notebook_keras.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "\n",
    "from keras import layers\n",
    "from keras import regularizers\n",
    "from keras import backend as K\n",
    "from keras import Sequential, Input\n",
    "from keras.optimizers import SGD, Adam\n",
    "from keras.losses import categorical_crossentropy\n",
    "from keras.callbacks import LearningRateScheduler\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import math\n",
    "\n",
    "from utils.constants import CLASSES, IMAGE_SIZE\n",
    "from utils.load_dataset import load_dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# make the result reproducible\n",
    "keras.utils.set_random_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'), PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]\n"
     ]
    }
   ],
   "source": [
    "physical_devices = tf.config.list_physical_devices()\n",
    "print(physical_devices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.config.set_visible_devices(physical_devices[1], 'GPU')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Attributes description\n",
    "\n",
    "- a1 - face containing flag: (1-with face, 0-without face),\n",
    "- a2 - image number in current class (person) beginning from 0,\n",
    "- a3 - class (person) number beginning from 0,\n",
    "- a4 - sex (0 - woman, 1 - man)\n",
    "- a5 - race (0- white, 1 - black, 2 - indian, ...)\n",
    "- a6 - age (0 - baby, 1 - young, 2 - middle-age, 3 - old) \n",
    "- a7 - binokulars (0 - without, 1 - transparent, 2 - dark)\n",
    "- a8 - emotional expression (not state!) (0 - sad, 1 - neutral, 2 - happy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "DIRPATH = './data'\n",
    "X_train, X_test, y_train, y_test = load_dataset(DIRPATH)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building a model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Testing LeNet-5 Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " conv2d_6 (Conv2D)           (None, 24, 24, 6)         156       \n",
      "                                                                 \n",
      " max_pooling2d_4 (MaxPooling  (None, 12, 12, 6)        0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " dropout_6 (Dropout)         (None, 12, 12, 6)         0         \n",
      "                                                                 \n",
      " conv2d_7 (Conv2D)           (None, 12, 12, 16)        2416      \n",
      "                                                                 \n",
      " max_pooling2d_5 (MaxPooling  (None, 6, 6, 16)         0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " dropout_7 (Dropout)         (None, 6, 6, 16)          0         \n",
      "                                                                 \n",
      " conv2d_8 (Conv2D)           (None, 6, 6, 120)         48120     \n",
      "                                                                 \n",
      " flatten_2 (Flatten)         (None, 4320)              0         \n",
      "                                                                 \n",
      " dropout_8 (Dropout)         (None, 4320)              0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 84)                362964    \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 48)                4080      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 417,736\n",
      "Trainable params: 417,736\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from functools import partial\n",
    "\n",
    "conv_regularizer = regularizers.l2(0.0006)\n",
    "dense_regularizer = regularizers.l2(0.01)\n",
    "\n",
    "activation_def = 'tanh'\n",
    "\n",
    "# {'conv_regularizer': 0.0006124219567560658, 'dense_regularizer': 0.03011157028527752, 'dropout_1': 0.3433425343566813, 'dropout_2': 0.1625206873530064, 'dropout_3': 0.019169937499455086}.\n",
    "\n",
    "DefaultConv2D = partial(tf.keras.layers.Conv2D, kernel_size=5, padding=\"same\", activation=activation_def,\n",
    "                        kernel_regularizer=conv_regularizer) # kernel_initializer=\"he_normal\")\n",
    "\n",
    "model = Sequential(\n",
    "  [\n",
    "    Input(shape=(IMAGE_SIZE, IMAGE_SIZE, 1)),\n",
    "    DefaultConv2D(6),\n",
    "    layers.MaxPooling2D(pool_size=2, strides=2),\n",
    "\n",
    "    layers.Dropout(0.45),\n",
    "    DefaultConv2D(16),\n",
    "    layers.MaxPooling2D(pool_size=2, strides=2),\n",
    "\n",
    "    layers.Dropout(0.3),\n",
    "    DefaultConv2D(120),\n",
    "\n",
    "    layers.Flatten(),\n",
    "    layers.Dropout(0.15),\n",
    "    layers.Dense(84, activation=activation_def, kernel_regularizer=dense_regularizer),\n",
    "    layers.Dense(CLASSES, activation='softmax'),\n",
    "  ]\n",
    ")\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def get_lr_metric(optimizer):\n",
    "#   def lr(y_true, y_pred):\n",
    "#     return optimizer._decayed_lr(tf.float32) # I use ._decayed_lr method instead of .lr\n",
    "#   return lr\n",
    "\n",
    "# lr_metric = get_lr_metric(optimizer)\n",
    "# model.compile(loss=categorical_crossentropy, optimizer=optimizer, metrics=[\"accuracy\", lr_metric])\n",
    "\n",
    "\n",
    "optimizer = Adam(learning_rate=0.001)\n",
    "model.compile(loss=categorical_crossentropy, optimizer=optimizer, metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def step_decay(epoch):\n",
    "#   initial_lrate = 0.1\n",
    "#   drop = 0.9\n",
    "#   epochs_drop = 10.0\n",
    "#   lrate = initial_lrate * math.pow(drop,  \n",
    "#           math.floor((1+epoch)/epochs_drop))\n",
    "#   return lrate\n",
    " \n",
    "# lrate_scheduler_callback = LearningRateScheduler(step_decay)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "CHECKPOINT_PATH = 'models/checkpoints/famous48_test_model_v2_best.keras'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_checkpoint_callback = keras.callbacks.ModelCheckpoint(\n",
    "  filepath=CHECKPOINT_PATH,\n",
    "  save_weights_only=False,\n",
    "  monitor='val_accuracy',\n",
    "  mode='max',\n",
    "  save_best_only=True,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "22/22 [==============================] - 3s 23ms/step - loss: 4.9976 - accuracy: 0.0569 - val_loss: 4.3955 - val_accuracy: 0.0704\n",
      "Epoch 2/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 3.9023 - accuracy: 0.1763 - val_loss: 3.3405 - val_accuracy: 0.2550\n",
      "Epoch 3/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 3.1167 - accuracy: 0.2849 - val_loss: 2.7844 - val_accuracy: 0.3592\n",
      "Epoch 4/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.7607 - accuracy: 0.3640 - val_loss: 2.5245 - val_accuracy: 0.4086\n",
      "Epoch 5/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.5380 - accuracy: 0.4099 - val_loss: 2.3141 - val_accuracy: 0.4625\n",
      "Epoch 6/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.3782 - accuracy: 0.4458 - val_loss: 2.2226 - val_accuracy: 0.4872\n",
      "Epoch 7/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.2923 - accuracy: 0.4678 - val_loss: 2.0761 - val_accuracy: 0.5210\n",
      "Epoch 8/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.1880 - accuracy: 0.4945 - val_loss: 1.9677 - val_accuracy: 0.5558\n",
      "Epoch 9/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.1323 - accuracy: 0.5146 - val_loss: 1.9160 - val_accuracy: 0.5558\n",
      "Epoch 10/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 2.0543 - accuracy: 0.5299 - val_loss: 1.8691 - val_accuracy: 0.5823\n",
      "Epoch 11/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.9782 - accuracy: 0.5478 - val_loss: 1.7503 - val_accuracy: 0.6170\n",
      "Epoch 12/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.9362 - accuracy: 0.5604 - val_loss: 1.7645 - val_accuracy: 0.6097\n",
      "Epoch 13/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.9162 - accuracy: 0.5601 - val_loss: 1.7622 - val_accuracy: 0.6051\n",
      "Epoch 14/100\n",
      "22/22 [==============================] - 0s 11ms/step - loss: 1.8757 - accuracy: 0.5732 - val_loss: 1.6878 - val_accuracy: 0.6417\n",
      "Epoch 15/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.8364 - accuracy: 0.5800 - val_loss: 1.6692 - val_accuracy: 0.6298\n",
      "Epoch 16/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.8009 - accuracy: 0.5878 - val_loss: 1.5799 - val_accuracy: 0.6709\n",
      "Epoch 17/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.7782 - accuracy: 0.6052 - val_loss: 1.5930 - val_accuracy: 0.6600\n",
      "Epoch 18/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.7339 - accuracy: 0.6116 - val_loss: 1.5447 - val_accuracy: 0.6709\n",
      "Epoch 19/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.7114 - accuracy: 0.6107 - val_loss: 1.5317 - val_accuracy: 0.6645\n",
      "Epoch 20/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.6974 - accuracy: 0.6198 - val_loss: 1.5055 - val_accuracy: 0.6691\n",
      "Epoch 21/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.6466 - accuracy: 0.6365 - val_loss: 1.4393 - val_accuracy: 0.6993\n",
      "Epoch 22/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.6403 - accuracy: 0.6262 - val_loss: 1.4395 - val_accuracy: 0.6910\n",
      "Epoch 23/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.6117 - accuracy: 0.6392 - val_loss: 1.4179 - val_accuracy: 0.7020\n",
      "Epoch 24/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.6093 - accuracy: 0.6484 - val_loss: 1.4108 - val_accuracy: 0.7121\n",
      "Epoch 25/100\n",
      "22/22 [==============================] - 0s 11ms/step - loss: 1.5657 - accuracy: 0.6635 - val_loss: 1.4036 - val_accuracy: 0.7002\n",
      "Epoch 26/100\n",
      "22/22 [==============================] - 0s 11ms/step - loss: 1.5563 - accuracy: 0.6630 - val_loss: 1.4214 - val_accuracy: 0.6993\n",
      "Epoch 27/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.5601 - accuracy: 0.6587 - val_loss: 1.4071 - val_accuracy: 0.6984\n",
      "Epoch 28/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.5355 - accuracy: 0.6696 - val_loss: 1.3783 - val_accuracy: 0.7121\n",
      "Epoch 29/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.5110 - accuracy: 0.6676 - val_loss: 1.3545 - val_accuracy: 0.7239\n",
      "Epoch 30/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.5067 - accuracy: 0.6717 - val_loss: 1.3298 - val_accuracy: 0.7276\n",
      "Epoch 31/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4779 - accuracy: 0.6875 - val_loss: 1.3271 - val_accuracy: 0.7276\n",
      "Epoch 32/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4610 - accuracy: 0.6966 - val_loss: 1.3397 - val_accuracy: 0.7313\n",
      "Epoch 33/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4519 - accuracy: 0.6877 - val_loss: 1.3215 - val_accuracy: 0.7276\n",
      "Epoch 34/100\n",
      "22/22 [==============================] - 0s 11ms/step - loss: 1.4445 - accuracy: 0.6941 - val_loss: 1.2964 - val_accuracy: 0.7386\n",
      "Epoch 35/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4279 - accuracy: 0.6991 - val_loss: 1.2576 - val_accuracy: 0.7532\n",
      "Epoch 36/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4032 - accuracy: 0.7062 - val_loss: 1.2973 - val_accuracy: 0.7303\n",
      "Epoch 37/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4152 - accuracy: 0.7010 - val_loss: 1.2935 - val_accuracy: 0.7340\n",
      "Epoch 38/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.4052 - accuracy: 0.6989 - val_loss: 1.2457 - val_accuracy: 0.7550\n",
      "Epoch 39/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3843 - accuracy: 0.7149 - val_loss: 1.2460 - val_accuracy: 0.7486\n",
      "Epoch 40/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3627 - accuracy: 0.7149 - val_loss: 1.2503 - val_accuracy: 0.7651\n",
      "Epoch 41/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3658 - accuracy: 0.7154 - val_loss: 1.2455 - val_accuracy: 0.7441\n",
      "Epoch 42/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3593 - accuracy: 0.7188 - val_loss: 1.2296 - val_accuracy: 0.7687\n",
      "Epoch 43/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3491 - accuracy: 0.7202 - val_loss: 1.2411 - val_accuracy: 0.7495\n",
      "Epoch 44/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3460 - accuracy: 0.7151 - val_loss: 1.2340 - val_accuracy: 0.7495\n",
      "Epoch 45/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3285 - accuracy: 0.7279 - val_loss: 1.1943 - val_accuracy: 0.7733\n",
      "Epoch 46/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3082 - accuracy: 0.7309 - val_loss: 1.1954 - val_accuracy: 0.7687\n",
      "Epoch 47/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3187 - accuracy: 0.7215 - val_loss: 1.2101 - val_accuracy: 0.7532\n",
      "Epoch 48/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3132 - accuracy: 0.7243 - val_loss: 1.2281 - val_accuracy: 0.7468\n",
      "Epoch 49/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.3121 - accuracy: 0.7337 - val_loss: 1.1790 - val_accuracy: 0.7715\n",
      "Epoch 50/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2829 - accuracy: 0.7346 - val_loss: 1.2124 - val_accuracy: 0.7495\n",
      "Epoch 51/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2821 - accuracy: 0.7426 - val_loss: 1.1772 - val_accuracy: 0.7724\n",
      "Epoch 52/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2680 - accuracy: 0.7364 - val_loss: 1.2119 - val_accuracy: 0.7614\n",
      "Epoch 53/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2777 - accuracy: 0.7369 - val_loss: 1.1833 - val_accuracy: 0.7596\n",
      "Epoch 54/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2814 - accuracy: 0.7371 - val_loss: 1.1683 - val_accuracy: 0.7733\n",
      "Epoch 55/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2725 - accuracy: 0.7396 - val_loss: 1.1819 - val_accuracy: 0.7715\n",
      "Epoch 56/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2680 - accuracy: 0.7369 - val_loss: 1.1616 - val_accuracy: 0.7733\n",
      "Epoch 57/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2500 - accuracy: 0.7497 - val_loss: 1.1372 - val_accuracy: 0.7943\n",
      "Epoch 58/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2380 - accuracy: 0.7494 - val_loss: 1.1515 - val_accuracy: 0.7788\n",
      "Epoch 59/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2408 - accuracy: 0.7492 - val_loss: 1.1370 - val_accuracy: 0.7943\n",
      "Epoch 60/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2187 - accuracy: 0.7615 - val_loss: 1.1398 - val_accuracy: 0.7907\n",
      "Epoch 61/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2247 - accuracy: 0.7526 - val_loss: 1.1583 - val_accuracy: 0.7843\n",
      "Epoch 62/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2223 - accuracy: 0.7565 - val_loss: 1.1474 - val_accuracy: 0.7962\n",
      "Epoch 63/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.2150 - accuracy: 0.7570 - val_loss: 1.1484 - val_accuracy: 0.7916\n",
      "Epoch 64/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1979 - accuracy: 0.7501 - val_loss: 1.1389 - val_accuracy: 0.7879\n",
      "Epoch 65/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1817 - accuracy: 0.7641 - val_loss: 1.1155 - val_accuracy: 0.7806\n",
      "Epoch 66/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1877 - accuracy: 0.7666 - val_loss: 1.0962 - val_accuracy: 0.7934\n",
      "Epoch 67/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1602 - accuracy: 0.7741 - val_loss: 1.1318 - val_accuracy: 0.7879\n",
      "Epoch 68/100\n",
      "22/22 [==============================] - 0s 11ms/step - loss: 1.1548 - accuracy: 0.7714 - val_loss: 1.1142 - val_accuracy: 0.7925\n",
      "Epoch 69/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1535 - accuracy: 0.7734 - val_loss: 1.1300 - val_accuracy: 0.7815\n",
      "Epoch 70/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1460 - accuracy: 0.7794 - val_loss: 1.1068 - val_accuracy: 0.7879\n",
      "Epoch 71/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1429 - accuracy: 0.7746 - val_loss: 1.1092 - val_accuracy: 0.7888\n",
      "Epoch 72/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1283 - accuracy: 0.7775 - val_loss: 1.0840 - val_accuracy: 0.8016\n",
      "Epoch 73/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1373 - accuracy: 0.7775 - val_loss: 1.0884 - val_accuracy: 0.7916\n",
      "Epoch 74/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1427 - accuracy: 0.7775 - val_loss: 1.0947 - val_accuracy: 0.7916\n",
      "Epoch 75/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1302 - accuracy: 0.7780 - val_loss: 1.0541 - val_accuracy: 0.8044\n",
      "Epoch 76/100\n",
      "22/22 [==============================] - 0s 9ms/step - loss: 1.1284 - accuracy: 0.7780 - val_loss: 1.0950 - val_accuracy: 0.7943\n",
      "Epoch 77/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1190 - accuracy: 0.7892 - val_loss: 1.0686 - val_accuracy: 0.8007\n",
      "Epoch 78/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1279 - accuracy: 0.7858 - val_loss: 1.0752 - val_accuracy: 0.7852\n",
      "Epoch 79/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1046 - accuracy: 0.7926 - val_loss: 1.0908 - val_accuracy: 0.8007\n",
      "Epoch 80/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1132 - accuracy: 0.7846 - val_loss: 1.0879 - val_accuracy: 0.8053\n",
      "Epoch 81/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1333 - accuracy: 0.7759 - val_loss: 1.0679 - val_accuracy: 0.8007\n",
      "Epoch 82/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1225 - accuracy: 0.7862 - val_loss: 1.0594 - val_accuracy: 0.8062\n",
      "Epoch 83/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.1038 - accuracy: 0.7865 - val_loss: 1.0836 - val_accuracy: 0.7989\n",
      "Epoch 84/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0923 - accuracy: 0.7940 - val_loss: 1.0812 - val_accuracy: 0.8026\n",
      "Epoch 85/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0808 - accuracy: 0.7904 - val_loss: 1.0716 - val_accuracy: 0.8062\n",
      "Epoch 86/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0998 - accuracy: 0.7876 - val_loss: 1.0513 - val_accuracy: 0.8099\n",
      "Epoch 87/100\n",
      "22/22 [==============================] - 0s 12ms/step - loss: 1.0835 - accuracy: 0.7933 - val_loss: 1.0445 - val_accuracy: 0.8071\n",
      "Epoch 88/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0796 - accuracy: 0.7892 - val_loss: 1.0865 - val_accuracy: 0.7962\n",
      "Epoch 89/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0673 - accuracy: 0.7986 - val_loss: 1.0590 - val_accuracy: 0.8053\n",
      "Epoch 90/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0927 - accuracy: 0.7945 - val_loss: 1.0559 - val_accuracy: 0.8007\n",
      "Epoch 91/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0721 - accuracy: 0.8048 - val_loss: 1.0582 - val_accuracy: 0.8099\n",
      "Epoch 92/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0834 - accuracy: 0.7913 - val_loss: 1.0472 - val_accuracy: 0.8181\n",
      "Epoch 93/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0788 - accuracy: 0.7965 - val_loss: 1.0542 - val_accuracy: 0.8117\n",
      "Epoch 94/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0390 - accuracy: 0.8171 - val_loss: 1.0678 - val_accuracy: 0.8053\n",
      "Epoch 95/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0552 - accuracy: 0.8018 - val_loss: 1.0562 - val_accuracy: 0.8199\n",
      "Epoch 96/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0445 - accuracy: 0.8082 - val_loss: 1.0827 - val_accuracy: 0.7907\n",
      "Epoch 97/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0559 - accuracy: 0.8064 - val_loss: 1.0392 - val_accuracy: 0.8172\n",
      "Epoch 98/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0366 - accuracy: 0.8022 - val_loss: 1.0578 - val_accuracy: 0.8190\n",
      "Epoch 99/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0294 - accuracy: 0.8137 - val_loss: 1.0511 - val_accuracy: 0.8080\n",
      "Epoch 100/100\n",
      "22/22 [==============================] - 0s 10ms/step - loss: 1.0429 - accuracy: 0.8093 - val_loss: 1.0580 - val_accuracy: 0.8172\n",
      "CPU times: total: 33.9 s\n",
      "Wall time: 25.3 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "batch_size = 200\n",
    "epochs = 100\n",
    "\n",
    "history = model.fit(\n",
    "  X_train, y_train,\n",
    "  batch_size=batch_size,\n",
    "  epochs=epochs,\n",
    "  validation_split=0.2,\n",
    "  shuffle=True,\n",
    "  # callbacks=[model_checkpoint_callback, lrate_scheduler_callback]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train set accuracy: 0.809\n",
      "train loss: 1.043\n",
      "validation set accuracy: 0.817\n",
      "validation set loss: 1.058\n"
     ]
    }
   ],
   "source": [
    "length = len(history.history['accuracy'])\n",
    "print(\"train set accuracy:\", round(history.history['accuracy'][length-1], 3))\n",
    "print(\"train loss:\", round(history.history['loss'][length-1], 3))\n",
    "print(\"validation set accuracy:\", round(history.history['val_accuracy'][length-1], 3))\n",
    "print(\"validation set loss:\", round(history.history['val_loss'][length-1], 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "43/43 [==============================] - 0s 4ms/step - loss: 1.0520 - accuracy: 0.8149\n",
      "test set accuracy: 0.815\n",
      "test set loss: 1.052\n"
     ]
    }
   ],
   "source": [
    "results = model.evaluate(X_test, y_test)\n",
    "print(\"test set accuracy:\", round(results[1], 3))\n",
    "print(\"test set loss:\", round(results[0], 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'val'], loc='upper left')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "env_torch",
   "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}