homework/Day078_HW.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Work\n",
    "1. 請嘗試將 preproc_x 替換成以每筆資料的 min/max 進行標準化至 -1 ~ 1 間，再進行訓練\n",
    "2. 請嘗試將 mlp 疊更深 (e.g 5~10 層)，進行訓練後觀察 learning curve 的走勢\n",
    "3. (optional) 請改用 GPU 進行訓練 (如果你有 GPU 的話)，比較使用 CPU 與 GPU 的訓練速度"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\n確認硬體資源\\n!nvidia-smi\\n'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## \n",
    "\"\"\"\n",
    "確認硬體資源\n",
    "!nvidia-smi\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import keras\n",
    "\n",
    "# 請嘗試設定 GPU：os.environ\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = keras.datasets.cifar10.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 資料前處理\n",
    "def preproc_x(x):\n",
    "    x=2*(x-x.min())/(x.max()-x.min())-1\n",
    "    x = x.reshape((len(x), -1))\n",
    "    return x\n",
    "\n",
    "def preproc_y(y, num_classes=10):\n",
    "    if y.shape[-1] == 1:\n",
    "        y = keras.utils.to_categorical(y, num_classes)\n",
    "    return y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, y_train = train\n",
    "x_test, y_test = test\n",
    "\n",
    "# Preproc the inputs\n",
    "x_train = preproc_x(x_train)\n",
    "x_test = preproc_x(x_test)\n",
    "\n",
    "# Preprc the outputs\n",
    "y_train = preproc_y(y_train)\n",
    "y_test = preproc_y(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 3072)              0         \n",
      "_________________________________________________________________\n",
      "hidden_layer1 (Dense)        (None, 512)               1573376   \n",
      "_________________________________________________________________\n",
      "hidden_layer2 (Dense)        (None, 256)               131328    \n",
      "_________________________________________________________________\n",
      "hidden_layer3 (Dense)        (None, 128)               32896     \n",
      "_________________________________________________________________\n",
      "hidden_layer4 (Dense)        (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "hidden_layer5 (Dense)        (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "output (Dense)               (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 1,748,266\n",
      "Trainable params: 1,748,266\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "def build_mlp(input_shape, output_units=10, num_neurons=[512, 256, 128,64,32]):\n",
    "    input_layer = keras.layers.Input(input_shape)\n",
    "    \n",
    "    for i, n_units in enumerate(num_neurons):\n",
    "        if i == 0:\n",
    "            x = keras.layers.Dense(units=n_units, activation=\"relu\", name=\"hidden_layer\"+str(i+1))(input_layer)\n",
    "        else:\n",
    "            x = keras.layers.Dense(units=n_units, activation=\"relu\", name=\"hidden_layer\"+str(i+1))(x)\n",
    "    \n",
    "    out = keras.layers.Dense(units=output_units, activation=\"softmax\", name=\"output\")(x)\n",
    "    \n",
    "    model = keras.models.Model(inputs=[input_layer], outputs=[out])\n",
    "    return model\n",
    "model = build_mlp(input_shape=x_train.shape[1:])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 超參數設定\n",
    "LEARNING_RATE = 0.001\n",
    "EPOCHS = 100\n",
    "BATCH_SIZE = 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = keras.optimizers.Adam(lr=LEARNING_RATE)\n",
    "model.compile(loss=\"categorical_crossentropy\", metrics=[\"accuracy\"], optimizer=optimizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "50000/50000 [==============================] - 32s 632us/step - loss: 2.2603 - accuracy: 0.1378 - val_loss: 2.2354 - val_accuracy: 0.1605\n",
      "Epoch 2/100\n",
      "50000/50000 [==============================] - 21s 422us/step - loss: 2.1875 - accuracy: 0.1835 - val_loss: 2.1564 - val_accuracy: 0.2076\n",
      "Epoch 3/100\n",
      "50000/50000 [==============================] - 20s 395us/step - loss: 2.1116 - accuracy: 0.2258 - val_loss: 2.1041 - val_accuracy: 0.2260\n",
      "Epoch 4/100\n",
      "50000/50000 [==============================] - 19s 384us/step - loss: 2.0513 - accuracy: 0.2527 - val_loss: 2.0450 - val_accuracy: 0.2528\n",
      "Epoch 5/100\n",
      "50000/50000 [==============================] - 19s 383us/step - loss: 1.9950 - accuracy: 0.2768 - val_loss: 2.0287 - val_accuracy: 0.2599\n",
      "Epoch 6/100\n",
      "50000/50000 [==============================] - 20s 395us/step - loss: 1.9613 - accuracy: 0.2895 - val_loss: 1.9916 - val_accuracy: 0.2752\n",
      "Epoch 7/100\n",
      "50000/50000 [==============================] - 20s 402us/step - loss: 1.9268 - accuracy: 0.3039 - val_loss: 1.9685 - val_accuracy: 0.2915\n",
      "Epoch 8/100\n",
      "50000/50000 [==============================] - 19s 382us/step - loss: 1.8948 - accuracy: 0.3197 - val_loss: 1.9852 - val_accuracy: 0.2818\n",
      "Epoch 9/100\n",
      "50000/50000 [==============================] - 21s 429us/step - loss: 1.8694 - accuracy: 0.3280 - val_loss: 1.9755 - val_accuracy: 0.2933\n",
      "Epoch 10/100\n",
      "50000/50000 [==============================] - 19s 379us/step - loss: 1.8348 - accuracy: 0.3409 - val_loss: 1.9680 - val_accuracy: 0.2959\n",
      "Epoch 11/100\n",
      "50000/50000 [==============================] - 19s 386us/step - loss: 1.8015 - accuracy: 0.3532 - val_loss: 1.9808 - val_accuracy: 0.2942\n",
      "Epoch 12/100\n",
      "50000/50000 [==============================] - 20s 391us/step - loss: 1.7732 - accuracy: 0.3636 - val_loss: 1.9656 - val_accuracy: 0.2981\n",
      "Epoch 13/100\n",
      "50000/50000 [==============================] - 22s 446us/step - loss: 1.7391 - accuracy: 0.3762 - val_loss: 1.9731 - val_accuracy: 0.3023\n",
      "Epoch 14/100\n",
      "50000/50000 [==============================] - 20s 391us/step - loss: 1.7152 - accuracy: 0.3862 - val_loss: 2.0132 - val_accuracy: 0.2926\n",
      "Epoch 15/100\n",
      "50000/50000 [==============================] - 19s 378us/step - loss: 1.6869 - accuracy: 0.3968 - val_loss: 1.9772 - val_accuracy: 0.3038\n",
      "Epoch 16/100\n",
      "50000/50000 [==============================] - 19s 387us/step - loss: 1.6451 - accuracy: 0.4135 - val_loss: 2.0462 - val_accuracy: 0.2902\n",
      "Epoch 17/100\n",
      "50000/50000 [==============================] - 20s 392us/step - loss: 1.6145 - accuracy: 0.4221 - val_loss: 1.9744 - val_accuracy: 0.3092\n",
      "Epoch 18/100\n",
      "50000/50000 [==============================] - 20s 397us/step - loss: 1.5868 - accuracy: 0.4341 - val_loss: 2.0116 - val_accuracy: 0.3136\n",
      "Epoch 19/100\n",
      "50000/50000 [==============================] - 20s 398us/step - loss: 1.5479 - accuracy: 0.4491 - val_loss: 2.0435 - val_accuracy: 0.3066\n",
      "Epoch 20/100\n",
      "50000/50000 [==============================] - 22s 444us/step - loss: 1.5149 - accuracy: 0.4594 - val_loss: 2.0448 - val_accuracy: 0.3096\n",
      "Epoch 21/100\n",
      "50000/50000 [==============================] - 20s 397us/step - loss: 1.4834 - accuracy: 0.4720 - val_loss: 2.0426 - val_accuracy: 0.2985\n",
      "Epoch 22/100\n",
      "50000/50000 [==============================] - 24s 470us/step - loss: 1.4599 - accuracy: 0.4804 - val_loss: 2.1561 - val_accuracy: 0.3004\n",
      "Epoch 23/100\n",
      "50000/50000 [==============================] - 20s 398us/step - loss: 1.4079 - accuracy: 0.4972 - val_loss: 2.1213 - val_accuracy: 0.3017\n",
      "Epoch 24/100\n",
      "50000/50000 [==============================] - 22s 431us/step - loss: 1.3858 - accuracy: 0.5084 - val_loss: 2.1683 - val_accuracy: 0.3052\n",
      "Epoch 25/100\n",
      "50000/50000 [==============================] - 21s 423us/step - loss: 1.3480 - accuracy: 0.5220 - val_loss: 2.2136 - val_accuracy: 0.2992\n",
      "Epoch 26/100\n",
      "50000/50000 [==============================] - 20s 391us/step - loss: 1.3145 - accuracy: 0.5309 - val_loss: 2.1927 - val_accuracy: 0.3012\n",
      "Epoch 27/100\n",
      "50000/50000 [==============================] - 20s 394us/step - loss: 1.2688 - accuracy: 0.5509 - val_loss: 2.2926 - val_accuracy: 0.2961\n",
      "Epoch 28/100\n",
      "50000/50000 [==============================] - 19s 388us/step - loss: 1.2443 - accuracy: 0.5563 - val_loss: 2.3513 - val_accuracy: 0.2952\n",
      "Epoch 29/100\n",
      "50000/50000 [==============================] - 21s 414us/step - loss: 1.2283 - accuracy: 0.5617 - val_loss: 2.3645 - val_accuracy: 0.2922\n",
      "Epoch 30/100\n",
      "50000/50000 [==============================] - 20s 390us/step - loss: 1.1740 - accuracy: 0.5842 - val_loss: 2.4756 - val_accuracy: 0.2849\n",
      "Epoch 31/100\n",
      "50000/50000 [==============================] - 21s 412us/step - loss: 1.1593 - accuracy: 0.5891 - val_loss: 2.5159 - val_accuracy: 0.2949\n",
      "Epoch 32/100\n",
      "50000/50000 [==============================] - 20s 398us/step - loss: 1.1172 - accuracy: 0.6028 - val_loss: 2.4948 - val_accuracy: 0.2994\n",
      "Epoch 33/100\n",
      "50000/50000 [==============================] - 22s 443us/step - loss: 1.0951 - accuracy: 0.6096 - val_loss: 2.6217 - val_accuracy: 0.2998\n",
      "Epoch 34/100\n",
      "50000/50000 [==============================] - 20s 407us/step - loss: 1.0641 - accuracy: 0.6209 - val_loss: 2.6137 - val_accuracy: 0.2978\n",
      "Epoch 35/100\n",
      "50000/50000 [==============================] - 20s 401us/step - loss: 1.0329 - accuracy: 0.6322 - val_loss: 2.7162 - val_accuracy: 0.2862\n",
      "Epoch 36/100\n",
      "50000/50000 [==============================] - 27s 531us/step - loss: 0.9929 - accuracy: 0.6481 - val_loss: 2.8126 - val_accuracy: 0.2806\n",
      "Epoch 37/100\n",
      "50000/50000 [==============================] - 20s 407us/step - loss: 0.9574 - accuracy: 0.6628 - val_loss: 2.8975 - val_accuracy: 0.2971\n",
      "Epoch 38/100\n",
      "50000/50000 [==============================] - 23s 468us/step - loss: 0.9219 - accuracy: 0.6761 - val_loss: 2.8267 - val_accuracy: 0.2895\n",
      "Epoch 39/100\n",
      "50000/50000 [==============================] - 20s 394us/step - loss: 0.9172 - accuracy: 0.6765 - val_loss: 2.9165 - val_accuracy: 0.2949\n",
      "Epoch 40/100\n",
      "50000/50000 [==============================] - 19s 377us/step - loss: 0.9017 - accuracy: 0.6795 - val_loss: 3.0014 - val_accuracy: 0.2948\n",
      "Epoch 41/100\n",
      "50000/50000 [==============================] - 19s 377us/step - loss: 0.8630 - accuracy: 0.6932 - val_loss: 3.0028 - val_accuracy: 0.2840\n",
      "Epoch 42/100\n",
      "50000/50000 [==============================] - 19s 389us/step - loss: 0.8264 - accuracy: 0.7089 - val_loss: 3.2776 - val_accuracy: 0.2893\n",
      "Epoch 43/100\n",
      "50000/50000 [==============================] - 19s 385us/step - loss: 0.8094 - accuracy: 0.7116 - val_loss: 3.2381 - val_accuracy: 0.2857\n",
      "Epoch 44/100\n",
      "50000/50000 [==============================] - 19s 380us/step - loss: 0.7926 - accuracy: 0.7172 - val_loss: 3.3719 - val_accuracy: 0.2936\n",
      "Epoch 45/100\n",
      "50000/50000 [==============================] - 24s 476us/step - loss: 0.7579 - accuracy: 0.7304 - val_loss: 3.5700 - val_accuracy: 0.2824\n",
      "Epoch 46/100\n",
      "50000/50000 [==============================] - 24s 473us/step - loss: 0.7484 - accuracy: 0.7336 - val_loss: 3.5302 - val_accuracy: 0.2922\n",
      "Epoch 47/100\n",
      "50000/50000 [==============================] - 19s 388us/step - loss: 0.7150 - accuracy: 0.7453 - val_loss: 3.5741 - val_accuracy: 0.2874\n",
      "Epoch 48/100\n",
      "50000/50000 [==============================] - 23s 465us/step - loss: 0.6954 - accuracy: 0.7541 - val_loss: 3.7676 - val_accuracy: 0.2822\n",
      "Epoch 49/100\n",
      "50000/50000 [==============================] - 21s 422us/step - loss: 0.6803 - accuracy: 0.7601 - val_loss: 3.6623 - val_accuracy: 0.2858\n",
      "Epoch 50/100\n",
      "50000/50000 [==============================] - 22s 446us/step - loss: 0.6495 - accuracy: 0.7701 - val_loss: 3.9352 - val_accuracy: 0.2793\n",
      "Epoch 51/100\n",
      "50000/50000 [==============================] - 28s 562us/step - loss: 0.6514 - accuracy: 0.7679 - val_loss: 4.0367 - val_accuracy: 0.2840\n",
      "Epoch 52/100\n",
      "50000/50000 [==============================] - 24s 490us/step - loss: 0.6103 - accuracy: 0.7849 - val_loss: 3.8969 - val_accuracy: 0.2790\n",
      "Epoch 53/100\n",
      "50000/50000 [==============================] - 20s 405us/step - loss: 0.6006 - accuracy: 0.7869 - val_loss: 4.0248 - val_accuracy: 0.2820\n",
      "Epoch 54/100\n",
      "50000/50000 [==============================] - 19s 390us/step - loss: 0.6093 - accuracy: 0.7853 - val_loss: 4.0185 - val_accuracy: 0.2915\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 55/100\n",
      "50000/50000 [==============================] - 19s 373us/step - loss: 0.5445 - accuracy: 0.8092 - val_loss: 4.3085 - val_accuracy: 0.2865\n",
      "Epoch 56/100\n",
      "50000/50000 [==============================] - 20s 390us/step - loss: 0.5509 - accuracy: 0.8066 - val_loss: 4.3677 - val_accuracy: 0.2836\n",
      "Epoch 57/100\n",
      "50000/50000 [==============================] - 21s 429us/step - loss: 0.5322 - accuracy: 0.8111 - val_loss: 4.3836 - val_accuracy: 0.2791\n",
      "Epoch 58/100\n",
      "50000/50000 [==============================] - 19s 373us/step - loss: 0.5228 - accuracy: 0.8154 - val_loss: 4.4515 - val_accuracy: 0.2769\n",
      "Epoch 59/100\n",
      "50000/50000 [==============================] - 18s 370us/step - loss: 0.4988 - accuracy: 0.8246 - val_loss: 4.7665 - val_accuracy: 0.2696\n",
      "Epoch 60/100\n",
      "50000/50000 [==============================] - 22s 440us/step - loss: 0.5186 - accuracy: 0.8157 - val_loss: 5.0576 - val_accuracy: 0.2764\n",
      "Epoch 61/100\n",
      "50000/50000 [==============================] - 21s 414us/step - loss: 0.4768 - accuracy: 0.8317 - val_loss: 4.9802 - val_accuracy: 0.2767\n",
      "Epoch 62/100\n",
      "50000/50000 [==============================] - 19s 375us/step - loss: 0.4481 - accuracy: 0.8444 - val_loss: 4.9259 - val_accuracy: 0.2748\n",
      "Epoch 63/100\n",
      "50000/50000 [==============================] - 19s 375us/step - loss: 0.4234 - accuracy: 0.8520 - val_loss: 5.4138 - val_accuracy: 0.2900\n",
      "Epoch 64/100\n",
      "50000/50000 [==============================] - 19s 387us/step - loss: 0.4156 - accuracy: 0.8556 - val_loss: 5.3770 - val_accuracy: 0.2829\n",
      "Epoch 65/100\n",
      "50000/50000 [==============================] - 20s 395us/step - loss: 0.4978 - accuracy: 0.8270 - val_loss: 5.0235 - val_accuracy: 0.2745\n",
      "Epoch 66/100\n",
      "50000/50000 [==============================] - 19s 386us/step - loss: 0.4132 - accuracy: 0.8565 - val_loss: 5.1637 - val_accuracy: 0.2743\n",
      "Epoch 67/100\n",
      "50000/50000 [==============================] - 22s 438us/step - loss: 0.4184 - accuracy: 0.8538 - val_loss: 5.3890 - val_accuracy: 0.2804\n",
      "Epoch 68/100\n",
      "50000/50000 [==============================] - 23s 468us/step - loss: 0.3917 - accuracy: 0.8657 - val_loss: 5.5496 - val_accuracy: 0.2819\n",
      "Epoch 69/100\n",
      "50000/50000 [==============================] - 20s 398us/step - loss: 0.4213 - accuracy: 0.8533 - val_loss: 5.4384 - val_accuracy: 0.2739\n",
      "Epoch 70/100\n",
      "50000/50000 [==============================] - 20s 398us/step - loss: 0.4146 - accuracy: 0.8545 - val_loss: 5.7483 - val_accuracy: 0.2701\n",
      "Epoch 71/100\n",
      "50000/50000 [==============================] - 19s 388us/step - loss: 0.4238 - accuracy: 0.8538 - val_loss: 5.7489 - val_accuracy: 0.2766\n",
      "Epoch 72/100\n",
      "50000/50000 [==============================] - 19s 390us/step - loss: 0.3283 - accuracy: 0.8880 - val_loss: 5.7366 - val_accuracy: 0.2680\n",
      "Epoch 73/100\n",
      "50000/50000 [==============================] - 19s 388us/step - loss: 0.3622 - accuracy: 0.8746 - val_loss: 5.8393 - val_accuracy: 0.2845\n",
      "Epoch 74/100\n",
      "50000/50000 [==============================] - 20s 403us/step - loss: 0.3465 - accuracy: 0.8806 - val_loss: 5.9495 - val_accuracy: 0.2732\n",
      "Epoch 75/100\n",
      "50000/50000 [==============================] - 20s 395us/step - loss: 0.3714 - accuracy: 0.8712 - val_loss: 5.7853 - val_accuracy: 0.2737\n",
      "Epoch 76/100\n",
      "50000/50000 [==============================] - 19s 382us/step - loss: 0.3086 - accuracy: 0.8943 - val_loss: 6.0755 - val_accuracy: 0.2794\n",
      "Epoch 77/100\n",
      "50000/50000 [==============================] - 20s 397us/step - loss: 0.3185 - accuracy: 0.8903 - val_loss: 6.6472 - val_accuracy: 0.2733\n",
      "Epoch 78/100\n",
      "50000/50000 [==============================] - 19s 388us/step - loss: 0.3639 - accuracy: 0.8750 - val_loss: 6.1828 - val_accuracy: 0.2687\n",
      "Epoch 79/100\n",
      "50000/50000 [==============================] - 21s 412us/step - loss: 0.3213 - accuracy: 0.8889 - val_loss: 6.4898 - val_accuracy: 0.2750\n",
      "Epoch 80/100\n",
      "50000/50000 [==============================] - 21s 427us/step - loss: 0.3214 - accuracy: 0.8893 - val_loss: 6.5009 - val_accuracy: 0.2772\n",
      "Epoch 81/100\n",
      "50000/50000 [==============================] - 24s 490us/step - loss: 0.3242 - accuracy: 0.8872 - val_loss: 6.5164 - val_accuracy: 0.2709\n",
      "Epoch 82/100\n",
      "50000/50000 [==============================] - 22s 449us/step - loss: 0.3256 - accuracy: 0.8866 - val_loss: 6.9900 - val_accuracy: 0.2691\n",
      "Epoch 83/100\n",
      "50000/50000 [==============================] - 20s 406us/step - loss: 0.3505 - accuracy: 0.8789 - val_loss: 6.4923 - val_accuracy: 0.2700\n",
      "Epoch 84/100\n",
      "50000/50000 [==============================] - 20s 401us/step - loss: 0.2952 - accuracy: 0.8984 - val_loss: 6.6135 - val_accuracy: 0.2717\n",
      "Epoch 85/100\n",
      "50000/50000 [==============================] - 26s 522us/step - loss: 0.2559 - accuracy: 0.9122 - val_loss: 6.9058 - val_accuracy: 0.2786\n",
      "Epoch 86/100\n",
      "50000/50000 [==============================] - 20s 410us/step - loss: 0.2408 - accuracy: 0.9184 - val_loss: 7.2265 - val_accuracy: 0.2781\n",
      "Epoch 87/100\n",
      "50000/50000 [==============================] - 20s 402us/step - loss: 0.3377 - accuracy: 0.8831 - val_loss: 6.5418 - val_accuracy: 0.2795\n",
      "Epoch 88/100\n",
      "50000/50000 [==============================] - 21s 412us/step - loss: 0.2609 - accuracy: 0.9122 - val_loss: 7.1450 - val_accuracy: 0.2786\n",
      "Epoch 89/100\n",
      "50000/50000 [==============================] - 21s 418us/step - loss: 0.2435 - accuracy: 0.9169 - val_loss: 7.0414 - val_accuracy: 0.2767\n",
      "Epoch 90/100\n",
      "50000/50000 [==============================] - 21s 427us/step - loss: 0.3932 - accuracy: 0.8676 - val_loss: 6.3446 - val_accuracy: 0.2733\n",
      "Epoch 91/100\n",
      "50000/50000 [==============================] - 20s 402us/step - loss: 0.2678 - accuracy: 0.9100 - val_loss: 6.8267 - val_accuracy: 0.2764\n",
      "Epoch 92/100\n",
      "50000/50000 [==============================] - 20s 408us/step - loss: 0.2179 - accuracy: 0.9259 - val_loss: 7.3458 - val_accuracy: 0.2864\n",
      "Epoch 93/100\n",
      "50000/50000 [==============================] - 22s 448us/step - loss: 0.2858 - accuracy: 0.9029 - val_loss: 6.7977 - val_accuracy: 0.2797\n",
      "Epoch 94/100\n",
      "50000/50000 [==============================] - 22s 443us/step - loss: 0.2356 - accuracy: 0.9212 - val_loss: 7.0896 - val_accuracy: 0.2739\n",
      "Epoch 95/100\n",
      "50000/50000 [==============================] - 21s 418us/step - loss: 0.2500 - accuracy: 0.9148 - val_loss: 7.0874 - val_accuracy: 0.2765\n",
      "Epoch 96/100\n",
      "50000/50000 [==============================] - 22s 436us/step - loss: 0.2552 - accuracy: 0.9130 - val_loss: 7.3355 - val_accuracy: 0.2786\n",
      "Epoch 97/100\n",
      "50000/50000 [==============================] - 20s 406us/step - loss: 0.2819 - accuracy: 0.9030 - val_loss: 7.1103 - val_accuracy: 0.2715\n",
      "Epoch 98/100\n",
      "50000/50000 [==============================] - 21s 410us/step - loss: 0.3065 - accuracy: 0.8975 - val_loss: 7.1671 - val_accuracy: 0.2748\n",
      "Epoch 99/100\n",
      "50000/50000 [==============================] - 20s 400us/step - loss: 0.1748 - accuracy: 0.9412 - val_loss: 7.6615 - val_accuracy: 0.2742\n",
      "Epoch 100/100\n",
      "50000/50000 [==============================] - 20s 404us/step - loss: 0.2049 - accuracy: 0.9304 - val_loss: 7.3047 - val_accuracy: 0.2749\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.callbacks.History at 0x1443e0160>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(x_train, y_train, \n",
    "          epochs=EPOCHS, \n",
    "          batch_size=BATCH_SIZE, \n",
    "          validation_data=(x_test, y_test), \n",
    "          shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "train_loss = model.history.history[\"loss\"]\n",
    "valid_loss = model.history.history[\"val_loss\"]\n",
    "\n",
    "train_acc = model.history.history[\"accuracy\"]\n",
    "valid_acc = model.history.history[\"val_accuracy\"]\n",
    "\n",
    "plt.plot(range(len(train_loss)), train_loss, label=\"train loss\")\n",
    "plt.plot(range(len(valid_loss)), valid_loss, label=\"valid loss\")\n",
    "plt.legend()\n",
    "plt.title(\"Loss\")\n",
    "plt.show()\n",
    "\n",
    "plt.plot(range(len(train_acc)), train_acc, label=\"train accuracy\")\n",
    "plt.plot(range(len(valid_acc)), valid_acc, label=\"valid accuracy\")\n",
    "plt.legend()\n",
    "plt.title(\"Accuracy\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}