\n",
      "Epoch 9/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5488 - sparse_categorical_accuracy: 0.8107 - val_loss: 0.5648 - val_sparse_categorical_accuracy: 0.7924\n",
      "Epoch 10/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5466 - sparse_categorical_accuracy: 0.8103 - val_loss: 0.5708 - val_sparse_categorical_accuracy: 0.8026\n",
      "Epoch 11/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5526 - sparse_categorical_accuracy: 0.8098 - val_loss: 0.5857 - val_sparse_categorical_accuracy: 0.7924\n",
      "Epoch 12/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5578 - sparse_categorical_accuracy: 0.8070 - val_loss: 0.5749 - val_sparse_categorical_accuracy: 0.7950\n",
      "Epoch 13/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5538 - sparse_categorical_accuracy: 0.8050 - val_loss: 0.6089 - val_sparse_categorical_accuracy: 0.7880\n",
      "Epoch 14/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5667 - sparse_categorical_accuracy: 0.8020 - val_loss: 0.6251 - val_sparse_categorical_accuracy: 0.7810\n",
      "Epoch 15/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5701 - sparse_categorical_accuracy: 0.8002 - val_loss: 0.6371 - val_sparse_categorical_accuracy: 0.7678\n",
      "Epoch 16/30\n",
      "1875/1875 [==============================] - 7s 3ms/step - loss: 0.5799 - sparse_categorical_accuracy: 0.7947 - val_loss: 0.5990 - val_sparse_categorical_accuracy: 0.7732\n",
      "Epoch 17/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5665 - sparse_categorical_accuracy: 0.7974 - val_loss: 0.5951 - val_sparse_categorical_accuracy: 0.7866\n",
      "Epoch 18/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5586 - sparse_categorical_accuracy: 0.8032 - val_loss: 0.6287 - val_sparse_categorical_accuracy: 0.7790\n",
      "Epoch 19/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5722 - sparse_categorical_accuracy: 0.7947 - val_loss: 0.5938 - val_sparse_categorical_accuracy: 0.7872\n",
      "Epoch 20/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5714 - sparse_categorical_accuracy: 0.7975 - val_loss: 0.6193 - val_sparse_categorical_accuracy: 0.7842\n",
      "Epoch 21/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5801 - sparse_categorical_accuracy: 0.7943 - val_loss: 0.6196 - val_sparse_categorical_accuracy: 0.7782\n",
      "Epoch 22/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5797 - sparse_categorical_accuracy: 0.7925 - val_loss: 0.6114 - val_sparse_categorical_accuracy: 0.7790\n",
      "Epoch 23/30\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.5781 - sparse_categorical_accuracy: 0.7901 - val_loss: 0.6204 - val_sparse_categorical_accuracy: 0.7678\n",
      "Epoch 24/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5920 - sparse_categorical_accuracy: 0.7852 - val_loss: 0.6238 - val_sparse_categorical_accuracy: 0.7730\n",
      "Epoch 25/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5860 - sparse_categorical_accuracy: 0.7915 - val_loss: 0.6145 - val_sparse_categorical_accuracy: 0.7756\n",
      "Epoch 26/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5778 - sparse_categorical_accuracy: 0.7941 - val_loss: 0.6163 - val_sparse_categorical_accuracy: 0.7738\n",
      "Epoch 27/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5877 - sparse_categorical_accuracy: 0.7847 - val_loss: 0.6264 - val_sparse_categorical_accuracy: 0.7690\n",
      "Epoch 28/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5808 - sparse_categorical_accuracy: 0.7910 - val_loss: 0.6128 - val_sparse_categorical_accuracy: 0.7820\n",
      "Epoch 29/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5920 - sparse_categorical_accuracy: 0.7873 - val_loss: 0.6304 - val_sparse_categorical_accuracy: 0.7706\n",
      "Epoch 30/30\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.5938 - sparse_categorical_accuracy: 0.7883 - val_loss: 0.6338 - val_sparse_categorical_accuracy: 0.7702\n",
      "157/157 [==============================] - 0s 2ms/step - loss: 0.6360 - sparse_categorical_accuracy: 0.7702\n",
      "0.7702000141143799 0.6359674334526062 (5000, 28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXUAAAD4CAYAAAATpHZ6AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd1wU18L/8c9hF1g6SLVjbyAqWLHGWKJpaoyJMUZz1XtjiYnp3XiT57m/mzypNz3XnmJJT0yisZtYwIoNu4ii9A5bz++PwRUUpLiAwHm/XvPaNjPn7KLfOXNm5oyQUqIoiqLUD061XQFFURTFcVSoK4qi1CMq1BVFUeoRFeqKoij1iAp1RVGUekRfWwUHBATI0NDQ2ipeURSlTtq9e3eqlDKwrM9rLdRDQ0OJjY2treIVRVHqJCHE2et9rrpfFEVR6hEV6oqiKPVIuaEuhFgohEgWQhws4/OOQojtQgijEOJJx1dRURRFqaiK9KkvBv4DLC3j83TgUeBuB9VJaUDMZjOJiYkUFhbWdlUU5aZiMBho1qwZzs7OlVqu3FCXUm4RQoRe5/NkIFkIMbpSJSsKkJiYiJeXF6GhoQghars6inJTkFKSlpZGYmIirVq1qtSyNdqnLoSYIYSIFULEpqSk1GTRyk2qsLAQf39/FeiKUowQAn9//yrtwdZoqEspP5VSRkkpowIDyzzNUmlgVKAryrWq+v+izp39cvRiNm/8fpT0PFNtV0VRFOWmU+dC/UxqHh9sPElSVkFtV0WpJzw9PWu7CoriMOUeKBVCfAUMBgKEEInAK4AzgJTyYyFECBALeAM2IcRjQGcpZXZ1VNjbTTsSnFVgro7VK4qi1GnlttSllPdLKRtLKZ2llM2klP+VUn4spfy46POLRe97Syl9i55XS6AD+BSFerYKdcXBpJQ89dRThIWFER4ezooVKwBISkpi4MCBdOvWjbCwMLZu3YrVamXKlCn2ed9+++1arr2iaGpt7Jeq8lEt9Xrr1Z8OcfiCY9sDnZt488odXSo077fffsu+ffvYv38/qamp9OzZk4EDB/Lll18yYsQIXnjhBaxWK/n5+ezbt4/z589z8KB2TV5mZqZD660oVVXn+tRVqCvVZdu2bdx///3odDqCg4MZNGgQMTEx9OzZk0WLFjF//nzi4uLw8vKidevWnDp1ijlz5vDbb7/h7e1d29VXFKAOttQ9XfXonATZBZbaroriYBVtUde0gQMHsmXLFn755RemTJnCvHnzmDx5Mvv37+f333/n448/ZuXKlSxcuLC2q6ooda+lLoTA26BXLXXF4QYMGMCKFSuwWq2kpKSwZcsWevXqxdmzZwkODmb69OlMmzaNPXv2kJqais1mY9y4cbz22mvs2bOntquvKEAdbKmD1gWjQl1xtDFjxrB9+3YiIiIQQvDvf/+bkJAQlixZwhtvvIGzszOenp4sXbqU8+fPM3XqVGw2GwD/+7//W8u1VxSNkFLWSsFRUVGyqjfJuOs/2/B1d2HJw70cXCulph05coROnTrVdjUU5aZU2v8PIcRuKWVUWcvUue4X0M5VVy11RVGUa9XJUPdxc1bnqSuKopSiToa6aqkriqKUrk6G+uUDpbV1PEBRFOVmVWdD3WKT5JustV0VRVGUm0qdDXVQV5UqiqJcTYW6oihKPaJCXVHqgX379rFmzZoaKWvatGkcPny40stt2rSJ22+/vRpqpBSnQl1RHMxiqflxiWoq1K1WK59//jmdO3eu9rKqk9Vaf4/HqVBXbh6/PguLRjt2+vXZcovNy8tj9OjRREREEBYWxooVKwgNDeXpp58mPDycXr16ceLECQB++uknevfuTffu3bn11lu5dOkSAPPnz+fBBx8kOjqaBx98kEOHDtGrVy+6detG165dOX78OADLly+3v//3v//9uuHy22+/0aNHDyIiIhg6dCgAu3btom/fvnTv3p1+/foRHx+PyWTi5ZdfZsWKFXTr1o0VK1aQl5fHww8/TK9evejevTs//PADAPn5+dx777107tyZMWPG0Lt3by5f2f3VV18RHh5OWFgYzzzzjL0enp6ePPHEE0RERLB9+3YGDx5sX6aidayIspazWq08+eSThIWF0bVrV95//30AYmJi6NevHxEREfTq1YucnBwWL17M7Nmz7eu8/fbb2bRpU6nfY8GCBfTs2ZOwsDBmzJhhP5vuxIkT3HrrrURERNCjRw9OnjzJ5MmT+f777+3rfeCBB+y/6U1HSlkrU2RkpKyqzHyTbPnMz/KzLServA7l5nD48OErL9Y8I+XCUY6d1jxTbh1Wr14tp02bZn+dmZkpW7ZsKV977TUppZRLliyRo0ePllJKmZ6eLm02m5RSys8++0zOmzdPSinlK6+8Inv06CHz8/OllFLOnj1bLl++XEoppdFolPn5+fLw4cPy9ttvlyaTSUop5SOPPCKXLFlSap2Sk5Nls2bN5KlTp6SUUqalpUkppczKypJms1lKKeW6devk2LFjpZRSLlq0SM6aNcu+/HPPPSeXLVsmpZQyIyNDtmvXTubm5so33nhDzpgxQ0opZVxcnNTpdDImJkaeP39eNm/eXCYnJ0uz2SyHDBkiv/vuOymllIBcsWKFfd2DBg2SMTExla7jxo0b7b9jacpa7sMPP5Tjxo2zf5aWliaNRqNs1aqV3LVrV4llr/4dRo8eLTdu3Fjq97hcXymlnDRpkvzxxx+llFL26tVLfvvtt1JKKQsKCmReXp7ctGmTvOuuu6SU2r+P0NBQe32qU4n/H0WAWHmdbK2TA3p5ueoRQt39qN657V+1Umx4eDhPPPEEzzzzDLfffjsDBgwA4P7777c/Pv744wAkJiYyYcIEkpKSMJlMtGrVyr6eO++8Ezc3NwD69u3L66+/TmJiImPHjqVdu3asX7+e3bt307NnTwAKCgoICgoqtU47duxg4MCB9vU3atQIgKysLB566CGOHz+OEAKzufT/A2vXruXHH3/kzTffBKCwsJCEhAS2bdvG3LlzAewtX9BavYMHDyYwMBDQWqJbtmzh7rvvRqfTMW7cOIfX8WplLffHH3/wj3/8A71eby8nLi6Oxo0b23/Lioxnf/X32LhxI//+97/Jz88nPT2dLl26MHjwYM6fP8+YMWMAMBgMAAwaNIiZM2eSkpLCN998w7hx4+z1udnUye4XJyeBt0FdVao4Rvv27dmzZw/h4eG8+OKLLFiwANCGeb7s8vM5c+Ywe/Zs4uLi+OSTTygsLLTP4+HhYX8+ceJEfvzxR9zc3Bg1ahQbNmxASslDDz3Evn372LdvH/Hx8cyfP79SdX3ppZcYMmQIBw8e5KeffipRfnFSSr755ht7WQkJCVUeOM1gMKDT6RxeR0ctV5xer7ePnAmUWEfx71FYWMjMmTNZvXo1cXFxTJ8+vdzyJk+ezPLly1m0aBEPP/xwpetWU+pkqIMafldxnAsXLuDu7s6kSZN46qmn7GOjX75H6YoVK+jbty+gtSabNm0KwJIlS8pc56lTp2jdujWPPvood911FwcOHGDo0KGsXr2a5ORkANLT0zl79mypy/fp04ctW7Zw+vRp+7xXl7948WL7/F5eXuTk5Nhfjxgxgvfff9/eT7x3714AoqOjWblyJQCHDx8mLi4OgF69erF582ZSU1OxWq189dVXDBo06Lq/W2XrWJ6ylhs2bBiffPKJ/QB0eno6HTp0ICkpiZiYGABycnKwWCyEhoayb98+bDYb586dY9euXaWWdTnAAwICyM3NZfXq1YD2OzZr1szef240GsnPzwdgypQpvPPOOwA39YFiFepKgxcXF2c/ePnqq6/y4osvApCRkUHXrl1599137TeWnj9/PuPHjycyMpKAgIAy17ly5UrCwsLo1q0bBw8eZPLkyXTu3JnXXnuN4cOH07VrV4YNG0ZSUlKpywcGBvLpp58yduxYIiIimDBhAgBPP/00zz33HN27dy9xls2QIUM4fPiw/UDpSy+9hNlspmvXrnTp0oWXXnoJwN6F0LlzZ1588UW6dOmCj48PjRs35l//+hdDhgwhIiKCyMhI7rrrruv+bpWtY3nKWm7atGm0aNGCrl27EhERwZdffomLiwsrVqxgzpw5REREMGzYMAoLC4mOjqZVq1Z07tyZRx99lB49epRalq+vL9OnTycsLIwRI0bYu3EAli1bxnvvvUfXrl3p168fFy9eBCA4OJhOnToxderUCn+n2lAnx1MHmPT5TvJNFr6dGe3AWik17WYdTz00NJTY2NjrBnddZLVaMZvNGAwGTp48ya233kp8fDwuLi61XbWbXn5+PuHh4ezZswcfH58aKbMq46nfnD39FeDtpicpq6C2q6EodUp+fj5DhgzBbNYGxPvwww9VoFfAH3/8wd/+9jcef/zxGgv0qqqzoa51v6ibTyvV48yZMzVWVu/evTEajSXeW7ZsGeHh4Q4vy8vLixvZQ3aERYsW8e6775Z4Lzo6mg8++KCWalS+W2+9tczjHzebOhvq3kU3ypBSljhLQVHqmp07d9Z2FWrU1KlTb/p+6bqsTh8oNVltFJpt5c+sKIrSQJQb6kKIhUKIZCHEwTI+F0KI94QQJ4QQB4QQpR9udjA1VICiKMq1KtJSXwyMvM7ntwHtiqYZwEc3Xq3yqVBXFEW5VrmhLqXcAqRfZ5a7gKVFwxLsAHyFEI0dVcGyqFBXFEW5liP61JsC54q9Tix6r1qpUFdqi6enZ5mfnTlzhrCwsBqsjaKUVKMHSoUQM4QQsUKI2JSUlBtalwp1RVGUaznilMbzQPNir5sVvXcNKeWnwKegXVF6I4WqUK9//t+u/8fR9KMOXWfHRh15ptcz153n2WefpXnz5syaNQvQhgLQ6/Vs3LiRjIwMzGYzr732WrmXzV+tsLCQRx55hNjYWPR6PW+99RZDhgzh0KFDTJ06FZPJhM1m45tvvqFJkybce++9JCYmYrVaeemll+yX3StKZTgi1H8EZgshvgZ6A1lSytIHtHAgL4MW6mr4XeVGTZgwgccee8we6itXruT333/n0Ucfxdvbm9TUVPr06cOdd95ZqWsiPvjgA4QQxMXFcfToUYYPH86xY8f4+OOPmTt3Lg888AAmkwmr1cqaNWto0qQJv/zyC6ANbqUoVVFuqAshvgIGAwFCiETgFcAZQEr5MbAGGAWcAPKBGrmqQOck8DLoVUu9HimvRV1dunfvTnJyMhcuXCAlJQU/Pz9CQkJ4/PHH2bJlC05OTpw/f55Lly4REhJS4fVu27aNOXPmANCxY0datmzJsWPHSh1rvawx3RWlssoNdSnl/eV8LoFZDqtRJfgUXVWqKDdq/PjxrF69mosXLzJhwgS++OILUlJS2L17N87OzoSGhlZpfO/STJw4kd69e/PLL78watQoPvnkE2655Rb27NnDmjVrePHFFxk6dCgvv/yyQ8pTGpY6O0wAqOF3FceZMGEC06dPJzU1lc2bN7Ny5UqCgoJwdnZm48aNVRr3Y8CAAXzxxRfccsstHDt2jISEBDp06FBirPWEhAQOHDhAx44dadSoEZMmTcLX15fPP/+8Gr6l0hDU6VBXdz9SHKVLly7k5OTQtGlTGjduzAMPPMAdd9xBeHg4UVFRdOzYsdLrnDlzJo888gjh4eHo9XoWL16Mq6srK1euZNmyZTg7OxMSEsLzzz9PTEwMTz31FE5OTjg7O/PRRzVyDZ9SD9XZ8dQB/rFsNydTclk37/p3aFFuXjfreOqKcjOoynjqdXZAL1DdL4qiKFer090vPu4q1JXaERcXx4MPPljiPVdX1wY3jK5y86nboe7mjNFio9BsxeBc8budK8qNCg8PZ9++fbVdDUW5Rp3ufvF2UxcgKYqiFFenQ10NFaAoilKSCnVFUZR6RIW6oihKPaJCXVEq6XrjqdcXmzZt4q+//qqRskaNGkVmZmall1u8eDGzZ8+uhhrVbfUi1NWBUqU+s1gsNV5mTYS6lBKbzcaaNWvw9fWt1rKq0+XvcbOo06c0ehu06mcV1Pw/esXxLv7P/2A84tjx1F07dSTk+eevO48jx1NPSkpiwoQJZGdnY7FY+OijjxgwYACenp5Mnz6dtWvXEhISwtdff01gYCCfffYZn376KSaTibZt27Js2TLc3d2ZMmUKBoOBvXv3Eh0dzV133cXcuXMBEEKwZcsWvLy8eOONN1i5ciVGo5ExY8bw6quvllm3pUuX8uabbyKEoGvXrixbtoyffvqJ1157DZPJhL+/P1988QUFBQV8/PHH6HQ6li9fzvvvv0/Hjh35xz/+QUJCAgDvvPMO0dHRpKSkMHHiRC5cuEDfvn1Zt24du3fvJiAggLfeeouFCxcCMG3aNB577DHOnDnDiBEj6N27N7t372bNmjUMGjSI2NhYAgICKlzH4ODgcv8WZS2Xm5vLnDlziI2NRQjBK6+8wrhx4/jtt994/vnnsVqtBAQEsH79eubPn4+npydPPvkkAGFhYfz8888A13yPf/3rX8TExFBQUMA999xj/1vExMQwd+5c8vLycHV1Zf369YwePZr33nuPbt26AdC/f38++OADIiIiyv1e5ZJS1soUGRkpHaHLy7/JV3885JB1KTXv8OHD9udJr78uz0x60KFT0uuvl1uHPXv2yIEDB9pfd+rUSSYkJMisrCwppZQpKSmyTZs20mazSSml9PDwKHNdb775pnzttdeklFJaLBaZnZ0tpZQSkMuXL5dSSvnqq6/KWbNmSSmlTE1NtS/7wgsvyPfee09KKeVDDz0kR48eLS0Wi5RSyttvv11u27ZNSillTk6ONJvN8vfff5fTp0+XNptNWq1WOXr0aLl58+ZS63Xw4EHZrl07mZKSIqWUMi0tTUopZXp6uv17ffbZZ3LevHlSSilfeeUV+cYbb9iXv//+++XWrVullFKePXtWduzYUUop5axZs+T//M//SCml/PXXXyUgU1JSZGxsrAwLC5O5ubkyJydHdu7cWe7Zs0eePn1aCiHk9u3b7etu2bKlTElJqXQdFy1aZP8dS1PWck8//bScO3duifmSk5Nls2bN5KlTp0qUffXv0KVLF3n69OlSv8flZSwWixw0aJDcv3+/NBqNslWrVnLXrl1SSimzsrKk2WyWixcvttchPj5elpWHxf9/XAbEyutka51uqYMaKqA+Ka9FXV0cOZ56z549efjhhzGbzdx99932lpiTk5P9TkaTJk1i7NixABw8eJAXX3yRzMxMcnNzGTFihH1d48ePR6fTLqqLjo5m3rx5PPDAA4wdO5ZmzZqxdu1a1q5dS/fu3QHIzc3l+PHjDBw48Jp6bdiwgfHjxxMQEABAo0aNAEhMTGTChAkkJSVhMplo1apVqd/rjz/+4PDhw/bX2dnZ5Obmsm3bNr777jsARo4ciZ+fH6CNJT9mzBg8PDwAGDt2LFu3buXOO++kZcuW9OnTx+F1vFpZy/3xxx98/fXX9vn8/Pz46aefGDhwoH2ey2Vfz9XfY+XKlXz66adYLBaSkpI4fPgwQggaN25Mz549AfD29ga0v+0///lP3njjDRYuXMiUKVMq9J0qok73qQPqRhmKQ1weT33FihXXjKe+b98+goODKzSe+sCBA9myZQtNmzZlypQpLF26tNT5Lt9BacqUKfznP/8hLi6OV155pUQZlwMRtC6izz//nIKCAqKjozl69ChSSp577jn27dvHvn37OHHiBH/7298q9b3nzJnD7NmziYuL45NPPinzO9psNnbs2GEv6/z581U+YFz8ezmyjo5arji9Xl+iv7ysv8/p06d58803Wb9+PQcOHGD06NHXLc/d3Z1hw4bxww8/sHLlSh544IFK160sdT7U1Y0yFEeYMGECX3/9NatXr2b8+PFkZWVVaTz1s2fPEhwczPTp05k2bRp79uwBtFBcvXo1AF9++SX9+/cHICcnh8aNG2M2m/niiy/KXO/JkycJDw/nmWeeoWfPnhw9epQRI0awcOFCcnNzATh//jzJycmlLn/LLbewatUq0tLSAEhPTwe02+Y1bdoUgCVLltjn9/LyIicnx/56+PDhvP/++/bXl4dIiI6OZuXKlQCsXbuWjIwMQBtL/vvvvyc/P5+8vDy+++67cu/mVNk6lqes5YYNG8YHH3xgf52RkUGfPn3YsmULp0+fLlF2aGio/W+4Z88e++dXy87OxsPDAx8fHy5dusSvv/4KQIcOHUhKSiImJgbQ/t6XD3xPmzaNRx99lJ49e9r3cByhXoS6aqkrN6q08dRjY2MJDw9n6dKlFR5PfdOmTURERNC9e3dWrFhhP7jp4eHBrl27CAsLY8OGDfa7Gv3zn/+kd+/eREdHX7eMd955h7CwMLp27YqzszO33XYbw4cPZ+LEifTt25fw8HDuueeeEkF89fd74YUXGDRoEBEREcybNw/QDgqPHz+eyMhIe7cHwB133MF3331Ht27d2Lp1K++99x6xsbF07dqVzp078/HHHwPwyiuvsHbtWsLCwli1ahUhISF4eXnRo0cPpkyZQq9evejduzfTpk2zdxNd729QmTqWp6zlXnzxRTIyMggLCyMiIoKNGzcSGBjIp59+ytixY4mIiLB3lY0bN4709HS6dOnCf/7zH9q3b19qWZf/5h07dmTixIlER0cD4OLiwooVK5gzZw4REREMGzbM3oKPjIzE29ubqVMdewfQOj2eOsBTq/az9XgqO54f6oBaKTWtoYyn7unpaW9R1ydGoxGdToder2f79u088sgjaqCzCrpw4QKDBw/m6NGjODmV3r6uynjq6kCpoihVlpCQwL333ovNZsPFxYXPPvustqtUJyxdupQXXniBt956q8xAr6p6EeoFZismiw0XfZ3vTVLqiMqOp15TrfS0tDSGDr12r3X9+vX4+/s7vLx27dqxd+9eh6+3Ml5//XVWrVpV4r3x48fzwgsv1FKNyjd58mQmT55cLeuu+6HufmWogEAv11qujVIVUkr72SB1xc06nrq/v/9NWa/q9MILL9zUAV5VVe0ar/NNWzX+S91mMBhIS0ur8j9gRamPpJSkpaVhMBgqvWydb6l7q1Cv05o1a0ZiYiIpKSm1XRVFuakYDAaaNWtW6eXqfKirQb3qNmdn5wpfIagoSvlU94uiKEo9Um9CPbtQhbqiKEqFQl0IMVIIES+EOCGEeLaUz1sKIdYLIQ4IITYJISrfEVRF9pZ6vgp1RVGUckNdCKEDPgBuAzoD9wshOl8125vAUillV2AB8L+OrmhZnHVOuLvoVPeLoigKFWup9wJOSClPSSlNwNfA1XcL6AxsKHq+sZTPq5W3QV1VqiiKAhUL9abAuWKvE4veK24/MLbo+RjASwhxzeVrQogZQohYIUSsI09hU0MFKIqiaBx1oPRJYJAQYi8wCDgPWK+eSUr5qZQySkoZFRgY6KCiVagriqJcVpHz1M8DzYu9blb0np2U8gJFLXUhhCcwTkpZ+duDV5G3mzOJGfk1VZyiKMpNqyIt9RignRCilRDCBbgP+LH4DEKIACHE5XU9Byx0bDWvT90oQ1EURVNuqEspLcBs4HfgCLBSSnlICLFACHFn0WyDgXghxDEgGHi9mupbKtX9oiiKoqnQMAFSyjXAmqvee7nY89XAasdWreJ83JzJM1kxW2046+r89VSKoihVVi8S0MdN2zapLhhFURq6+hHq7mr8F0VRFKgvoa4G9VIURQFUqCuKotQrKtQVRVHqkXoR6t724XcttVwTRVGU2lU/Qt2g7n6kKIoC9STUDc46XPVOqvtFUZQGr16EOhRdVapulKEoSgNXv0JdtdQVRWngVKgriqLUIyrUFUVR6hEV6oqiKPVIvQl1bzWmuqIoSv0JdR83Z3KMFqw2WdtVURRFqTX1KtRBXYCkKErDVu9CXfWrK4rSkKlQVxRFqUfqT6irG2UoiqLUo1C3j9SoQl1RlIar3oW6aqkritKQqVBXFEWpR+pNqLvqnXDRqeF3FUVp2PS1XQFHEUI0vKtKjbmQcxFykiD3kvaYc1Gbci+BdxNoNxza3ALujWq7toqi1IB6E+oAPm76ut1SlxKM2ZCXWjSlFE3Fn6dcCW9T7rXr0LuBVwh4BsGJP+DAChBO0DRKC/h2t0JIBDjVm500RVGKqWehXo2DekkJWYlwbqcWqJYCMF+e8sFcWPRYAJai5xajtpy2gmvXd+UFGHO0wLaaSi/f4AMegdoUEg5th2nh7dUYvIK1R89gbT4htGVsVriwF46vg+NrYeNr2uQRBG1vhXbDoM0QcPO7Uqf89KKW/8WSrf6cJMhNBr0BfFuAb8uix6LJKwScdI78xRVFqYJ6F+opuUbHrMxqhotxWoif2wkJOyHnQsl5hA6c3cHZ7arJHQy+WgBeDlgo+Vx748pTVy/wCLgS3MWfuweA3qXy38FJB82itGnIc5CbAifXawEfvwb2f6l9h8COUJilhbetlI2iwadooxGkbayOr9XmLVGWM/g0uxLyPs21Lh/3RuB21aOzeym/haIojlChUBdCjATeBXTA51LKf131eQtgCeBbNM+zUso1Dq5ruXzcnDmRUkqXREXkp0NiLJzbAed2wfndWoCBFlAt+0Hz3tCit9ZKdfEAnbPjKl8TPAMh4j5tslq073hiHSTth8YRWovfM6RoD6Bo8gzWNlRXMxdoey6ZZyEzoeRUWugXp3O9Kux9tY2gm1+x50WvLz83+GobF7U3oCjXVW6oCyF0wAfAMCARiBFC/CilPFxstheBlVLKj4QQnYE1QGg11Pe6qnSf0vN7YNvbcOQnQGot18ZdocdkLcSb9wafptVS31ql02sbqBa9y53VkppK9q+/kbNuHTofb7xHjcJz8GCcAtpBQLsyFjJBQQYUpGsbzFIfiz5PPQ4FmVCYqXVdlcVJD43aaGUGdoCA9kVTO21PR1GUCrXUewEnpJSnAIQQXwN3AcVDXQLeRc99gKv6KWrG5eF3bTaJk9N1du+lhNObtTA/tQlcfSB6rtbP3LSH1gpv4KxZWeSsW0fWL7+Qv3MX2Gy4tmuL6fRpctb9gXB3x+uWW/AePQrP6GiEy1XdQ3qXor7+4MoVbC64EvDFHwsyIC9Z2wCkHoNjv4HNcgUWXbUAACAASURBVGU576ZauAd00B6d3cCUrx1MNuVpe12m3KL38rTn5nztuIN3E21vzKcZ+Da/8twjqOwDyhYTZJ/X9layzl15zDynHdjWOYPeVZt0rtrvoTcUPS82eYZAo1bQqLW2B1iVbjZFKaYiod4UOFfsdSJwdfNuPrBWCDEH8ABuLW1FQogZwAyAFi1aVLau5fJ2c0ZKyCm02MeCKcFmg6M/a2F+YY/WtTBsAUROBYP3tfNXMyklplOnyNuxA2tWFk4uLojLk/PlR+cr77k4I5ycsGZnY83IxJqZgTUz88qUkYk1KxNLZia2zCycvL1xbdcO17Zttcd2bXFt2xadd+nf1ZafT87GjWT/sobcrVvBbMa5RQv8/z4Dn1GjcG3XDmm1kh8TS/aaNeT8/jvZP/+Mk48P3sOH4T1qFO69eiF0FesisRmNWNPT0fn64uRW1MVz+biEd+PrL2w1Q/ppSI2HlPiisI+HfV+UclaQABdPcHHXNtguHuDsoXXpCAEZZ+HMNu3Mo+J0LtrGwqeZFvSWoi6nrETtAPLVB789goo2Ck21DY7FqG1ErBnaRsBSqB0ItxiLpsKSxzCEk1ZWo9bXTh5B2l5NbrK2cctLvfI89/KZUUXvO7tp9fZuqtXFuwl4NyvaeDUFryZ1a+MhpToGUwlCyuvfVEIIcQ8wUko5rej1g0BvKeXsYvPMK1rX/wkh+gL/BcKklLay1hsVFSVjY2Md8R3sVsWe46nVB9jy1BBa+Ltf+cBigriVsO0dSDsOfq2Q/R6FLuORQo+0WJBGI9bsHKxZmdiys7XgzMzCmp2FNSsLW1bRe9nZ6Hx9MXTuhKFLFwydO+McFFThOpovXSJv+3byt28nb/sOLMnJN/alnZ3R+fqg9/VF5+OLzs8Xna8vOh8fLBkZGE+cwHT8BLb8fPsi+uDgEmHv5OFOztp15GzciCwoQB8cjPdtt+E9ejSGsC6IMv5DSZOJ3L/+InvNGnL/WI8tPx9dYADeI2/Dc0B/bAWFWNJSsaalF3tMw5qWhiUtDVtuUfjq9Ri6dMY9Mgr3qEjcundH7+dXtd9DSsi+oIWli6cW4FcfsC5LYZbW0ra3vs9dCfGsRK1l7VPUkvctas1fDnzvpuBsqHxd89Mg/dRV02ntsSD9+ssLJ3D31wLfI0A7kO0RqO2JZJ/Xfoes82DMunZZjyAt4O1nMLUseZDb1bNy38XRci7Boe8gbhVcPKB1uwV1gqDOENRRe/QLbZDHWIQQu6WUUWV+XoFQ7wvMl1KOKHr9HICU8n+LzXMILfjPFb0+BfSRUpaZWNUR6msPXWTGst38NLs/XQJcSXruWfJ3/QUFOUirFSn1SPRImwSLpfwVFnHy9ETn7Y2Trw86L28sKSmYTp+2n5aoCwzArXMXDF0624NeHxKCEAJrTg75u3aRt30Hedu3Yzp5UlvGzw+Pvn1w79sXj759cQ4JQZpMSJMJm8mENJm112aT/X1pMiFtNnTePkXh7YeTh3uZoXuZtNmwJCVRePw4phMnMB4/jvH4CYwnTyKN2tlCOl9fvEaOwGf0aNwiIxGVPI/dVlBA7ubNWit/82akqdipmUJoGxr/Ruj9A9D7+6Pz99ce/fwwJyaSv3s3hXFxSLPWcnVt1xa3HpG4R0XiHhmJc5MmlapPvVCQcSXg81KKAjzwSni7+1cs1Iw5WsBnn9dCPvsCZBdtqDLPaQe3rVedNebuX/KUVXd/bSPp6lX06Fm00Sz23NXrxkK2MEs7thW3Ck5vAWmDkK4Q2h8yzkDyYe3xMr1BO6YS1Lko8DuBqzdIq9atJm3acymLXhe9Z7NyZS+r6P+OEGU8d9L2GgM6aHt6NwFHhLoeOAYMBc4DMcBEKeWhYvP8CqyQUi4WQnQC1gNN5XVWXh2hvvNUGhM+3cHySd1o8dbL5P31F94t8nHyDYKm3RABoVq3hl6vTS7OCL0e9HqEi4sWlj7e6Ly90fn44OTjg87LS5vnKtbcPIzxRyk8dJjCQ4coPHwY48mTWhcPoGvUCH1wMMb4eLDZEG5uuEdF4dG3Lx59++DaoUOlg9PRpNWKOTERS3o6bmFhCGfHnM1jzc2l8NBhdD7e9uAu7Te8ms1opDAujvzY3eTv3k3B3r321ry+SWOtJR+pBb1L69a1/vuVRtpsWpeSv3+5G9sKrU9K+4ZY18gffVAQ+qAgdJ4OPO5js2kbjcyE0s9mKi30y+LqDf5trhzAvnwwu1FrbU/nauZCOP67FuTH1mrl+LWC8PEQfo92QLw4U57W3ZZ8RAv55COQclTbYFUroW3cgjppdQrspO0xBLSv2DE4KbXjRcZsbSPr6qWdXVaVmtxoqBetZBTwDtrpigullK8LIRYAsVLKH4vOePkM8ETbBD4tpVx7vXVWR6gfvZjNHf+3ga8TvsX9QCyNe6bjO/MV6POIQ8spi62gAGN8PAWHtaC3JCVhiIjAo29f3Lp1w+nqg4nKdUmrFWN8PPm795C/ezf5u2OxpqQC2p6FW48e9pA3dO7ssI1SVdiMRrJ++IH0RYsxnT6NvkljPHr3waNPb9x798Y5pOL/ga1ZWeRt307utm3kbfsTy8WL18zj5O5uD3h9cDD6oECci167tG6Da5vWFdqQVoiURQea87RAMuVqQ1SYcq96nQf5qUXHN45rewOXCSeti+dy2DdqpZ15duQnLeg8giBsnBbmTXtUvg+9IFMLe3O+trcgnLQz2YRT0WuhvbZ/5lTywsCyntusWjfc5Q1JSrzWhWu/SLAo7AM7al1xpjzt+xRma91expyi59klD+z3fxxunV/Zv4RWoiNCvTpUR6hfSM5k/T1TiEw5RpOxbfH13AvzjtR+/6DiEFJKzAkJV1ryu3djOnsWAGEw4BYRYQ95t8hInFxLaRk6mCUjg8wVK0hf/gXW1FRcO3fCe/gICo8cIX/nTqyZmQC4hIbi3qc3Hn364N6rF/pGV8bikVYrhXFx5G77k7ytWymIiwObDScvL23Prn80buHhWLOysaQkY0nWJvOlK88tyckluryEwYChUycMYWG4hXXBEBaGS6tWNbt3Y8yFtBPalHqsaCoKfKtRa9V3ulNrkbcaeNP1j1uzs8le8yv6kGA8+/e/spG0WiDj9JWQTzkCyUe1ixNdvLSTLly9ix69ij0v9hjcRZuqoMGEus1o5OzMWRT++ScnH5jO7fwLejwIo//PYWUoNx9LSkqJlrzxqNbdpfP3x2/i/fjdf3+JAHUUU2Ii6YsWk/ntt8iCAjwGDsD/4Ydx793b3u0ibTaMx46Rt2MH+Tt2kh8Tgy0vDwDXDh1w79ULS2oKeX9tx5aVBUJgCA/Hs380Hv3749a1a4Vb21JKbFlZmC9dwnjsOIUH4yg4qHULyoICQGvdG7poAW8I64JrmzbasabrHMOxmUxgNmPo0kU71uKIs1BsVq27xCOo8geXa4DpzBnSly4j8/vvkUUnGOiDgvAZMwbfcWNxqYYz9yqjQYS6zWQicfZs8rZs5b0e93LPMBh69h34x58QEuaQMpS6wZqbS35sLJlffU3u5s0IV1d87rqLRlMewrV16xtef0FcHGkLF5Lz+1rQ6fC5/XYaTZ2CoX37cpeVFguFBw+St2MneTt3ULBnLzpvbzz698ejfzQe/fpV/ayfssq0WjGdOqUF/MGD2nT0qP0AeWU4t2yB75gx+Nx9d6W6k6pC2mzaabppadjy8nDy9MTJyxudtxfCYHDMxqV4eVKSv3MX6UuWkLtpE0Kvx3v0aPwmTcJyMYnMVau103xtNtx79cL3nnF4DR+Ok6HmN0r1PtRtJhOJc+aQt3kLIf9cwK1HfPjN+QkCA4Nh2h8OqKlSVxlPnSJ98RKyfvgBaTTiOWgQjaZOxb13rwqHgrTZMCckUHDwEJlff01+bCxOXl743TcBv0mTcA6u5MVVxddtsYBO5/CAKrdcsxnjyZOYzpxFOOtLXhdRdD1E8WsmEILcrVvJ+vY78nftAiHwiI7Gd+wYPIcOrXQ3lyU1FePx45iTLmJNT8OSmlbiVFdLehrW9AywWktfgV6PzssLJy8v7dHbC52n9ujctCmurdvg2rYNLi1aXHtR3FVsJhPZv6whfelSjEeOoPPzw+/++7Q9vMDAEvOaL10i67vvyfz2W8wJCTh5eeF9+2h8x92DoUvnGvs71utQt5lMnJ/zKLmbNxOy4FX87r2Xx//1Pm8Xvgh3fwTdJjqotkpdZklPJ+PLr8j48kus6em4du6E/5QpeN92W4mDq9bcPIzH4ik8ehTj0XgK449iPHbc3n2hb9yYRg9Nxvee8Y49+6QOMZ07R9Z335H5/fdYLiTh5OODz+jR+Iwde02w2fLzMZ44gfHYMQqPHcN47DjGY8ewppc8/14YDNpZUgH+6Bv5ow8oOuW16LmThwe2vDys2TnYcnNKPFpzsrHl5GLLyS465pBy5UCnXo9Ly5a4tm6NS9s2uLZpq4V9aCi2/Hwyvv6ajK++wpqSimu7tvhNnozPHXeU2/qWNhv5MbFkfrOanN/XIo1GXDt2xHvEcJzc3cFJh9DrSj7qnLQNuE4HOh2uoaG4titjiI1y1NtQt5lMnH90LrmbNhHy6qv4TbgXgC3/eyeRpt14PH+i9IGolAbLVlhI1k8/kb54CaaTJ9EHB+M1Yrh2Dv/ReMznrlw47eTtjaFDB1w7dsTQsQOuHbRHh51RUsdJm438HTvI/OZbctatQ5pMuLZvj0f//pjPJVB47BjmhHP2gBVubtrFbu3bYWjfHtf27XFu1gx9o0Y4eThuA2krKMB0+jTGkycxnjiJ6VTRY0LClZa/EKDXg9mMx4ABNHroITyi+1WppW3Nzib7l1/IXP0NhYcOlb9AEf/p0wh64olKlwf1NNSlyUTi3MfI3biRkPnz8btvgvZBXiqWNzqwxjCKO59d5sDaKvWJtNnI27qVtMWLyd+5C5cWLa6Ed/sOGDp2QN+4cY13i9RV2lkia8j85lsKDx3Sfs+i4Hbt0B5D+/Y4N29eq9cV2EwmTGfOYDp5EuPJU9gK8vEdOxbXNm0cV0Z+vnZ1utUKVivSagOrBWmzgUV7lBaLdiDfz6/KXXflhXqda3ZIk4nExx7XAv2Vl68EOsC+L9BjYbUYxp21V0XlJiecnPAcNAjPQYOQVmuFx6pRSqfz9sbvvvvwu+++m/b3dHJxwdC+fYUOaFe5DPeb44rTm++SvHJk/fgjuRs2EPzyS/jdf/+VD2w2iF3EGY8IDhir98i8Un/cjAFUl6nfs/bVuZa6z7hxuLRsiXvPniU/OL0ZMk5zqO0Csg+Zyx9+V1EUpR6qcy11IcS1gQ6wexG4NeJS0+HYJOSaKj5gl6IoSn1R50K9VDmX4Ogv0G0inp7akACVvgOSoihKPVA/Qn3vMm2wnMipeLtp5x1nFahQVxSl4an7oW6zwZ4lEDoAAtriUxTq2SrUFUVpgOp+qJ/coI33HDUV4EqoF6pQVxSl4an7ob57EbgHQMc7APB2007oUd0viqI0RHU71LMvQPyv0H2S/Ua6PqpPXVGUBqzOhfr2C9uZ8PMEMgszYc8y7b6DkQ/ZP/d01aNzEirUFUVpkOpcqBv0Bg6nHWb3xRjYsxRaD9Huf1hECIG3Qa9CXVGUBqnOhXqYfxgGnYHY+G+1eyBGPXzNPD5uzmQVqIuPFEVpeOpcqDvrnIkIjCD2Ygx4BkOH266ZRwt11VJXFKXhqXOhDhDp2454WUhWxATQXXsHeW8V6oqiNFB1MtSj0i4ghWBv866lfu7j5qwuPlIUpUGqe6FuNdP18G+4IIjNOVvqLKr7RVGUhqruhfqx33DNvUhXr1bEXIopdZbLoV5bd3VSFEWpLXUv1EO6wqBniQq9laPpR8kx5Vwzi4+bM1abJM9Uxt3IFUVR6qm6F+p+LWHIc0Q17oVN2tibvPeaWdRVpYqiNFQVCnUhxEghRLwQ4oQQ4tlSPn9bCLGvaDomhMh0fFVL6hrYFb2TnthL19682h7qakx1RVEamHJvZyeE0AEfAMOARCBGCPGjlPLw5XmklI8Xm38O0L0a6lqCm96NrgFdib1YdqirkRoVRWloKtJS7wWckFKeklKagK+Bu64z//3AV46oXHkigyM5nHaYPHNeiffVjTIURWmoKhLqTYFzxV4nFr13DSFES6AVsKGMz2cIIWKFELEpKSmVres1okKisEor+5L3lXhf9akritJQOfpA6X3AaillqaedSCk/lVJGSSmjAgMDb7iwboHd0Itr+9W91d2PFEVpoCoS6ueB5sVeNyt6rzT3UUNdLwDuzu50Duh8Tb+6l6seIVRLXVGUhqcioR4DtBNCtBJCuKAF949XzySE6Aj4AdsdW8Xr6xnck4OpB8k359vfc3ISeBvUVaWKojQ85Ya6lNICzAZ+B44AK6WUh4QQC4QQdxab9T7ga1nDl3FGhURhkRb2p+wv8b4aKkBRlIao3FMaAaSUa4A1V7338lWv5zuuWhXXPag7OqEj9lIsfZv0tb+vQl1RlIao7l1RehUPZw86Nep0Tb+6CnVFURqiOh/qoHXBxKXGUWgptL+nQl1RlIaofoR6cBRmm5kDKQfs73mrMdUVRWmA6kWodw/ujpNwKnG+uhp+V1GUhqhehLq3izcd/DpcE+pmq6TArIbfVRSl4agXoQ5av/qBlAMYrUZADRWgKErDVH9CPTgKo9VIXEocoEJdUZSGqd6EemRwJAJh74LxdtNOwc8usNRmtRRFUWpUvQl1H1cf2vu1t4e6aqkritIQ1ZtQB61ffX/yfsxWswp1RVEapPoV6sFRFFoLOZR2SIW6oigNUr0K9cjgSABiLsbgZVChrihKw1OvQt3P4Edb37bEXopF5yTwMujVVaWKojQo9SrUQWut703ei9mm9aun5Zlqu0qKoig1pt6Fes+QnhRYCjiSdoSoln5sOHKJzHwV7IqiNAz1LtSL96v/fVAb8kxWlvx1tpZrpSiKUjPqXagHuAXQyqcVsZdi6dTYm6Edg1j012nyjOoiJOWKo+lHWbB9AZPWTOJS3qXaro6iOEyF7nxU10QFR7Hm9BosNgszh7Rl3Ed/8dWuBKYNaF3bVVNqUYGlgN/P/M6q+FUcSD2Aq84VgeDJzU+ycORCnJ2ca7uKSj1itplJzk8mKTeJpLwkLuVfsj+/teWtjG03tlrKrbehvurYKuLT44ls2YU+rRvx2dZTPNi3Ja56XW1XT6lhpzJPserYKn44+QM5phxa+bTimZ7PcEebO9h+YTtPbXmKt2Lf4plez9R2VZU6Kj49nl9O/cKFvAsk5SVxMfciKQUpSEoO/e3n6keIRwhma/WdlVc/Qz0kCoDYS7F0CejCzMFtmbxwF9/tOc99vVrUcu2UmmC2mlmfsJ6Vx1YSczEGvZOeYS2GMb7DeKKCoxBCADCy1Uj2p+xn+ZHlRARFMDJ0ZI3UL9eUyxObnyAyOJIZXWfUSJmKRkqJ2WbGRedyw+vKNmXzn73/YUX8CvRCTxPPJoR4hNCvaT8aezSmsUdjQjxC7JOb3s0B3+D66mWoB7kH0dK7JTEXY3ioy0MMaBdAeFMfPt58kvFRzdE5idquolJNck25LD60mFXHVpFemE5Tz6bM7TGXMW3H4O/mX+oy8yLncTD1IC//+TLtfdvT2rd6u+nMNjPzNs1je9J2/rrwF4FugYxpN6Zay7yZWGwWUgtS8Tf446yrmS4vKSXHMo7x25nf+PX0r1zKv8TtrW9napepVfp726SNH078wDt73iHTmMmEDhOY1W0WPq4+1VD7yqmXoQ5aF8zaM2ux2qzonHTMHNyGR77Yw5q4JO6IaFLb1VMczGwzsyp+FR/v/5gMYwaDmw9mQocJ9GvSDydx/fMBnHXOvDnoTe79+V4e2/QYX43+Cg9nj2qpp5SSBdsXsD1pOy/3fZl1Z9axYMcCmnk1o2dIz2opsyZJKckwZmhdEHkXS0yX30spSMEmbXg6ezKg6QBuaXkL/Zv0x9PF0+H1OZN1xh7kp7JOoRM6+jTuQ5/Gffjl1C98f+J7BjcfzMNhD9M9qHuF1nk47TCv73ydAykH6B7UnU96f0LHRh0dXveqErV1u7eoqCgZGxtb/oxV9NPJn3h+2/OsvH0lnfw7YbNJhr29GRe9jjWP9rfvfit1m5SS9QnreWfPO5zNPkuvkF7Mi5pHF/8ulV7XrqRdTF83neEth/Pvgf+uln8jH+37iA/3f8g/Iv7BrG6zyDZlM2nNJNIL0/ly1Je08K5696CUklXHVnE66zRWacUmbaVP2LDZbBj0BnqG9KRfk35l7sVUxIXcC2w6t4lN5zaxN3kvhdbCEp+7OLnQ2LMxIe4hBHsE09ijMYFugRxJP8LGcxtJL0zH2cmZ3o17M7TFUAY3H0yAW0CV65OUm2QP8iPpRxAIegT34LbQ2xgWOoxGhkYAZBRm8NXRr/jq6FdkGjPpFtiNh8MeZlDzQaU2BLKMWby/931Wxq+kkaER86LmcUfrO2o8S4QQu6WUUWV+Xl9D/WLeRYatHsbTPZ/mwc4PArB6dyJPrtrPoik9GdIxqNrKVmrGvuR9/F/s/7EvZR9tfNowL2oeA5oOuKH/ZP+N+y/v7HmHZ3s9ywOdHnBgbeH7E9/z0p8vcWebO3kt+jV7Pc9ln2Pimon4uvqyfNTyKu3Cm6wmXvrzJdacXoOnsyc6Jx1OOOEkyp6yjFlkGjMB6NSoE9FNo+nXpB/dArtdt1tESsnh9MNsTNjIpnObiM+IB6CVTyv6NelHc6/m9j7kxh6N8XP1K/NvYrVZ2Z+ynw0JG1ifsJ7E3EQEgm5B3bil+S3c0uIW+4ZOSkmBpYAcUw655lxyTDn2KdecS0ZhBtvOb2Nfyj4AwvzDuK3VbQwPHU6IR0iZ3yffnM93J75j6aGlXMi7QGuf1kzpMoXRrUfjonPBJm18e/xb3t3zLjmmHO7veD8zu83Ey8Wr0n8nR2iwoQ5wx3d3YLaZWTxycdERZxuD39hEiI+B1f/oq1rrddTZ7LO8u+dd1p1dR4BbALO6zeLutnejd7rx3kQpJXM3zmVr4lYWjVxEt6BuDqgx/HXhL2b9MYuokCg+HPrhNaG5+9Jupq2dRmRwJB/d+lGlTq/MLMxk7sa57Enew9wec/lb2N8q9G/bJm0cST/CX+f/4s8Lf7I/eT8WacFd707vxr2JbhJNv6ZaSButRnYl7dJa5ImbSM5Pxkk40S2wG7e0uIVBzQYR6hNa2Z+lBCklxzOPsz5hPRsTNnIk/QgAQW5BGG1Gck25WOX17znc1rcto1qNYmToSJp7N69U+RabhbVn1rLw4ELiM+IJcgving73sDVxK3GpcfQI6sHzvZ+nQ6MOVf6OjtCgQ/1AygFmrJtBgFsAC0csJMg9iCV/neGVHw+xYkYfereu+i6nUvPSC9P5ZP8nrIxfibPOmalhU3mo80O4O7s7tJxsUzb3/XwfRouRFXesuKGuANBOd3vot4do4tmEJSOXlNnC++HED7z454vc0/4eXu7zcoWC+Vz2OWaun8mF3Au83v91Rraq+tk7uaZcdl7caQ/587nnAWjm2Yy0wjQKLAW46d3o37Q/g5sPZkDTAfgZ/KpcXnku5F5gQ8IGjqQfwV3vjpeLF54unni5eOHl7HXldbHnjji7RErJ9gvbWXhoITuTdhLgFsATUU8wutXom6Ih2KBDHbRd9L+v+ztB7kEsHLEQT30j+v+/DYQ19WHJw72qvXzFMbYmbuXpLU9TYClgbLuxzOw284bD9nri0+OZtGYSXQO78smwT6q8F3Ax7yIPrNG6cb4Y9cV1uwEA3t3zLp/Hfc5TUU8xucvk6867L3kfj254FBs23hvyHj2Ce1SpjqWRUnI2+yx/XviTnUk7CXQLZEiLIfQM6YmrztVh5dzsEnMSaWRo5PCGw40oL9SRUpY7ASOBeOAE8GwZ89wLHAYOAV+Wt87IyEhZU3Zf3C17Lu8p7/juDpmSnyL/s+G4bPnMzzIuMbPG6qBU3aHUQ7Ln8p7ynh/vkSczTtZYuT+c+EGGLQ6Tb8e+XaXls43ZcswPY2TvL3rLo2lHK7SM1WaVj214TIYvDpcbEzaWOd9vp3+TPZb2kKO+GSXPZJ2pUv2UugmIldfJ1nLHfhFC6IAPgNuAzsD9QojOV83TDngOiJZSdgEeq/Tmpxr1CO7Bh0M/5GLeRab9Po3R3T3xctXz0aaTtV01pRwX8y4ye/1sfFx9+HDoh9V+Dnlxd7a5k/Htx/Pfg/9lQ8KGSi1rtmrnop/OPM1bg9+qcD+sk3Di9f6v08m/E09veZr49PgSn0spWXhwIU9ufpLO/p1ZPmo5Lb1bVqpuSv1WkQG9egEnpJSnpJQm4GvgrqvmmQ58IKXMAJBSJju2mjcuKiSKD4Z+wPnc8zy++RHu6e3LmoNJnEzJre2qKWXIMeXwyB+PUGAp4IOhHxDoHljjdXim1zN08e/CC9teYNv5bVzMu4jVdv2DdVJK5m+fz46kHczvN59+TfpVqkx3Z3fev+V9vFy8mL1hNqkFqYB2IO+fO/7J27vfZmToSD4f8Xm19mkrFWOzSX4/dJGXfzjIO38c46tdCWw4eomD57NIyTFis9VsF3e5fepCiHuAkVLKaUWvHwR6SylnF5vne+AYEA3ogPlSyt9KWdcMYAZAixYtIs+erfkhcXcm7WT2+tk08WjG0b0PcFfXdvz7nogar4dyfWabmVl/zCLmYgwf3vohfZv0rbW6XMi9wH0/30eGMQMAZydnmno2palXU5p5NtMmr6LJsxlLDi/h4/0fM7PbTB6JeKTK5R5OO8yU36bQ1rct79/yPi/8+QJ/nv+Tv4X9jUd7PFruRVVK9boc5u+uP87Rizm4u+jIN127wdc7CQK9XAn2NhDsrT0O6RBU5dOqb/hAaQVD/WfAjNav3gzYAoRLKTPLyzebzgAAE7lJREFUWm9NHSgtzY6kHcxePxuDDOZi/BS2PHk7TXyrf0wGpWIut3S/Pf4tC/otuCkuoc8szORI+hEScxNJzCmaip5nm7Kvmf/utnezoN+CGz5bYn3Ceh7f+DgGvQGT1cSLfbSzY5TaY7NJfj14kffWHyf+Ug6tAzyYfUtb7oxogk1CSq6RS9mFJGcXcilbe34p20hyTqH9+UP9Qpk3rH2Vyi8v1CtySP88UPyEz2ZF7xWXCOyUUpqB00KIY0A7IKaS9a0RfRr34b0h7zF7wxxcm3/OB5tb8PpdvSu0bIGlAL2TXg3TWo3+e/C/fHv8W2Z0nXFTBDqAr8G3zL2FLGMW53PP24NeJ3RM7DTRIae/DW0xlCejnuS/B//L24PfJrpp9A2vs7ZIKTmSlMPawxfZciyF9sFezBzclhb+N8+ZJddjtUl+iUvi/fXHOZ6cS+tAD96Z0I07IpqUGE+qqa8bTctpJJbXmL4RFWmp69G6VoaihXkMMFFKeajYPCOB+6WUDwkhAoC9QDcpZVpZ6/3/7Z17cFPXncc/R5Yt25Is+SH5bfwEjIkTkkCB0oSQAVoSmqSPJE1Jkz7S7mzT7e7OTruPP9rNTmcznc1OsztJ03aTNk2Tps/QwDazFPIiMBAIMQbjp4zxA+vhl16WLEs6+4eEwQQbDAZZ4nxm7tx7j66uz08/36/O/d3f+SmRI/UzvDfwHn+9+1tEg4Xs/PxLVObFYrbuCTd93j56Pb2xtbeXfm8/vd5ehgJDZKZlsix/GY2WxthS0EihvjChtqQKf+7+M9/d+122VG3hyU88uSDyghcCUsqk/CwiUcnhnhF2nXCw64SdvpEAQsDyEhPtDi+RqOSzN5fy+B11C1bcI1HJzubT/PebXXQ5fdRaDXxrQy13N5YkpDjgvOSpCyG2AD8iFi9/QUr5AyHEE8RSa14Xsf+2p4ilPkaAH0gpX53tnAtB1AF+c3wX/3boO5jSrZSb8+jz9n3kdtqabaXcWE6FsYIyYxmjwVGah5ppHW5lMhqri1yYXTgl8I2WRurz669Jmc2FyBHHEQ4MHuDjpR+nsaDxksXoA8cHPLbrMW4ouIGfbfrZvJRGVVw+p4b9/PZwH68dGSAclVTkZceW/Oyz23nZWIy6aT4OTkbY2znErhY7e9qcjPhDZKRpWFdXwKZlhdxZX4jFqMPhCfLjt2288n4vkajkMytKeXxDLYvyr04xtbkQCEXodHo52u/m5/tO0u3ys7jQwLc21LHlhuKEVnq97icfXQoPvPQ8JwK/45ayMmrzKik3lk+JeKmxdEZxDkVCtI20cWzoGEddR2l2NU/NwksTadTn1bOlegt3Vd81VUQolWlyNvFM0zMcGDww1VaZU8nWmq3cXX03JYaZq2P2uHvY9sY2cnW5l13/RHHlBCcj7Drh4DeHetnXNYxGwPolVvL1GfSOjNM3Ms6gJ8i5spGZrqE8NybwQgj2dQ0RmIxgzNRy51IrmxqKuG2xBYPuwtFehyfIc+/YeOVgL+FrLO6RqKRn2E+73Uub3UuH3Uu7w0vPsH/KxqVFRv7mzjo+2VCEZgGU7Vaifgl0OLx89tn95GSl89JXV1FtufwSoMOBYY4NHaPZ1cz+0/tpGW5Bq9Gyvmw999Xdx9qStfNSo2Qh0exq5tmmZ9l3eh95mXl8ZflXuKv6Lvb27+V12+scdsT8vLJoJVurt7Jx0cZpZVZHgiNs+/M2/JN+frXlV5Qb51azQ3HltNu9vHqol9c+HGBsfJLyvCweuLWcz91STpEpc9qxE+EIA6OBKZHvnVoCjIfC3FZnYXNDEauq8sjQXnqGjtMT5Ll3unn54CnCUcl9K0p5/I5aKgumi/t4KIzTM4HTG3v46PRM4PJN4PJOMBmJohECIUAg0AjO7sfXGgHjExHaHV66nD4mwlEg1l6Zr2dxoZElRUaWFhlZXGSkKl+/IMT8DErUL5HjA24eeeF9AF78yiqWl87PSLFztJPtXdvZ2b2TkeAIliwLW2u2cm/tvVSZqublbySKlqEWnml6hr0De8nV5fLl5V/mgSUPfGRK9YBvgJ22nezo3sEpzyky0zLZULGBT9d8mhXWFXz9L1+nbaSN5zc/z42WhZleGghF0Gk1C+rivlL8E2F2HD3Nq4f6aOobIyNNw6aGQh5cWcHamvyE2Xq+uK+rLSAwGcHlncDpCeKfIW2wwKBDl65BSohKiZSxZxHRM/ucmUEPGVoNtVYDS4uMLCnKYUmhkbpCA5npC//nLpWoz4Ful49t/3MQbzDM84+uZFXV/IVMJiOTvNv/Ltu7trN3YC8RGWGFdQX31t7L5srNV+1HGa4GrcOtPNv0LG/3v41JZ+LRhkd5aOlDF62PIaXkqOsoO2w7eKPnDbwhL7o0HaFIiKfWP8XGRRuvqF/ByQh/ONLPL/b1EIpEqbUYqLUaqLHG1rVWAzmZs2cthcJRuod8tNu9tNu9dDhit+X9owFKzVl8YVU5968sx2rMnPU8CxXfRJi32538X4uDN1sd+EMR6qwGHlxVwX0rSsnTL5znGE5PkJ+82827HS5y9RlYjDqsRh1WYyZWoy62nxPbN2elp9QX7mwoUZ8jp8cCPPz8QfpHAzy37ZarUnd9KDDEDtsOXut6jZPuk2Rps6jPq8ekM8WWDNPZ7XPazDozhgwDWo0WrdDGamZfowkoE5EJjg8d55ctv+TNvjcxZhinxPxyfrEmFAnxTv87vHHyDVYXr+b+Jfdfdt/c45O8dKCHX+zvYcgX4sYyE2V52dicPrpdfkKR6NSxVqNuSuBrrQYKDDq6XT7aHT7a7R66XX7C8RmAWo2g2hK7Ha+xGDh8aoR9XcNoNYLNDUV8cXUFa6rz55yVEolKWk67OdA9zLAvRCgSJRyRhKNRQuHYOhyR8fYo4ahEqxEsK8mhsczMjWXmj4REZsPlnWB3q4NdLXb2dQ0TikTJ12ewcVkhn7+1nJsrzEmZWXO9okT9Mhj2TfDozw/ROujhqftv5J6bSq/K3zl35HrScxL3hBv3hBtPyEMgHLikcwgEaSKNNE1abH3OtiXbQo25hlpzLdWmamrNtZQaSknTzH6LGYqE6BztpGW4hZbhFk4Mn6BrtIuwDGNMN/Jww8Nsq9+WsB8JOMPpsQAvvHeSX7/fiz8U4fbFFv7q9hpWV+dNiVQ4EqVvNECX03d2cfmwOX34JsJT5yrLzYrFUOPx1CVFRqoK9Oi00z8rm8vHrw/28rsP+nEHJqku0PPQxyr43C1lmLMvPMqVUtLh8LHfNsR+2zAHuofxBmN/OzNdQ7pGgzZNkJ6mIT0ttq3VTN+PZWP4iMS/cKxGHY1lJhrLzDSWmbixzEzuOaPsniE/u07Y2dXi4IPeUaSE8rwsNi8rYvPyIm6uyFW/1ZukKFG/TLzBSb724mHe7xnhiXuW8/Dqa1s0KRgO4gl5poTePeHGHXLjDXkJR8NEZCS2RM9bywjhaJhwNIx93I5tzIbdb586ry5NR5WpaprYF2QVTBPxjtEOwtGY6Jh1ZhryG1iWv4yG/AZWFq8kJyPnmn4W59Ph8PKTd7r5U9MAEtjaWMw3bq+hvvjS+yWlxO4JMuwLUVmgnzEzYyaCkxH+t3mQlw+e4kjvGDqthrsbS/ji6gpWlJs5NTzOftsw+21DHOgeZsgXAmBRfjZra/JZU1PAmup8LMZLL2MbnIzQctpDc/8Yzf1umvvHsLn8U6+X52WxvMREt8tPu8MLQENJDpuWFbGpoZClRUY1Ik8BlKhfAcHJCI+/coTdrU7+YdNivnlHbVJeFL6Qj253N7YxG7YxG13uLrrHuhn0D047zphhnBLvhvwGGgoaKNGXLAibz0xi+em73expc5KVnsaDq8r56roqynITO2nlxGkPLx88xfYPB/CHIuRkavHER+KFOTrW1hSwpiaftTX5895Xb3CSYwPuKZE/PuCh2JTJ5oYiNi4rpDxvYU7oUVw+StSvkMlIlO/8vpnXPhzga+uq+Je76heEyM0H/kk/tjEbroCLxebFlBnLEm7bqD9E95CfbpePk0N+ul1+Tg75OTnsJxSOkqfP4NG1lTy8etG0cMNCwDcR5k9NAxw5NcZNFWbW1uRTXaBP+GeqSC2UqM8D0ajkiZ0n+MX+Hj5/Sxn//pkb0KapCnmXSzQq6R8N0OHw0un0YZsScB+j45NTx2k1goq8bKoteqotBpYUGtlyQzFZGQs/7UyhuFrMR0Gv6x6NRvC9rcswZaXz9J5Odrc6KM3Noigni2JTJkWmTIpyMqe2i01ZSniIhUz6RsbpdProiE/06HTG1sHJsxkpFqOO6gI9n1xeRHWBgWqLnqoCPeV52aSrL0+FYk4oUb9EhBD83cbF1FoN7LcNY3cH6B8d5/CpEcbOGV2ewZSVTnleFutqLdxZb2VFuXnBju6jUXnFOb5DvglaBz20DXpptcfWNtfZ2XoAxaZMaq0GvvixRdRZDdQVGqm1GjBlqYqXCsV8ocIv80AgFMHuCTLoDmB3Bxl0B7G7g3Q6vRzuGSUclZiz01m/2MKG+kJur7Ngyk68kB0fcPP0nk72tDrIzc6I32Wcvds4d7soJ5OsjDQmwhFsTj9tdk9MxO1eWge9DPkmps5bmKOLz9IzUGc1UltooM5qwHiRiT8KheLiqPDLNSArI42qgljI4Hw8wUn2dgyxp83B2+0utjedJk0juHVRLnfWW9mwtJAay7V9mNbcP8Z/7elkd6uTnEwtX1pTSSgSZXAsQP9ogMOnRi9492HOTscXDE9NzsnQalhSaOSOJRaWFudQX2RkaXHOgpqVqFBcb6iR+jUkEpU09Y3xZpuDPa1O2uyxXOJF+dncsiiX6gI9VfGYcmW+ft7j8k19Yzy9u4O32l2YstL52roqHvl45QWnzgdCkel3HvE7EWNmOvXFOSwrNlKZr1+wISWFIlVR2S8LmIGxAG+2OXmrzcmJ0x7snuC010tMmVTFHxpWFRioLtBTWaCnxJz5kZmOs3Gkd5Snd3fyTocLc3Y6j32imi+tWaTCIQpFEqJEPYnwT4TpGY7nZcfzs8/kbJ+ZzAIgBBQaMynLzYov2dPWJeYsMrQaDveM8PSeTvZ2DpGbnc5jt1XzpTWVc549qVAoFg4qpp5E6HVaGkpMNJRML/srpWR0fHJqQs5APPYdy74ZZUfz4FRNEIiJfr4+gyFfiHx9Bv/0qaVsW70IvRJzhSLlUVd5EiCEIE+fQZ4+j1srP1oOOByJYvcE40IfE/uB0QBLiow89LEKsjOUmxWK6wV1tacA2jRNPPSi6nwoFNc7KnVBoVAoUggl6gqFQpFCKFFXKBSKFEKJukKhUKQQStQVCoUihVCirlAoFCmEEnWFQqFIIZSoKxQKRQqRsNovQggXcOoy314ADM1jdxYCqWZTqtkDqWdTqtkDqWfThexZJKW0zPSGhIn6lSCEODxbQZtkJNVsSjV7IPVsSjV7IPVsuhx7VPhFoVAoUggl6gqFQpFCJKuo/zTRHbgKpJpNqWYPpJ5NqWYPpJ5Nc7YnKWPqCoVCobgwyTpSVygUCsUFUKKuUCgUKUTSiboQ4pNCiHYhRJcQ4h8T3Z/5QAjRI4Q4JoRoEkIk3Q+3CiFeEEI4hRDHz2nLE0L8RQjRGV/nJrKPc2UGm74vhBiI+6lJCLElkX2cC0KIciHEW0KIE0KIFiHEt+PtSemnWexJZh9lCiHeF0Icjdv0r/H2KiHEwbjm/UYIkTHreZIppi6ESAM6gI1AP3AI+IKU8kRCO3aFCCF6gFullEk5aUIIcRvgA34ppVweb/shMCKlfDL+5ZsrpfxuIvs5F2aw6fuAT0r5H4ns2+UghCgGiqWUR4QQRuAD4F7gUZLQT7PYcz/J6yMB6KWUPiFEOvAe8G3g74E/SilfFUI8BxyVUv54pvMk20h9FdAlpeyWUoaAV4F7Etyn6x4p5bvAyHnN9wAvxrdfJHbBJQ0z2JS0SCkHpZRH4tteoBUoJUn9NIs9SYuM4YvvpscXCWwAfh9vv6iPkk3US4G+c/b7SXJHxpHALiHEB0KIrye6M/NEoZRyML5tBwoT2Zl55HEhRHM8PJMUoYrzEUJUAiuAg6SAn86zB5LYR0KINCFEE+AE/gLYgDEpZTh+yEU1L9lEPVVZJ6W8GfgU8M34rX/KIGMxvuSJ883Mj4Ea4CZgEHgqsd2ZO0IIA/AH4G+llJ5zX0tGP13AnqT2kZQyIqW8CSgjFplYOtdzJJuoDwDl5+yXxduSGinlQHztBF4j5sxkxxGPe56JfzoT3J8rRkrpiF90UeBnJJmf4nHaPwAvSyn/GG9OWj9dyJ5k99EZpJRjwFvAGsAshNDGX7qo5iWbqB8C6uJPgzOAB4HXE9ynK0IIoY8/6EEIoQc2Acdnf1dS8DrwSHz7EeBPCezLvHBG/OLcRxL5Kf4Q7nmgVUr5n+e8lJR+msmeJPeRRQhhjm9nEUsIaSUm7p+LH3ZRHyVV9gtAPEXpR0Aa8IKU8gcJ7tIVIYSoJjY6B9ACrySbTUKIXwPriZUJdQDfA7YDvwUqiJVYvl9KmTQPHmewaT2x23oJ9ADfOCcevaARQqwD9gLHgGi8+Z+JxaGTzk+z2PMFktdHjcQehKYRG3D/Vkr5RFwjXgXygA+BbVLKiRnPk2yirlAoFIqZSbbwi0KhUChmQYm6QqFQpBBK1BUKhSKFUKKuUCgUKYQSdYVCoUghlKgrFApFCqFEXaFQKFKI/wcirHPyxTq5mgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='sigmoid'),\n",
    "  tf.keras.layers.Dense(100,activation='sigmoid'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.SGD(),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    trainX,trainy,\n",
    "    epochs=30,\n",
    "    validation_data=(validX, validy)\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 486
    },
    "id": "QMo6BDfHlsfM",
    "outputId": "c7333428-8e47-42b8-f15c-74dc5bab4b63"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "157/157 [==============================] - 0s 2ms/step - loss: 0.6360 - sparse_categorical_accuracy: 0.7702\n",
      "0.7702000141143799 0.6359674334526062 (5000, 28, 28)\n",
      "[[435   0   8  40   3   0  12   0   9   0]\n",
      " [  9 436   7  27   1   0   1   0   0   0]\n",
      " [ 29   0 364   7  83   0  38   0   0   0]\n",
      " [ 48   6   2 424   7   0  10   0   3   0]\n",
      " [ 24   0 166  40 241   0  46   0   4   0]\n",
      " [  1   0   0   1   0 434   0  20   1  28]\n",
      " [158   1  78  22  66   0 147   0  10   0]\n",
      " [  0   0   0   0   0  24   0 447   0  29]\n",
      " [  6   1  15   3   6   2   6   2 483   2]\n",
      " [  0   0   0   2   0   5   0  30   0 440]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fb5306426d8>"
      ]
     },
     "execution_count": 23,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)\n",
    "predict = model.predict(test1X)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wYw-CoUypHmz"
   },
   "source": [
    "\n",
    "\n",
    "1.   hidden layer: elu\n",
    "2.   optimizer: SGD\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "EX7RHOdKo3KK",
    "outputId": "1e93b795-7a1c-4e8f-e686-84e84840bc42"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1500/1500 - 5s - loss: 0.6646 - sparse_categorical_accuracy: 0.7772 - val_loss: 0.5097 - val_sparse_categorical_accuracy: 0.8183\n",
      "Epoch 2/30\n",
      "1500/1500 - 4s - loss: 0.4846 - sparse_categorical_accuracy: 0.8310 - val_loss: 0.4609 - val_sparse_categorical_accuracy: 0.8340\n",
      "Epoch 3/30\n",
      "1500/1500 - 4s - loss: 0.4482 - sparse_categorical_accuracy: 0.8422 - val_loss: 0.4424 - val_sparse_categorical_accuracy: 0.8406\n",
      "Epoch 4/30\n",
      "1500/1500 - 4s - loss: 0.4278 - sparse_categorical_accuracy: 0.8505 - val_loss: 0.4425 - val_sparse_categorical_accuracy: 0.8411\n",
      "Epoch 5/30\n",
      "1500/1500 - 4s - loss: 0.4109 - sparse_categorical_accuracy: 0.8555 - val_loss: 0.4345 - val_sparse_categorical_accuracy: 0.8464\n",
      "Epoch 6/30\n",
      "1500/1500 - 4s - loss: 0.3993 - sparse_categorical_accuracy: 0.8584 - val_loss: 0.4042 - val_sparse_categorical_accuracy: 0.8547\n",
      "Epoch 7/30\n",
      "1500/1500 - 4s - loss: 0.3881 - sparse_categorical_accuracy: 0.8617 - val_loss: 0.3974 - val_sparse_categorical_accuracy: 0.8586\n",
      "Epoch 8/30\n",
      "1500/1500 - 4s - loss: 0.3791 - sparse_categorical_accuracy: 0.8651 - val_loss: 0.3892 - val_sparse_categorical_accuracy: 0.8603\n",
      "Epoch 9/30\n",
      "1500/1500 - 4s - loss: 0.3695 - sparse_categorical_accuracy: 0.8693 - val_loss: 0.3883 - val_sparse_categorical_accuracy: 0.8593\n",
      "Epoch 10/30\n",
      "1500/1500 - 4s - loss: 0.3632 - sparse_categorical_accuracy: 0.8705 - val_loss: 0.3792 - val_sparse_categorical_accuracy: 0.8632\n",
      "Epoch 11/30\n",
      "1500/1500 - 4s - loss: 0.3554 - sparse_categorical_accuracy: 0.8728 - val_loss: 0.3786 - val_sparse_categorical_accuracy: 0.8617\n",
      "Epoch 12/30\n",
      "1500/1500 - 4s - loss: 0.3491 - sparse_categorical_accuracy: 0.8756 - val_loss: 0.3749 - val_sparse_categorical_accuracy: 0.8668\n",
      "Epoch 13/30\n",
      "1500/1500 - 4s - loss: 0.3435 - sparse_categorical_accuracy: 0.8763 - val_loss: 0.3725 - val_sparse_categorical_accuracy: 0.8652\n",
      "Epoch 14/30\n",
      "1500/1500 - 4s - loss: 0.3368 - sparse_categorical_accuracy: 0.8783 - val_loss: 0.3672 - val_sparse_categorical_accuracy: 0.8658\n",
      "Epoch 15/30\n",
      "1500/1500 - 4s - loss: 0.3326 - sparse_categorical_accuracy: 0.8799 - val_loss: 0.3730 - val_sparse_categorical_accuracy: 0.8673\n",
      "Epoch 16/30\n",
      "1500/1500 - 4s - loss: 0.3272 - sparse_categorical_accuracy: 0.8819 - val_loss: 0.3546 - val_sparse_categorical_accuracy: 0.8722\n",
      "Epoch 17/30\n",
      "1500/1500 - 4s - loss: 0.3225 - sparse_categorical_accuracy: 0.8836 - val_loss: 0.3538 - val_sparse_categorical_accuracy: 0.8705\n",
      "Epoch 18/30\n",
      "1500/1500 - 4s - loss: 0.3183 - sparse_categorical_accuracy: 0.8848 - val_loss: 0.3519 - val_sparse_categorical_accuracy: 0.8707\n",
      "Epoch 19/30\n",
      "1500/1500 - 4s - loss: 0.3137 - sparse_categorical_accuracy: 0.8862 - val_loss: 0.3474 - val_sparse_categorical_accuracy: 0.8746\n",
      "Epoch 20/30\n",
      "1500/1500 - 4s - loss: 0.3110 - sparse_categorical_accuracy: 0.8874 - val_loss: 0.3670 - val_sparse_categorical_accuracy: 0.8667\n",
      "Epoch 21/30\n",
      "1500/1500 - 5s - loss: 0.3065 - sparse_categorical_accuracy: 0.8889 - val_loss: 0.3557 - val_sparse_categorical_accuracy: 0.8698\n",
      "Epoch 22/30\n",
      "1500/1500 - 4s - loss: 0.3016 - sparse_categorical_accuracy: 0.8910 - val_loss: 0.3472 - val_sparse_categorical_accuracy: 0.8737\n",
      "Epoch 23/30\n",
      "1500/1500 - 5s - loss: 0.2991 - sparse_categorical_accuracy: 0.8913 - val_loss: 0.3394 - val_sparse_categorical_accuracy: 0.8788\n",
      "Epoch 24/30\n",
      "1500/1500 - 4s - loss: 0.2956 - sparse_categorical_accuracy: 0.8926 - val_loss: 0.3380 - val_sparse_categorical_accuracy: 0.8790\n",
      "Epoch 25/30\n",
      "1500/1500 - 4s - loss: 0.2927 - sparse_categorical_accuracy: 0.8939 - val_loss: 0.3392 - val_sparse_categorical_accuracy: 0.8789\n",
      "Epoch 26/30\n",
      "1500/1500 - 4s - loss: 0.2890 - sparse_categorical_accuracy: 0.8953 - val_loss: 0.3368 - val_sparse_categorical_accuracy: 0.8796\n",
      "Epoch 27/30\n",
      "1500/1500 - 4s - loss: 0.2857 - sparse_categorical_accuracy: 0.8965 - val_loss: 0.3357 - val_sparse_categorical_accuracy: 0.8787\n",
      "Epoch 28/30\n",
      "1500/1500 - 4s - loss: 0.2826 - sparse_categorical_accuracy: 0.8970 - val_loss: 0.3391 - val_sparse_categorical_accuracy: 0.8784\n",
      "Epoch 29/30\n",
      "1500/1500 - 4s - loss: 0.2797 - sparse_categorical_accuracy: 0.8985 - val_loss: 0.3360 - val_sparse_categorical_accuracy: 0.8800\n",
      "Epoch 30/30\n",
      "1500/1500 - 4s - loss: 0.2759 - sparse_categorical_accuracy: 0.9000 - val_loss: 0.3358 - val_sparse_categorical_accuracy: 0.8792\n",
      "157/157 [==============================] - 0s 2ms/step - loss: 0.3635 - sparse_categorical_accuracy: 0.8746\n",
      "0.8745999932289124 0.3634796738624573 (5000, 28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='elu'),\n",
    "  tf.keras.layers.Dense(100,activation='elu'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.SGD(),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    x_train,y_train,\n",
    "    epochs=30,\n",
    "    validation_split=0.2,verbose=2\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 488
    },
    "id": "VkFJ5XXqpCK0",
    "outputId": "fb61a7f6-1fa5-48cc-d204-16f03a648737"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "157/157 [==============================] - 0s 2ms/step - loss: 0.3635 - sparse_categorical_accuracy: 0.8746\n",
      "0.8745999932289124 0.3634796738624573 (5000, 28, 28)\n",
      "[[418   1  10  25   4   0  39   0  10   0]\n",
      " [  0 466   0  15   0   0   0   0   0   0]\n",
      " [  7   0 393  16  69   0  32   1   3   0]\n",
      " [  7   1   6 467   7   1   7   0   4   0]\n",
      " [  0   0  33  30 433   0  23   0   2   0]\n",
      " [  0   0   0   0   0 465   0  12   1   7]\n",
      " [ 51   1  44  25  44   0 304   0  13   0]\n",
      " [  0   0   0   0   0  18   0 473   0   9]\n",
      " [  3   0   1   6   2   1   4   4 505   0]\n",
      " [  0   0   0   0   0   7   0  21   0 449]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7799b950b8>"
      ]
     },
     "execution_count": 40,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)\n",
    "predict = model.predict(test1X)\n",
    "result = []\n",
    "for s in range(len(predict)):\n",
    "  maximum = max(predict[s])\n",
    "  pre = [i for i, j in enumerate(predict[s]) if j == maximum]\n",
    "  result.append(pre[0])\n",
    "\n",
    "cf_matrix = confusion_matrix(test1y, result)\n",
    "print(cf_matrix)\n",
    "sns.heatmap(cf_matrix)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "z425GM3XqaBv",
    "outputId": "c22e46f8-7f56-4c6c-d8a3-995e3c56f895"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1500/1500 - 4s - loss: 0.7526 - sparse_categorical_accuracy: 0.7605 - val_loss: 0.5546 - val_sparse_categorical_accuracy: 0.8150\n",
      "Epoch 2/30\n",
      "1500/1500 - 4s - loss: 0.5210 - sparse_categorical_accuracy: 0.8253 - val_loss: 0.5018 - val_sparse_categorical_accuracy: 0.8238\n",
      "Epoch 3/30\n",
      "1500/1500 - 4s - loss: 0.4733 - sparse_categorical_accuracy: 0.8392 - val_loss: 0.4583 - val_sparse_categorical_accuracy: 0.8405\n",
      "Epoch 4/30\n",
      "1500/1500 - 4s - loss: 0.4473 - sparse_categorical_accuracy: 0.8471 - val_loss: 0.4493 - val_sparse_categorical_accuracy: 0.8390\n",
      "Epoch 5/30\n",
      "1500/1500 - 4s - loss: 0.4292 - sparse_categorical_accuracy: 0.8531 - val_loss: 0.4350 - val_sparse_categorical_accuracy: 0.8445\n",
      "Epoch 6/30\n",
      "1500/1500 - 4s - loss: 0.4151 - sparse_categorical_accuracy: 0.8574 - val_loss: 0.4246 - val_sparse_categorical_accuracy: 0.8521\n",
      "Epoch 7/30\n",
      "1500/1500 - 4s - loss: 0.4041 - sparse_categorical_accuracy: 0.8604 - val_loss: 0.4079 - val_sparse_categorical_accuracy: 0.8576\n",
      "Epoch 8/30\n",
      "1500/1500 - 4s - loss: 0.3938 - sparse_categorical_accuracy: 0.8650 - val_loss: 0.3990 - val_sparse_categorical_accuracy: 0.8612\n",
      "Epoch 9/30\n",
      "1500/1500 - 4s - loss: 0.3852 - sparse_categorical_accuracy: 0.8666 - val_loss: 0.3977 - val_sparse_categorical_accuracy: 0.8600\n",
      "Epoch 10/30\n",
      "1500/1500 - 4s - loss: 0.3766 - sparse_categorical_accuracy: 0.8702 - val_loss: 0.3874 - val_sparse_categorical_accuracy: 0.8638\n",
      "Epoch 11/30\n",
      "1500/1500 - 4s - loss: 0.3702 - sparse_categorical_accuracy: 0.8724 - val_loss: 0.3813 - val_sparse_categorical_accuracy: 0.8668\n",
      "Epoch 12/30\n",
      "1500/1500 - 4s - loss: 0.3624 - sparse_categorical_accuracy: 0.8755 - val_loss: 0.3770 - val_sparse_categorical_accuracy: 0.8668\n",
      "Epoch 13/30\n",
      "1500/1500 - 4s - loss: 0.3583 - sparse_categorical_accuracy: 0.8757 - val_loss: 0.3802 - val_sparse_categorical_accuracy: 0.8661\n",
      "Epoch 14/30\n",
      "1500/1500 - 4s - loss: 0.3519 - sparse_categorical_accuracy: 0.8769 - val_loss: 0.3728 - val_sparse_categorical_accuracy: 0.8685\n",
      "Epoch 15/30\n",
      "1500/1500 - 4s - loss: 0.3468 - sparse_categorical_accuracy: 0.8792 - val_loss: 0.3863 - val_sparse_categorical_accuracy: 0.8630\n",
      "Epoch 16/30\n",
      "1500/1500 - 4s - loss: 0.3415 - sparse_categorical_accuracy: 0.8810 - val_loss: 0.3743 - val_sparse_categorical_accuracy: 0.8675\n",
      "Epoch 17/30\n",
      "1500/1500 - 4s - loss: 0.3369 - sparse_categorical_accuracy: 0.8836 - val_loss: 0.3624 - val_sparse_categorical_accuracy: 0.8727\n",
      "Epoch 18/30\n",
      "1500/1500 - 4s - loss: 0.3319 - sparse_categorical_accuracy: 0.8843 - val_loss: 0.3566 - val_sparse_categorical_accuracy: 0.8754\n",
      "Epoch 19/30\n",
      "1500/1500 - 4s - loss: 0.3269 - sparse_categorical_accuracy: 0.8850 - val_loss: 0.3575 - val_sparse_categorical_accuracy: 0.8738\n",
      "Epoch 20/30\n",
      "1500/1500 - 4s - loss: 0.3228 - sparse_categorical_accuracy: 0.8872 - val_loss: 0.3589 - val_sparse_categorical_accuracy: 0.8737\n",
      "Epoch 21/30\n",
      "1500/1500 - 4s - loss: 0.3194 - sparse_categorical_accuracy: 0.8872 - val_loss: 0.3565 - val_sparse_categorical_accuracy: 0.8724\n",
      "Epoch 22/30\n",
      "1500/1500 - 4s - loss: 0.3145 - sparse_categorical_accuracy: 0.8894 - val_loss: 0.3551 - val_sparse_categorical_accuracy: 0.8729\n",
      "Epoch 23/30\n",
      "1500/1500 - 4s - loss: 0.3113 - sparse_categorical_accuracy: 0.8901 - val_loss: 0.3468 - val_sparse_categorical_accuracy: 0.8767\n",
      "Epoch 24/30\n",
      "1500/1500 - 4s - loss: 0.3077 - sparse_categorical_accuracy: 0.8911 - val_loss: 0.3483 - val_sparse_categorical_accuracy: 0.8749\n",
      "Epoch 25/30\n",
      "1500/1500 - 4s - loss: 0.3043 - sparse_categorical_accuracy: 0.8932 - val_loss: 0.3434 - val_sparse_categorical_accuracy: 0.8792\n",
      "Epoch 26/30\n",
      "1500/1500 - 4s - loss: 0.2992 - sparse_categorical_accuracy: 0.8949 - val_loss: 0.3396 - val_sparse_categorical_accuracy: 0.8799\n",
      "Epoch 27/30\n",
      "1500/1500 - 4s - loss: 0.2964 - sparse_categorical_accuracy: 0.8964 - val_loss: 0.3399 - val_sparse_categorical_accuracy: 0.8784\n",
      "Epoch 28/30\n",
      "1500/1500 - 4s - loss: 0.2939 - sparse_categorical_accuracy: 0.8971 - val_loss: 0.3543 - val_sparse_categorical_accuracy: 0.8759\n",
      "Epoch 29/30\n",
      "1500/1500 - 4s - loss: 0.2901 - sparse_categorical_accuracy: 0.8983 - val_loss: 0.3428 - val_sparse_categorical_accuracy: 0.8783\n",
      "Epoch 30/30\n",
      "1500/1500 - 4s - loss: 0.2875 - sparse_categorical_accuracy: 0.8990 - val_loss: 0.3385 - val_sparse_categorical_accuracy: 0.8787\n",
      "157/157 [==============================] - 0s 2ms/step - loss: 0.3652 - sparse_categorical_accuracy: 0.8648\n",
      "0.864799976348877 0.3652113080024719 (5000, 28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(300,activation='relu'),\n",
    "  tf.keras.layers.Dense(10, activation='softmax')\n",
    "])\n",
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.SGD(),\n",
    "    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "    metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    x_train,y_train,\n",
    "    epochs=30,\n",
    "    validation_split=0.2,verbose=2\n",
    ")\n",
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.plot()\n",
    "\n",
    "loss, accuracy = model.evaluate(test1X, test1y)\n",
    "print(accuracy, loss, test1X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "4JkltOAtzwIj"
   },
   "outputs": [],
   "source": [
    "def encoder(input_img):\n",
    "    #encoder\n",
    "    #input = 28 x 28 x 1 (wide and thin)\n",
    "    conv1 = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img) #28 x 28 x 32\n",
    "    conv1 = BatchNormalization()(conv1)\n",
    "    # conv1 = Conv2D(32, (3, 3), activation='relu', padding='same')(conv1)\n",
    "    # conv1 = BatchNormalization()(conv1)\n",
    "    pool1 = MaxPooling2D(pool_size=(2, 2))(conv1) #14 x 14 x 32\n",
    "    conv2 = Conv2D(64, (3, 3), activation='relu', padding='same')(pool1) #14 x 14 x 64\n",
    "    conv2 = BatchNormalization()(conv2)\n",
    "    # conv2 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv2)\n",
    "    # conv2 = BatchNormalization()(conv2)\n",
    "    pool2 = MaxPooling2D(pool_size=(2, 2))(conv2) #7 x 7 x 64\n",
    "    conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(pool2) #7 x 7 x 128 (small and thick)\n",
    "    conv3 = BatchNormalization()(conv3)\n",
    "    # conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv3)\n",
    "    # conv3 = BatchNormalization()(conv3)\n",
    "    # conv4 = Conv2D(256, (3, 3), activation='relu', padding='same')(conv3) #7 x 7 x 256 (small and thick)\n",
    "    # conv4 = BatchNormalization()(conv4)\n",
    "    # conv4 = Conv2D(256, (3, 3), activation='relu', padding='same')(conv4)\n",
    "    # conv4 = BatchNormalization()(conv4)\n",
    "    return conv3\n",
    "\n",
    "def decoder(conv4):    \n",
    "  #decoder\n",
    "  conv5 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv4) #7 x 7 x 128\n",
    "  conv5 = BatchNormalization()(conv5)\n",
    "  # conv5 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv5)\n",
    "  # conv5 = BatchNormalization()(conv5)\n",
    "  conv6 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv5) #7 x 7 x 64\n",
    "  conv6 = BatchNormalization()(conv6)\n",
    "  # conv6 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv6)\n",
    "  # conv6 = BatchNormalization()(conv6)\n",
    "  up1 = UpSampling2D((2,2))(conv6) #14 x 14 x 64\n",
    "  conv7 = Conv2D(32, (3, 3), activation='relu', padding='same')(up1) # 14 x 14 x 32\n",
    "  conv7 = BatchNormalization()(conv7)\n",
    "  # conv7 = Conv2D(32, (3, 3), activation='relu', padding='same')(conv7)\n",
    "  # conv7 = BatchNormalization()(conv7)\n",
    "  up2 = UpSampling2D((2,2))(conv7) # 28 x 28 x 32\n",
    "  decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(up2) # 28 x 28 x 1\n",
    "  return decoded"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "g1X7iqy6bH_W",
    "outputId": "8e162084-3a7d-4190-b487-f90d483143d2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train: X=(60000, 28, 28), y=(60000,)\n",
      "Test: X=(5000, 28, 28), y=(5000,)\n",
      "Validation: X=(5000, 28, 28), y=(5000,)\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Input,Dense,Flatten,Dropout,merge,Reshape,Conv2D,MaxPooling2D,UpSampling2D,Conv2DTranspose\n",
    "batch_size = 64\n",
    "epochs = 10\n",
    "inChannel = 1\n",
    "x, y = 28, 28\n",
    "input_img = Input(shape = (x, y, inChannel))\n",
    "num_classes = 10\n",
    "import keras\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from matplotlib import pyplot\n",
    "# Load dataset\n",
    "(trainX, trainy), (testX, testy) = keras.datasets.fashion_mnist.load_data()\n",
    "# Summarize loaded dataset\n",
    "(test1X, test1y) = (testX[0:5000], testy[0:5000])\n",
    "(validX, validy) = (testX[5000:], testy[5000:])\n",
    "print('Train: X=%s, y=%s' % (trainX.shape, trainy.shape))\n",
    "print('Test: X=%s, y=%s' % (test1X.shape, test1y.shape))\n",
    "print('Validation: X=%s, y=%s' % (validX.shape, validy.shape))\n",
    "\n",
    "train_data = trainX / np.max(trainX)\n",
    "test_data = testX / np.max(testy)\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "train_X,valid_X,train_ground,valid_ground = train_test_split(train_data,\n",
    "                                                             train_data,\n",
    "                                                             test_size=0.2,\n",
    "                                                             random_state=13)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "yg_eoMCebkwQ"
   },
   "outputs": [],
   "source": [
    "from keras.models import Model,Sequential\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.optimizers import RMSprop\n",
    "autoencoder = Model(input_img, decoder(encoder(input_img)))\n",
    "autoencoder.compile(loss='mean_squared_error', optimizer = RMSprop())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "fKWvebNpbk1i",
    "outputId": "665e6035-8850-42ab-ed1e-617d13c250c4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "750/750 [==============================] - 304s 405ms/step - loss: 0.0036 - val_loss: 0.0033\n"
     ]
    }
   ],
   "source": [
    "autoencoder_train = autoencoder.fit(train_X, train_ground, batch_size=batch_size,epochs=1,verbose=1,validation_data=(valid_X, valid_ground))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "3YaOVst159tX"
   },
   "outputs": [],
   "source": [
    "def fc(enco):\n",
    "    flat = Flatten()(enco)\n",
    "    den = Dense(128, activation='relu')(flat)\n",
    "    out = Dense(num_classes, activation='softmax')(den)\n",
    "    return out\n",
    "encode = encoder(input_img)\n",
    "full_model = Model(input_img,fc(encode))\n",
    "full_model.compile(loss=keras.losses.categorical_crossentropy, optimizer=keras.optimizers.Adam(),metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "XguwtS1Xxc2-"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "neural_network_project.txt",
   "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": "3.8.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}