03-NeuralNetworks/02-TensorFlow/KerasTF-Seminar.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Решение практических задач с Keras и TensorFlow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow_datasets as tdfs\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>features/carat</th>\n",
       "      <th>features/clarity</th>\n",
       "      <th>features/color</th>\n",
       "      <th>features/cut</th>\n",
       "      <th>features/depth</th>\n",
       "      <th>features/table</th>\n",
       "      <th>features/x</th>\n",
       "      <th>features/y</th>\n",
       "      <th>features/z</th>\n",
       "      <th>price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.26</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>2</td>\n",
       "      <td>60.599998</td>\n",
       "      <td>60.0</td>\n",
       "      <td>6.97</td>\n",
       "      <td>7.00</td>\n",
       "      <td>4.23</td>\n",
       "      <td>6546.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.80</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>62.099998</td>\n",
       "      <td>54.0</td>\n",
       "      <td>5.96</td>\n",
       "      <td>5.99</td>\n",
       "      <td>3.71</td>\n",
       "      <td>3030.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.56</td>\n",
       "      <td>4</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>61.700001</td>\n",
       "      <td>54.0</td>\n",
       "      <td>5.28</td>\n",
       "      <td>5.32</td>\n",
       "      <td>3.27</td>\n",
       "      <td>1915.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.51</td>\n",
       "      <td>3</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>64.000000</td>\n",
       "      <td>58.0</td>\n",
       "      <td>7.24</td>\n",
       "      <td>7.27</td>\n",
       "      <td>4.64</td>\n",
       "      <td>6936.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.33</td>\n",
       "      <td>6</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>62.200001</td>\n",
       "      <td>54.0</td>\n",
       "      <td>4.43</td>\n",
       "      <td>4.45</td>\n",
       "      <td>2.76</td>\n",
       "      <td>646.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   features/carat  features/clarity  features/color  features/cut  \\\n",
       "0            1.26                 2               4             2   \n",
       "1            0.80                 3               4             4   \n",
       "2            0.56                 4               2             4   \n",
       "3            1.51                 3               6             1   \n",
       "4            0.33                 6               5             4   \n",
       "\n",
       "   features/depth  features/table  features/x  features/y  features/z   price  \n",
       "0       60.599998            60.0        6.97        7.00        4.23  6546.0  \n",
       "1       62.099998            54.0        5.96        5.99        3.71  3030.0  \n",
       "2       61.700001            54.0        5.28        5.32        3.27  1915.0  \n",
       "3       64.000000            58.0        7.24        7.27        4.64  6936.0  \n",
       "4       62.200001            54.0        4.43        4.45        2.76   646.0  "
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = tdfs.as_dataframe(tdfs.load('diamonds',split='train'))\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "for c in df.columns:\n",
    "    if c.startswith('features/'):\n",
    "        df[c] = (df[c]-df[c].min())/(df[c].max()-df[c].min())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "# df['price'] = 1 + (df['price']-df['price'].min())/(df['price'].max()-df['price'].min())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>features/carat</th>\n",
       "      <th>features/clarity</th>\n",
       "      <th>features/color</th>\n",
       "      <th>features/cut</th>\n",
       "      <th>features/depth</th>\n",
       "      <th>features/table</th>\n",
       "      <th>features/x</th>\n",
       "      <th>features/y</th>\n",
       "      <th>features/z</th>\n",
       "      <th>price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.220374</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.488889</td>\n",
       "      <td>0.326923</td>\n",
       "      <td>0.648976</td>\n",
       "      <td>0.118846</td>\n",
       "      <td>0.133019</td>\n",
       "      <td>1.336271</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.124740</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.530555</td>\n",
       "      <td>0.211538</td>\n",
       "      <td>0.554935</td>\n",
       "      <td>0.101698</td>\n",
       "      <td>0.116667</td>\n",
       "      <td>1.146186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.074844</td>\n",
       "      <td>0.571429</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.519444</td>\n",
       "      <td>0.211538</td>\n",
       "      <td>0.491620</td>\n",
       "      <td>0.090323</td>\n",
       "      <td>0.102830</td>\n",
       "      <td>1.085906</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.272349</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.583333</td>\n",
       "      <td>0.288462</td>\n",
       "      <td>0.674115</td>\n",
       "      <td>0.123430</td>\n",
       "      <td>0.145912</td>\n",
       "      <td>1.357355</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.027027</td>\n",
       "      <td>0.857143</td>\n",
       "      <td>0.833333</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.211538</td>\n",
       "      <td>0.412477</td>\n",
       "      <td>0.075552</td>\n",
       "      <td>0.086792</td>\n",
       "      <td>1.017300</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   features/carat  features/clarity  features/color  features/cut  \\\n",
       "0        0.220374          0.285714        0.666667          0.50   \n",
       "1        0.124740          0.428571        0.666667          1.00   \n",
       "2        0.074844          0.571429        0.333333          1.00   \n",
       "3        0.272349          0.428571        1.000000          0.25   \n",
       "4        0.027027          0.857143        0.833333          1.00   \n",
       "\n",
       "   features/depth  features/table  features/x  features/y  features/z  \\\n",
       "0        0.488889        0.326923    0.648976    0.118846    0.133019   \n",
       "1        0.530555        0.211538    0.554935    0.101698    0.116667   \n",
       "2        0.519444        0.211538    0.491620    0.090323    0.102830   \n",
       "3        0.583333        0.288462    0.674115    0.123430    0.145912   \n",
       "4        0.533333        0.211538    0.412477    0.075552    0.086792   \n",
       "\n",
       "      price  \n",
       "0  1.336271  \n",
       "1  1.146186  \n",
       "2  1.085906  \n",
       "3  1.357355  \n",
       "4  1.017300  "
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df['price'].hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df[[x for x in df.columns if x.startswith('features/')]]\n",
    "Y = df['price']\n",
    "train_x, test_x, train_y, test_y = train_test_split(X,Y,test_size=0.2,shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.Sequential([\n",
    "    keras.layers.Dense(100,input_shape=(9,),activation='sigmoid'),\n",
    "    keras.layers.Dense(1,activation=None)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0212 - MAE: 0.0903 - MAPE: 7.0820 - val_loss: 0.0052 - val_MAE: 0.0526 - val_MAPE: 4.1999\n",
      "Epoch 2/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0052 - MAE: 0.0523 - MAPE: 4.2153 - val_loss: 0.0049 - val_MAE: 0.0508 - val_MAPE: 4.1027\n",
      "Epoch 3/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0051 - MAE: 0.0512 - MAPE: 4.1374 - val_loss: 0.0047 - val_MAE: 0.0490 - val_MAPE: 3.9433\n",
      "Epoch 4/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0049 - MAE: 0.0503 - MAPE: 4.0588 - val_loss: 0.0057 - val_MAE: 0.0514 - val_MAPE: 4.1265\n",
      "Epoch 5/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0048 - MAE: 0.0488 - MAPE: 3.9233 - val_loss: 0.0045 - val_MAE: 0.0472 - val_MAPE: 3.7776\n",
      "Epoch 6/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0047 - MAE: 0.0481 - MAPE: 3.8626 - val_loss: 0.0048 - val_MAE: 0.0517 - val_MAPE: 4.2115\n",
      "Epoch 7/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0047 - MAE: 0.0478 - MAPE: 3.8361 - val_loss: 0.0054 - val_MAE: 0.0566 - val_MAPE: 4.6309\n",
      "Epoch 8/10\n",
      "1349/1349 [==============================] - 4s 3ms/step - loss: 0.0046 - MAE: 0.0469 - MAPE: 3.7602 - val_loss: 0.0043 - val_MAE: 0.0446 - val_MAPE: 3.5477\n",
      "Epoch 9/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0046 - MAE: 0.0465 - MAPE: 3.7189 - val_loss: 0.0044 - val_MAE: 0.0477 - val_MAPE: 3.8375 loss: 0.0046 - MAE: 0.0\n",
      "Epoch 10/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0046 - MAE: 0.0467 - MAPE: 3.7375 - val_loss: 0.0049 - val_MAE: 0.0523 - val_MAPE: 4.2628\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1ee8b81d580>"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.compile(loss='MSE',optimizer='adam',metrics=['MAE','MAPE'])\n",
    "model.fit(train_x,train_y,validation_data=(test_x,test_y),epochs=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = tdfs.as_dataframe(tdfs.load('diamonds',split='train'))\n",
    "\n",
    "df['logprice'] = np.log(df['price'])\n",
    "df['logprice'].hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "for c in df.columns:\n",
    "    if c.startswith('features/'):\n",
    "        df[c] = (df[c]-df[c].min())/(df[c].max()-df[c].min())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df[[x for x in df.columns if x.startswith('features/')]]\n",
    "Y = df['logprice']\n",
    "train_x, test_x, train_y, test_y = train_test_split(X,Y,test_size=0.2,shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.Sequential([\n",
    "    keras.layers.Dense(100,input_shape=(9,),activation='tanh'),\n",
    "    keras.layers.Dense(1,activation=None)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.2944 - MAE: 0.3137 - MAPE: 4.1066 - val_loss: 0.0570 - val_MAE: 0.1827 - val_MAPE: 2.33650.8306 - MAE: 0.5 - E\n",
      "Epoch 2/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0533 - MAE: 0.1739 - MAPE: 2.2332 - val_loss: 0.0530 - val_MAE: 0.1683 - val_MAPE: 2.1816\n",
      "Epoch 3/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0500 - MAE: 0.1673 - MAPE: 2.1472 - val_loss: 0.0496 - val_MAE: 0.1648 - val_MAPE: 2.1244\n",
      "Epoch 4/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0473 - MAE: 0.1622 - MAPE: 2.0827 - val_loss: 0.0498 - val_MAE: 0.1691 - val_MAPE: 2.1496\n",
      "Epoch 5/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0448 - MAE: 0.1571 - MAPE: 2.0191 - val_loss: 0.0457 - val_MAE: 0.1540 - val_MAPE: 2.0015\n",
      "Epoch 6/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0426 - MAE: 0.1526 - MAPE: 1.9620 - val_loss: 0.0453 - val_MAE: 0.1508 - val_MAPE: 1.9738\n",
      "Epoch 7/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0406 - MAE: 0.1482 - MAPE: 1.9070 - val_loss: 0.0417 - val_MAE: 0.1452 - val_MAPE: 1.8822\n",
      "Epoch 8/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0388 - MAE: 0.1442 - MAPE: 1.8575 - val_loss: 0.0399 - val_MAE: 0.1433 - val_MAPE: 1.8447\n",
      "Epoch 9/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0371 - MAE: 0.1402 - MAPE: 1.8068 - val_loss: 0.0391 - val_MAE: 0.1379 - val_MAPE: 1.7978\n",
      "Epoch 10/10\n",
      "1349/1349 [==============================] - 3s 2ms/step - loss: 0.0355 - MAE: 0.1367 - MAPE: 1.7636 - val_loss: 0.0375 - val_MAE: 0.1346 - val_MAPE: 1.7528\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1ee9b7415e0>"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.compile(loss='MSE',optimizer='sgd',metrics=['MAE','MAPE'])\n",
    "model.fit(train_x,train_y,validation_data=(test_x,test_y),epochs=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## MNIST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "(train_x, train_y), (test_x, test_y) = tf.keras.datasets.mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((60000, 28, 28), 255)"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_x.shape, train_x.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x = train_x.astype(np.float32)/256.0\n",
    "test_x = test_x.astype(np.float32)/256.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 100)               78500     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                1010      \n",
      "=================================================================\n",
      "Total params: 79,510\n",
      "Trainable params: 79,510\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = keras.Sequential([\n",
    "    keras.layers.Flatten(input_shape=(28,28)),\n",
    "    keras.layers.Dense(100,activation='relu'),\n",
    "    keras.layers.Dense(10,activation='softmax')\n",
    "])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((60000,), (60000, 10))"
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_y_oh = keras.utils.to_categorical(train_y,10)\n",
    "test_y_oh = keras.utils.to_categorical(test_y,10)\n",
    "train_y.shape, train_y_oh.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.2741 - acc: 0.9220 - val_loss: 0.1496 - val_acc: 0.9545\n",
      "Epoch 2/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1239 - acc: 0.9635 - val_loss: 0.1080 - val_acc: 0.9686\n",
      "Epoch 3/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0849 - acc: 0.9746 - val_loss: 0.1049 - val_acc: 0.9669\n",
      "Epoch 4/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0653 - acc: 0.9797 - val_loss: 0.0873 - val_acc: 0.9726\n",
      "Epoch 5/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0509 - acc: 0.9844 - val_loss: 0.0756 - val_acc: 0.97630.\n",
      "Epoch 6/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0411 - acc: 0.9871 - val_loss: 0.0827 - val_acc: 0.9743 0.\n",
      "Epoch 7/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0342 - acc: 0.9893 - val_loss: 0.0805 - val_acc: 0.9768\n",
      "Epoch 8/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0277 - acc: 0.9912 - val_loss: 0.0754 - val_acc: 0.9768\n",
      "Epoch 9/10\n",
      "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0222 - acc: 0.9933 - val_loss: 0.0791 - val_acc: 0.9787\n",
      "Epoch 10/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0201 - acc: 0.9940 - val_loss: 0.0855 - val_acc: 0.9768\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='categorical_crossentropy',optimizer='adam',metrics=['acc'])\n",
    "hist = model.fit(train_x,train_y_oh,validation_data=(test_x,test_y_oh),epochs=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(hist.history['acc'])\n",
    "plt.plot(hist.history['val_acc'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1875/1875 [==============================] - 13s 7ms/step - loss: 0.2758 - acc: 0.9209 - val_loss: 0.1539 - val_acc: 0.9557\n",
      "Epoch 2/10\n",
      "1875/1875 [==============================] - 21s 11ms/step - loss: 0.1255 - acc: 0.9625 - val_loss: 0.1060 - val_acc: 0.9677\n",
      "Epoch 3/10\n",
      "1875/1875 [==============================] - 26s 14ms/step - loss: 0.0875 - acc: 0.9734 - val_loss: 0.0973 - val_acc: 0.9701\n",
      "Epoch 4/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0675 - acc: 0.9795 - val_loss: 0.0787 - val_acc: 0.9773\n",
      "Epoch 5/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0531 - acc: 0.9836 - val_loss: 0.0804 - val_acc: 0.9742\n",
      "Epoch 6/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0430 - acc: 0.9867 - val_loss: 0.0810 - val_acc: 0.9762\n",
      "Epoch 7/10\n",
      "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0352 - acc: 0.9892 - val_loss: 0.0786 - val_acc: 0.9777\n",
      "Epoch 8/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0288 - acc: 0.9910 - val_loss: 0.0869 - val_acc: 0.9759\n",
      "Epoch 9/10\n",
      "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0252 - acc: 0.9925 - val_loss: 0.0796 - val_acc: 0.9774\n",
      "Epoch 10/10\n",
      "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0199 - acc: 0.9939 - val_loss: 0.0840 - val_acc: 0.9775\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x22f000b1d90>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.compile(loss='sparse_categorical_crossentropy',optimizer='adam',metrics=['acc'])\n",
    "model.fit(train_x,train_y,validation_data=(test_x,test_y),epochs=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([6.97525522e-08, 1.45022785e-08, 6.36827565e-07, 8.77206912e-05,\n",
       "        2.63948086e-12, 7.80755016e-08, 5.54000364e-16, 9.99837518e-01,\n",
       "        7.80763969e-08, 7.38739982e-05], dtype=float32),\n",
       " <tf.Tensor: shape=(), dtype=int64, numpy=7>)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_p = model.predict(test_x)\n",
    "pred_c = tf.argmax(pred_p,axis=1)\n",
    "pred_p[0],pred_c[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.9399166703224182, 0.9326000213623047)"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def run(activation, n_neurons):\n",
    "    model = keras.Sequential([\n",
    "        keras.layers.Flatten(input_shape=(28,28)),\n",
    "        keras.layers.Dense(n_neurons,activation=activation),\n",
    "        keras.layers.Dense(10,activation='softmax')])\n",
    "    estop = tf.keras.callbacks.EarlyStopping(monitor='val_acc', patience=3)\n",
    "    model.compile(loss='sparse_categorical_crossentropy',optimizer='adam',metrics=['acc'])\n",
    "    hist = model.fit(train_x,train_y,validation_data=(test_x,test_y),epochs=10,verbose=False,callbacks=[estop])\n",
    "    return np.max(hist.history['acc']), np.max(hist.history['val_acc'])\n",
    "\n",
    "run('tanh',10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "relu,5 -> train=0.8934833407402039, val=0.8902999758720398\n",
      "relu,10 -> train=0.9408166408538818, val=0.9362999796867371\n",
      "relu,100 -> train=0.9935666918754578, val=0.9775999784469604\n",
      "tanh,5 -> train=0.8752666711807251, val=0.8736000061035156\n",
      "tanh,10 -> train=0.9403499960899353, val=0.9329000115394592\n",
      "tanh,100 -> train=0.995033323764801, val=0.9767000079154968\n",
      "sigmoid,5 -> train=0.8840333223342896, val=0.8819000124931335\n",
      "sigmoid,10 -> train=0.9314333200454712, val=0.9243999719619751\n",
      "sigmoid,100 -> train=0.9892500042915344, val=0.9768999814987183\n"
     ]
    }
   ],
   "source": [
    "for act in ['relu','tanh','sigmoid']:\n",
    "    for n in [5,10,100]:\n",
    "        acc, vacc = run(act,n)\n",
    "        print(f\"{act},{n} -> train={acc}, val={vacc}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.12 ('base')",
   "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": "3.9.12"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "d21c24a7952cc9fa93a33c3da71b373ffe761d6ab0c9cc9b2153e31da51ffb58"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}