Day_076-optimizer_HW.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 作業重點:\n",
    "\n",
    "(1)以, Adam, 為例, 調整 batch_size, epoch , 觀察accurancy, loss 的變化\n",
    "\n",
    "(2)以同一模型, 分別驗證 SGD, Adam, Rmsprop 的 accurancy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 作業目標:\n",
    "    \n",
    "    取得各種優化器的運算結果"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (1)以, Adam, 為例, 調整 batch_size, epoch , 觀察accurancy, loss 的變化"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "import keras\n",
    "from keras.datasets import cifar10\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "import os\n",
    "from keras import optimizers\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Blas GEMM launch failed , 避免動態分配GPU / CPU, 出現問題\n",
    "# import tensorflow as tf\n",
    "# gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)\n",
    "# sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))\n",
    "\n",
    "import tensorflow as tf\n",
    "gpu_options = tf.compat.v1.GPUOptions(per_process_gpu_memory_fraction=0.333) ##每个gpu占用0.333 的显存\n",
    "config=tf.compat.v1.ConfigProto(gpu_options=gpu_options, allow_soft_placement=True)\n",
    "sess=tf.compat.v1.Session(config=config)##如果电脑有多个GPU，tensorflow默认全部使用。如果想只使用部分GPU，可以设置CUDA_VISIBLE_DEVICES。\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "'''\n",
    "   宣告並設定\n",
    "   batch_size：對總的樣本數進行分組，每組包含的樣本數量\n",
    "   epochs ：訓練次數\n",
    "   \n",
    "''' \n",
    "batch_size = 16 #example = 32\n",
    "num_classes = 10\n",
    "epochs = 30 #example = 20\n",
    "data_augmentation = True\n",
    "num_predictions = 20\n",
    "save_dir = os.path.join(os.getcwd(), 'saved_models')\n",
    "model_name = 'keras_cifar10_trained_model_hw.h5'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape: (50000, 32, 32, 3)\n",
      "50000 train samples\n",
      "10000 test samples\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# The data, split between train and test sets:\n",
    "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
    "print('x_train shape:', x_train.shape)\n",
    "print(x_train.shape[0], 'train samples')\n",
    "print(x_test.shape[0], 'test samples')\n",
    "\n",
    "# Convert class vectors to binary class matrices.\n",
    "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#    第一步：選擇模型, 順序模型是多個網絡層的線性堆疊\n",
    " \n",
    "model = Sequential()\n",
    "\n",
    "#   第二步：構建網絡層\n",
    "model.add(Conv2D(32, (3, 3), padding='same',\n",
    "                 input_shape=x_train.shape[1:]))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(32, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Conv2D(64, (3, 3), padding='same'))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(64, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(512))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "model.add(Dense( 10)) # 輸出結果是10個類別，所以維度是10   \n",
    "model.add(Activation('softmax')) # 最後一層用softmax作為激活函數"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Parameters：1250858\n"
     ]
    }
   ],
   "source": [
    "# 模型建立完成後，統計參數總量\n",
    "print(\"Total Parameters：%d\" % model.count_params())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 32, 32, 32)        896       \n",
      "_________________________________________________________________\n",
      "activation (Activation)      (None, 32, 32, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 30, 30, 32)        9248      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 30, 30, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 15, 15, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 15, 15, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 13, 13, 64)        36928     \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 13, 13, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 2304)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 512)               1180160   \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                5130      \n",
      "_________________________________________________________________\n",
      "activation_5 (Activation)    (None, 10)                0         \n",
      "=================================================================\n",
      "Total params: 1,250,858\n",
      "Trainable params: 1,250,858\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 輸出模型摘要資訊\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#第三步編譯\n",
    "'''\n",
    " SGD(隨機梯度下降) - Arguments\n",
    "lr: float >= 0. Learning rate.\n",
    "momentum: float >= 0. Parameter that accelerates SGD in the relevant direction and dampens oscillations.\n",
    "decay: float >= 0. Learning rate decay over each update.\n",
    "nesterov: boolean. Whether to apply Nesterov momentum.\n",
    "'''\n",
    "\n",
    "'''\n",
    "RMSprop- Arguments\n",
    "lr: float >= 0. Learning rate.\n",
    "rho: float >= 0.\n",
    "epsilon: float >= 0. Fuzz factor. If None, defaults to K.epsilon().\n",
    "decay: float >= 0. Learning rate decay over each update.\n",
    "'''\n",
    "\n",
    "'''\n",
    "Example:\n",
    "opt = optimizers.SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "\n",
    "'''\n",
    "model.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 資料正規化\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "x_train /= 255\n",
    "x_test /= 255\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using real-time data augmentation.\n",
      "\n",
      "Epoch 1/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 1.5414 - accuracy: 0.4391 - val_loss: 1.1744 - val_accuracy: 0.5828\n",
      "Epoch 2/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 1.1676 - accuracy: 0.5857 - val_loss: 0.9548 - val_accuracy: 0.6656\n",
      "Epoch 3/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 1.0286 - accuracy: 0.6367 - val_loss: 0.8880 - val_accuracy: 0.6895\n",
      "Epoch 4/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.9569 - accuracy: 0.6635 - val_loss: 0.9434 - val_accuracy: 0.6717\n",
      "Epoch 5/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.8965 - accuracy: 0.6843 - val_loss: 0.8515 - val_accuracy: 0.7096\n",
      "Epoch 6/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.8588 - accuracy: 0.7002 - val_loss: 0.7872 - val_accuracy: 0.7286\n",
      "Epoch 7/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.8216 - accuracy: 0.7116 - val_loss: 0.7926 - val_accuracy: 0.7275\n",
      "Epoch 8/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.8044 - accuracy: 0.7193 - val_loss: 0.7535 - val_accuracy: 0.7394\n",
      "Epoch 9/30\n",
      "3125/3125 [==============================] - 98s 31ms/step - loss: 0.7728 - accuracy: 0.7299 - val_loss: 0.7599 - val_accuracy: 0.7400\n",
      "Epoch 10/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.7594 - accuracy: 0.7344 - val_loss: 0.7318 - val_accuracy: 0.7461\n",
      "Epoch 11/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.7326 - accuracy: 0.7443 - val_loss: 0.7710 - val_accuracy: 0.7390\n",
      "Epoch 12/30\n",
      "3125/3125 [==============================] - 95s 30ms/step - loss: 0.7219 - accuracy: 0.7502 - val_loss: 0.7767 - val_accuracy: 0.7421\n",
      "Epoch 13/30\n",
      "3125/3125 [==============================] - 95s 30ms/step - loss: 0.7067 - accuracy: 0.7550 - val_loss: 0.7487 - val_accuracy: 0.7517\n",
      "Epoch 14/30\n",
      "3125/3125 [==============================] - 95s 30ms/step - loss: 0.6940 - accuracy: 0.7590 - val_loss: 0.7517 - val_accuracy: 0.7437\n",
      "Epoch 15/30\n",
      "3125/3125 [==============================] - 98s 32ms/step - loss: 0.6885 - accuracy: 0.7608 - val_loss: 0.7301 - val_accuracy: 0.7505\n",
      "Epoch 16/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.6785 - accuracy: 0.7653 - val_loss: 0.7131 - val_accuracy: 0.7592\n",
      "Epoch 17/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.6712 - accuracy: 0.7649 - val_loss: 0.7102 - val_accuracy: 0.7530\n",
      "Epoch 18/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.6560 - accuracy: 0.7706 - val_loss: 0.6956 - val_accuracy: 0.7589\n",
      "Epoch 19/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.6533 - accuracy: 0.7737 - val_loss: 0.7268 - val_accuracy: 0.7534\n",
      "Epoch 20/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.6468 - accuracy: 0.7764 - val_loss: 0.7565 - val_accuracy: 0.7513\n",
      "Epoch 21/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.6345 - accuracy: 0.7791 - val_loss: 0.7602 - val_accuracy: 0.7445\n",
      "Epoch 22/30\n",
      "3125/3125 [==============================] - 99s 32ms/step - loss: 0.6323 - accuracy: 0.7824 - val_loss: 0.7356 - val_accuracy: 0.7532\n",
      "Epoch 23/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.6272 - accuracy: 0.7840 - val_loss: 0.6860 - val_accuracy: 0.7676\n",
      "Epoch 24/30\n",
      "3125/3125 [==============================] - 101s 32ms/step - loss: 0.6242 - accuracy: 0.7822 - val_loss: 0.7325 - val_accuracy: 0.7557\n",
      "Epoch 25/30\n",
      "3125/3125 [==============================] - 98s 31ms/step - loss: 0.6083 - accuracy: 0.7902 - val_loss: 0.7064 - val_accuracy: 0.7659\n",
      "Epoch 26/30\n",
      "3125/3125 [==============================] - 96s 31ms/step - loss: 0.6163 - accuracy: 0.7850 - val_loss: 0.7546 - val_accuracy: 0.7588\n",
      "Epoch 27/30\n",
      "3125/3125 [==============================] - 95s 31ms/step - loss: 0.6024 - accuracy: 0.7901 - val_loss: 0.7133 - val_accuracy: 0.7650\n",
      "Epoch 28/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.5961 - accuracy: 0.7940 - val_loss: 0.7044 - val_accuracy: 0.7629\n",
      "Epoch 29/30\n",
      "3125/3125 [==============================] - 100s 32ms/step - loss: 0.5930 - accuracy: 0.7969 - val_loss: 0.7107 - val_accuracy: 0.7675\n",
      "Epoch 30/30\n",
      "3125/3125 [==============================] - 98s 31ms/step - loss: 0.5863 - accuracy: 0.7958 - val_loss: 0.7285 - val_accuracy: 0.7589\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\n   第四步：訓練\\n   .fit的一些參數\\n   batch_size：對總的樣本數進行分組，每組包含的樣本數量\\n   epochs ：訓練次數\\n   shuffle：是否把數據隨機打亂之後再進行訓練\\n   validation_split：拿出百分之多少用來做交叉驗證\\n   verbose：屏顯模式 - 0：不輸出, 1：輸出進度, 2：輸出每次的訓練結果\\n'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 是否要做資料處理\n",
    "if not data_augmentation:\n",
    "    print('Not using data augmentation.')\n",
    "    history=model.fit(x_train, y_train,\n",
    "              batch_size=batch_size,\n",
    "              epochs=epochs,\n",
    "              validation_data=(x_test, y_test),\n",
    "              shuffle=True)\n",
    "else:\n",
    "    print('Using real-time data augmentation.')\n",
    "    print('')\n",
    "        \n",
    "    # This will do preprocessing and realtime data augmentation:\n",
    "    datagen = ImageDataGenerator(\n",
    "        featurewise_center=False,  # set input mean to 0 over the dataset\n",
    "        samplewise_center=False,  # set each sample mean to 0\n",
    "        featurewise_std_normalization=False,  # divide inputs by std of the dataset\n",
    "        samplewise_std_normalization=False,  # divide each input by its std\n",
    "        zca_whitening=False,  # apply ZCA whitening\n",
    "        zca_epsilon=1e-06,  # epsilon for ZCA whitening\n",
    "        rotation_range=0,  # randomly rotate images in the range (degrees, 0 to 180)\n",
    "        # randomly shift images horizontally (fraction of total width)\n",
    "        width_shift_range=0.1,\n",
    "        # randomly shift images vertically (fraction of total height)\n",
    "        height_shift_range=0.1,\n",
    "        shear_range=0.,  # set range for random shear\n",
    "        zoom_range=0.,  # set range for random zoom\n",
    "        channel_shift_range=0.,  # set range for random channel shifts\n",
    "        # set mode for filling points outside the input boundaries\n",
    "        fill_mode='nearest',\n",
    "        cval=0.,  # value used for fill_mode = \"constant\"\n",
    "        horizontal_flip=True,  # randomly flip images\n",
    "        vertical_flip=False,  # randomly flip images\n",
    "        # set rescaling factor (applied before any other transformation)\n",
    "        rescale=None,\n",
    "        # set function that will be applied on each input\n",
    "        preprocessing_function=None,\n",
    "        # image data format, either \"channels_first\" or \"channels_last\"\n",
    "        data_format=None,\n",
    "        # fraction of images reserved for validation (strictly between 0 and 1)\n",
    "        validation_split=0.0)\n",
    "\n",
    "    # Compute quantities required for feature-wise normalization\n",
    "    # (std, mean, and principal components if ZCA whitening is applied).\n",
    "    datagen.fit(x_train)\n",
    "    history=model.fit(x_train, y_train,\n",
    "              batch_size=batch_size,\n",
    "              epochs=epochs,\n",
    "              validation_data=(x_test, y_test),\n",
    "              shuffle=True)   \n",
    "\n",
    "'''\n",
    "   第四步：訓練\n",
    "   .fit的一些參數\n",
    "   batch_size：對總的樣本數進行分組，每組包含的樣本數量\n",
    "   epochs ：訓練次數\n",
    "   shuffle：是否把數據隨機打亂之後再進行訓練\n",
    "   validation_split：拿出百分之多少用來做交叉驗證\n",
    "   verbose：屏顯模式 - 0：不輸出, 1：輸出進度, 2：輸出每次的訓練結果\n",
    "''' \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saved trained model at C:\\Users\\cerbe\\Desktop\\ML_marathon\\saved_models\\keras_cifar10_trained_model_hw.h5 \n",
      "313/313 [==============================] - 2s 6ms/step - loss: 0.7285 - accuracy: 0.7589\n",
      "Test loss: 0.7285057902336121\n",
      "Test accuracy: 0.758899986743927\n"
     ]
    }
   ],
   "source": [
    "# Save model and weights\n",
    "if not os.path.isdir(save_dir):\n",
    "    os.makedirs(save_dir)\n",
    "model_path = os.path.join(save_dir, model_name)\n",
    "model.save(model_path)\n",
    "print('Saved trained model at %s ' % model_path)\n",
    "\n",
    "# Score trained model.\n",
    "scores = model.evaluate(x_test, y_test, verbose=1)\n",
    "print('Test loss:', scores[0])\n",
    "print('Test accuracy:', scores[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " test set \n",
      "\n",
      " The test loss is %f  [0.7285060286521912, 0.758899986743927]\n"
     ]
    }
   ],
   "source": [
    "#    第六步：輸出\n",
    "import numpy \n",
    "\n",
    "print ( \" test set \" )\n",
    "scores = model.evaluate(x_test,y_test,batch_size=200,verbose= 0)\n",
    "print ( \"\" )\n",
    "#print ( \" The test loss is %f \" % scores)\n",
    "print ( \" The test loss is %f \", scores)\n",
    "\n",
    "\n",
    "result = model.predict(x_test,batch_size=200,verbose= 0)\n",
    "\n",
    "result_max = numpy.argmax(result, axis = 1 )\n",
    "test_max = numpy.argmax(y_test, axis = 1 )\n",
    "\n",
    "result_bool = numpy.equal(result_max, test_max)\n",
    "true_num = numpy.sum(result_bool)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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",
    "%matplotlib inline\n",
    "\n",
    "# Plot training & validation accuracy values\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', 'Valiidation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Valiidation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'縮小批次，增加epochs並沒有說一定可以把模型做到更好的優化，比原始的成果差了一些'"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#批次大小縮減，epochs增加來檢視對於adam優化的影響程度\n",
    "# batch_size = 16 #example = 32\n",
    "# num_classes = 10\n",
    "# epochs = 30 #example = 20\n",
    "'''縮小批次，增加epochs並沒有說一定可以把模型做到更好的優化，比原始的成果差了一些'''"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (2)以同一模型, 分別驗證 SGD, Adam, Rmsprop 的 accurancy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape: (50000, 32, 32, 3)\n",
      "50000 train samples\n",
      "10000 test samples\n"
     ]
    }
   ],
   "source": [
    "batch_size = 32 #example = 32\n",
    "num_classes = 10\n",
    "epochs = 20 \n",
    "data_augmentation = True\n",
    "num_predictions = 20\n",
    "save_dir = os.path.join(os.getcwd(), 'saved_models')\n",
    "model_name = 'keras_cifar10_trained_model_hw.h5'\n",
    "\n",
    "\n",
    "# The data, split between train and test sets:\n",
    "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
    "print('x_train shape:', x_train.shape)\n",
    "print(x_train.shape[0], 'train samples')\n",
    "print(x_test.shape[0], 'test samples')\n",
    "\n",
    "# Convert class vectors to binary class matrices.\n",
    "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Parameters：1250858\n",
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_8 (Conv2D)            (None, 32, 32, 32)        896       \n",
      "_________________________________________________________________\n",
      "activation_12 (Activation)   (None, 32, 32, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_9 (Conv2D)            (None, 30, 30, 32)        9248      \n",
      "_________________________________________________________________\n",
      "activation_13 (Activation)   (None, 30, 30, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_10 (Conv2D)           (None, 15, 15, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_14 (Activation)   (None, 15, 15, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_11 (Conv2D)           (None, 13, 13, 64)        36928     \n",
      "_________________________________________________________________\n",
      "activation_15 (Activation)   (None, 13, 13, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "dropout_7 (Dropout)          (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 2304)              0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 512)               1180160   \n",
      "_________________________________________________________________\n",
      "activation_16 (Activation)   (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout_8 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 10)                5130      \n",
      "_________________________________________________________________\n",
      "activation_17 (Activation)   (None, 10)                0         \n",
      "=================================================================\n",
      "Total params: 1,250,858\n",
      "Trainable params: 1,250,858\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#    第一步：選擇模型, 順序模型是多個網絡層的線性堆疊\n",
    " \n",
    "model = Sequential()\n",
    "\n",
    "#   第二步：構建網絡層\n",
    "model.add(Conv2D(32, (3, 3), padding='same',\n",
    "                 input_shape=x_train.shape[1:]))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(32, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Conv2D(64, (3, 3), padding='same'))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(64, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(512))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "model.add(Dense( 10)) # 輸出結果是10個類別，所以維度是10   \n",
    "model.add(Activation('softmax')) # 最後一層用softmax作為激活函數\n",
    "\n",
    "# 模型建立完成後，統計參數總量\n",
    "print(\"Total Parameters：%d\" % model.count_params())\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "#第三步編譯\n",
    "'''\n",
    " SGD(隨機梯度下降) - Arguments\n",
    "lr: float >= 0. Learning rate.\n",
    "momentum: float >= 0. Parameter that accelerates SGD in the relevant direction and dampens oscillations.\n",
    "decay: float >= 0. Learning rate decay over each update.\n",
    "nesterov: boolean. Whether to apply Nesterov momentum.\n",
    "'''\n",
    "\n",
    "'''\n",
    "RMSprop- Arguments\n",
    "lr: float >= 0. Learning rate.\n",
    "rho: float >= 0.\n",
    "epsilon: float >= 0. Fuzz factor. If None, defaults to K.epsilon().\n",
    "decay: float >= 0. Learning rate decay over each update.\n",
    "'''\n",
    "\n",
    "'''\n",
    "Example:\n",
    "opt = optimizers.SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "\n",
    "'''\n",
    "opt = optimizers.SGD(lr = 0.01, decay = 1e-6, momentum = 0.9, nesterov = True)\n",
    "model.compile(optimizer = opt, loss = 'categorical_crossentropy', metrics = ['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using real-time data augmentation.\n",
      "\n",
      "Epoch 1/20\n",
      "1563/1563 [==============================] - 88s 56ms/step - loss: 1.7483 - accuracy: 0.3541 - val_loss: 1.3350 - val_accuracy: 0.5132\n",
      "Epoch 2/20\n",
      "1563/1563 [==============================] - 87s 56ms/step - loss: 1.3276 - accuracy: 0.5240 - val_loss: 1.2008 - val_accuracy: 0.5778\n",
      "Epoch 3/20\n",
      "1563/1563 [==============================] - 87s 56ms/step - loss: 1.1421 - accuracy: 0.5936 - val_loss: 1.0164 - val_accuracy: 0.6508\n",
      "Epoch 4/20\n",
      "1563/1563 [==============================] - 86s 55ms/step - loss: 1.0290 - accuracy: 0.6360 - val_loss: 0.9692 - val_accuracy: 0.6605\n",
      "Epoch 5/20\n",
      "1563/1563 [==============================] - 83s 53ms/step - loss: 0.9522 - accuracy: 0.6654 - val_loss: 0.8530 - val_accuracy: 0.7060\n",
      "Epoch 6/20\n",
      "1563/1563 [==============================] - 83s 53ms/step - loss: 0.9017 - accuracy: 0.6834 - val_loss: 0.8119 - val_accuracy: 0.7166\n",
      "Epoch 7/20\n",
      "1563/1563 [==============================] - 84s 54ms/step - loss: 0.8581 - accuracy: 0.6984 - val_loss: 0.8118 - val_accuracy: 0.7198\n",
      "Epoch 8/20\n",
      "1563/1563 [==============================] - 84s 54ms/step - loss: 0.8219 - accuracy: 0.7103 - val_loss: 0.7871 - val_accuracy: 0.7263\n",
      "Epoch 9/20\n",
      "1563/1563 [==============================] - 84s 54ms/step - loss: 0.7921 - accuracy: 0.7235 - val_loss: 0.7275 - val_accuracy: 0.7529\n",
      "Epoch 10/20\n",
      "1563/1563 [==============================] - 83s 53ms/step - loss: 0.7691 - accuracy: 0.7340 - val_loss: 0.7460 - val_accuracy: 0.7443\n",
      "Epoch 11/20\n",
      "1563/1563 [==============================] - 83s 53ms/step - loss: 0.7527 - accuracy: 0.7367 - val_loss: 0.7412 - val_accuracy: 0.7488\n",
      "Epoch 12/20\n",
      "1563/1563 [==============================] - 84s 54ms/step - loss: 0.7301 - accuracy: 0.7447 - val_loss: 0.7581 - val_accuracy: 0.7415\n",
      "Epoch 13/20\n",
      "1563/1563 [==============================] - 85s 54ms/step - loss: 0.7239 - accuracy: 0.7482 - val_loss: 0.7318 - val_accuracy: 0.7496\n",
      "Epoch 14/20\n",
      "1563/1563 [==============================] - 85s 54ms/step - loss: 0.7018 - accuracy: 0.7536 - val_loss: 0.7044 - val_accuracy: 0.7622\n",
      "Epoch 15/20\n",
      "1563/1563 [==============================] - 84s 54ms/step - loss: 0.6998 - accuracy: 0.7569 - val_loss: 0.7610 - val_accuracy: 0.7359\n",
      "Epoch 16/20\n",
      "1563/1563 [==============================] - 86s 55ms/step - loss: 0.6861 - accuracy: 0.7615 - val_loss: 0.7120 - val_accuracy: 0.7578\n",
      "Epoch 17/20\n",
      "1563/1563 [==============================] - 85s 55ms/step - loss: 0.6758 - accuracy: 0.7658 - val_loss: 0.7302 - val_accuracy: 0.7546\n",
      "Epoch 18/20\n",
      "1563/1563 [==============================] - 85s 55ms/step - loss: 0.6739 - accuracy: 0.7675 - val_loss: 0.6832 - val_accuracy: 0.7681\n",
      "Epoch 19/20\n",
      "1563/1563 [==============================] - 86s 55ms/step - loss: 0.6644 - accuracy: 0.7710 - val_loss: 0.6970 - val_accuracy: 0.7660\n",
      "Epoch 20/20\n",
      "1563/1563 [==============================] - 86s 55ms/step - loss: 0.6592 - accuracy: 0.7744 - val_loss: 0.7502 - val_accuracy: 0.7496\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\n   第四步：訓練\\n   .fit的一些參數\\n   batch_size：對總的樣本數進行分組，每組包含的樣本數量\\n   epochs ：訓練次數\\n   shuffle：是否把數據隨機打亂之後再進行訓練\\n   validation_split：拿出百分之多少用來做交叉驗證\\n   verbose：屏顯模式 - 0：不輸出, 1：輸出進度, 2：輸出每次的訓練結果\\n'"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 資料正規化\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "x_train /= 255\n",
    "x_test /= 255\n",
    "\n",
    "# 是否要做資料處理\n",
    "if not data_augmentation:\n",
    "    print('Not using data augmentation.')\n",
    "    history=model.fit(x_train, y_train,\n",
    "              batch_size=batch_size,\n",
    "              epochs=epochs,\n",
    "              validation_data=(x_test, y_test),\n",
    "              shuffle=True)\n",
    "else:\n",
    "    print('Using real-time data augmentation.')\n",
    "    print('')\n",
    "        \n",
    "    # This will do preprocessing and realtime data augmentation:\n",
    "    datagen = ImageDataGenerator(\n",
    "        featurewise_center=False,  # set input mean to 0 over the dataset\n",
    "        samplewise_center=False,  # set each sample mean to 0\n",
    "        featurewise_std_normalization=False,  # divide inputs by std of the dataset\n",
    "        samplewise_std_normalization=False,  # divide each input by its std\n",
    "        zca_whitening=False,  # apply ZCA whitening\n",
    "        zca_epsilon=1e-06,  # epsilon for ZCA whitening\n",
    "        rotation_range=0,  # randomly rotate images in the range (degrees, 0 to 180)\n",
    "        # randomly shift images horizontally (fraction of total width)\n",
    "        width_shift_range=0.1,\n",
    "        # randomly shift images vertically (fraction of total height)\n",
    "        height_shift_range=0.1,\n",
    "        shear_range=0.,  # set range for random shear\n",
    "        zoom_range=0.,  # set range for random zoom\n",
    "        channel_shift_range=0.,  # set range for random channel shifts\n",
    "        # set mode for filling points outside the input boundaries\n",
    "        fill_mode='nearest',\n",
    "        cval=0.,  # value used for fill_mode = \"constant\"\n",
    "        horizontal_flip=True,  # randomly flip images\n",
    "        vertical_flip=False,  # randomly flip images\n",
    "        # set rescaling factor (applied before any other transformation)\n",
    "        rescale=None,\n",
    "        # set function that will be applied on each input\n",
    "        preprocessing_function=None,\n",
    "        # image data format, either \"channels_first\" or \"channels_last\"\n",
    "        data_format=None,\n",
    "        # fraction of images reserved for validation (strictly between 0 and 1)\n",
    "        validation_split=0.0)\n",
    "\n",
    "    # Compute quantities required for feature-wise normalization\n",
    "    # (std, mean, and principal components if ZCA whitening is applied).\n",
    "    datagen.fit(x_train)\n",
    "    history=model.fit(x_train, y_train,\n",
    "              batch_size=batch_size,\n",
    "              epochs=epochs,\n",
    "              validation_data=(x_test, y_test),\n",
    "              shuffle=True)   \n",
    "\n",
    "'''\n",
    "   第四步：訓練\n",
    "   .fit的一些參數\n",
    "   batch_size：對總的樣本數進行分組，每組包含的樣本數量\n",
    "   epochs ：訓練次數\n",
    "   shuffle：是否把數據隨機打亂之後再進行訓練\n",
    "   validation_split：拿出百分之多少用來做交叉驗證\n",
    "   verbose：屏顯模式 - 0：不輸出, 1：輸出進度, 2：輸出每次的訓練結果\n",
    "''' \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saved trained model at C:\\Users\\cerbe\\Desktop\\ML_marathon\\saved_models\\keras_cifar10_trained_model_hw.h5 \n",
      "313/313 [==============================] - 2s 6ms/step - loss: 0.7502 - accuracy: 0.7496\n",
      "Test loss: 0.7502145171165466\n",
      "Test accuracy: 0.7495999932289124\n",
      " test set \n",
      "\n",
      " The test loss is %f  [0.7502143979072571, 0.7495999932289124]\n"
     ]
    }
   ],
   "source": [
    "# Save model and weights\n",
    "if not os.path.isdir(save_dir):\n",
    "    os.makedirs(save_dir)\n",
    "model_path = os.path.join(save_dir, model_name)\n",
    "model.save(model_path)\n",
    "print('Saved trained model at %s ' % model_path)\n",
    "\n",
    "# Score trained model.\n",
    "scores = model.evaluate(x_test, y_test, verbose=1)\n",
    "print('Test loss:', scores[0])\n",
    "print('Test accuracy:', scores[1])\n",
    "\n",
    "\n",
    "#    第六步：輸出\n",
    "import numpy \n",
    "print ( \" test set \" )\n",
    "scores = model.evaluate(x_test,y_test,batch_size=200,verbose= 0)\n",
    "print ( \"\" )\n",
    "#print ( \" The test loss is %f \" % scores)\n",
    "print ( \" The test loss is %f \", scores)\n",
    "\n",
    "\n",
    "result = model.predict(x_test,batch_size=200,verbose= 0)\n",
    "\n",
    "result_max = numpy.argmax(result, axis = 1 )\n",
    "test_max = numpy.argmax(y_test, axis = 1 )\n",
    "\n",
    "result_bool = numpy.equal(result_max, test_max)\n",
    "true_num = numpy.sum(result_bool)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "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",
    "%matplotlib inline\n",
    "\n",
    "# Plot training & validation accuracy values\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', 'Valiidation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Valiidation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Rmsprop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "#第三步編譯\n",
    "'''\n",
    " SGD(隨機梯度下降) - Arguments\n",
    "lr: float >= 0. Learning rate.\n",
    "momentum: float >= 0. Parameter that accelerates SGD in the relevant direction and dampens oscillations.\n",
    "decay: float >= 0. Learning rate decay over each update.\n",
    "nesterov: boolean. Whether to apply Nesterov momentum.\n",
    "'''\n",
    "\n",
    "'''\n",
    "RMSprop- Arguments\n",
    "lr: float >= 0. Learning rate.\n",
    "rho: float >= 0.\n",
    "epsilon: float >= 0. Fuzz factor. If None, defaults to K.epsilon().\n",
    "decay: float >= 0. Learning rate decay over each update.\n",
    "'''\n",
    "\n",
    "'''\n",
    "Example:\n",
    "opt = optimizers.SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "\n",
    "'''\n",
    "opt = optimizers.RMSprop(lr = 0.01, decay = 1e-6)\n",
    "model.compile(optimizer = opt, loss = 'categorical_crossentropy', metrics = ['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using real-time data augmentation.\n",
      "\n",
      "Epoch 1/20\n",
      "1563/1563 [==============================] - 92s 59ms/step - loss: 2.5640 - accuracy: 0.0981 - val_loss: 2.3039 - val_accuracy: 0.1000\n",
      "Epoch 2/20\n",
      "1563/1563 [==============================] - 91s 58ms/step - loss: 2.3039 - accuracy: 0.1013 - val_loss: 2.3042 - val_accuracy: 0.1000\n",
      "Epoch 3/20\n",
      "1563/1563 [==============================] - 92s 59ms/step - loss: 2.3041 - accuracy: 0.0991 - val_loss: 2.3034 - val_accuracy: 0.1000\n",
      "Epoch 4/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3039 - accuracy: 0.0971 - val_loss: 2.3039 - val_accuracy: 0.1000\n",
      "Epoch 5/20\n",
      "1563/1563 [==============================] - 93s 59ms/step - loss: 2.3040 - accuracy: 0.1012 - val_loss: 2.3037 - val_accuracy: 0.1000\n",
      "Epoch 6/20\n",
      "1563/1563 [==============================] - 93s 60ms/step - loss: 2.3040 - accuracy: 0.0987 - val_loss: 2.3031 - val_accuracy: 0.1000\n",
      "Epoch 7/20\n",
      "1563/1563 [==============================] - 92s 59ms/step - loss: 2.3039 - accuracy: 0.0994 - val_loss: 2.3034 - val_accuracy: 0.1000\n",
      "Epoch 8/20\n",
      "1563/1563 [==============================] - 92s 59ms/step - loss: 2.3038 - accuracy: 0.1017 - val_loss: 2.3047 - val_accuracy: 0.1000\n",
      "Epoch 9/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3040 - accuracy: 0.1012 - val_loss: 2.3042 - val_accuracy: 0.1000\n",
      "Epoch 10/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3039 - accuracy: 0.0994 - val_loss: 2.3040 - val_accuracy: 0.1000\n",
      "Epoch 11/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3040 - accuracy: 0.0996 - val_loss: 2.3045 - val_accuracy: 0.1000\n",
      "Epoch 12/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3040 - accuracy: 0.0991 - val_loss: 2.3035 - val_accuracy: 0.1000\n",
      "Epoch 13/20\n",
      "1563/1563 [==============================] - 95s 61ms/step - loss: 2.3038 - accuracy: 0.1014 - val_loss: 2.3036 - val_accuracy: 0.1000\n",
      "Epoch 14/20\n",
      "1563/1563 [==============================] - 100s 64ms/step - loss: 2.3040 - accuracy: 0.0986 - val_loss: 2.3037 - val_accuracy: 0.1000\n",
      "Epoch 15/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3041 - accuracy: 0.0985 - val_loss: 2.3037 - val_accuracy: 0.1000\n",
      "Epoch 16/20\n",
      "1563/1563 [==============================] - 93s 60ms/step - loss: 2.3041 - accuracy: 0.1004 - val_loss: 2.3038 - val_accuracy: 0.1000\n",
      "Epoch 17/20\n",
      "1563/1563 [==============================] - 95s 61ms/step - loss: 2.3039 - accuracy: 0.0988 - val_loss: 2.3059 - val_accuracy: 0.1000\n",
      "Epoch 18/20\n",
      "1563/1563 [==============================] - 95s 61ms/step - loss: 2.3039 - accuracy: 0.1006 - val_loss: 2.3037 - val_accuracy: 0.1000\n",
      "Epoch 19/20\n",
      "1563/1563 [==============================] - 93s 60ms/step - loss: 2.3040 - accuracy: 0.0981 - val_loss: 2.3030 - val_accuracy: 0.1000\n",
      "Epoch 20/20\n",
      "1563/1563 [==============================] - 94s 60ms/step - loss: 2.3039 - accuracy: 0.0983 - val_loss: 2.3036 - val_accuracy: 0.1000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\n   第四步：訓練\\n   .fit的一些參數\\n   batch_size：對總的樣本數進行分組，每組包含的樣本數量\\n   epochs ：訓練次數\\n   shuffle：是否把數據隨機打亂之後再進行訓練\\n   validation_split：拿出百分之多少用來做交叉驗證\\n   verbose：屏顯模式 - 0：不輸出, 1：輸出進度, 2：輸出每次的訓練結果\\n'"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 資料正規化\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "x_train /= 255\n",
    "x_test /= 255\n",
    "\n",
    "# 是否要做資料處理\n",
    "if not data_augmentation:\n",
    "    print('Not using data augmentation.')\n",
    "    history=model.fit(x_train, y_train,\n",
    "              batch_size=batch_size,\n",
    "              epochs=epochs,\n",
    "              validation_data=(x_test, y_test),\n",
    "              shuffle=True)\n",
    "else:\n",
    "    print('Using real-time data augmentation.')\n",
    "    print('')\n",
    "        \n",
    "    # This will do preprocessing and realtime data augmentation:\n",
    "    datagen = ImageDataGenerator(\n",
    "        featurewise_center=False,  # set input mean to 0 over the dataset\n",
    "        samplewise_center=False,  # set each sample mean to 0\n",
    "        featurewise_std_normalization=False,  # divide inputs by std of the dataset\n",
    "        samplewise_std_normalization=False,  # divide each input by its std\n",
    "        zca_whitening=False,  # apply ZCA whitening\n",
    "        zca_epsilon=1e-06,  # epsilon for ZCA whitening\n",
    "        rotation_range=0,  # randomly rotate images in the range (degrees, 0 to 180)\n",
    "        # randomly shift images horizontally (fraction of total width)\n",
    "        width_shift_range=0.1,\n",
    "        # randomly shift images vertically (fraction of total height)\n",
    "        height_shift_range=0.1,\n",
    "        shear_range=0.,  # set range for random shear\n",
    "        zoom_range=0.,  # set range for random zoom\n",
    "        channel_shift_range=0.,  # set range for random channel shifts\n",
    "        # set mode for filling points outside the input boundaries\n",
    "        fill_mode='nearest',\n",
    "        cval=0.,  # value used for fill_mode = \"constant\"\n",
    "        horizontal_flip=True,  # randomly flip images\n",
    "        vertical_flip=False,  # randomly flip images\n",
    "        # set rescaling factor (applied before any other transformation)\n",
    "        rescale=None,\n",
    "        # set function that will be applied on each input\n",
    "        preprocessing_function=None,\n",
    "        # image data format, either \"channels_first\" or \"channels_last\"\n",
    "        data_format=None,\n",
    "        # fraction of images reserved for validation (strictly between 0 and 1)\n",
    "        validation_split=0.0)\n",
    "\n",
    "    # Compute quantities required for feature-wise normalization\n",
    "    # (std, mean, and principal components if ZCA whitening is applied).\n",
    "    datagen.fit(x_train)\n",
    "    history=model.fit(x_train, y_train,\n",
    "              batch_size=batch_size,\n",
    "              epochs=epochs,\n",
    "              validation_data=(x_test, y_test),\n",
    "              shuffle=True)   \n",
    "\n",
    "'''\n",
    "   第四步：訓練\n",
    "   .fit的一些參數\n",
    "   batch_size：對總的樣本數進行分組，每組包含的樣本數量\n",
    "   epochs ：訓練次數\n",
    "   shuffle：是否把數據隨機打亂之後再進行訓練\n",
    "   validation_split：拿出百分之多少用來做交叉驗證\n",
    "   verbose：屏顯模式 - 0：不輸出, 1：輸出進度, 2：輸出每次的訓練結果\n",
    "''' \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saved trained model at C:\\Users\\cerbe\\Desktop\\ML_marathon\\saved_models\\keras_cifar10_trained_model_hw.h5 \n",
      "313/313 [==============================] - 2s 6ms/step - loss: 2.3036 - accuracy: 0.1000\n",
      "Test loss: 2.3035616874694824\n",
      "Test accuracy: 0.10000000149011612\n",
      " test set \n",
      "\n",
      " The test loss is %f  [2.303561210632324, 0.10000000149011612]\n"
     ]
    }
   ],
   "source": [
    "# Save model and weights\n",
    "if not os.path.isdir(save_dir):\n",
    "    os.makedirs(save_dir)\n",
    "model_path = os.path.join(save_dir, model_name)\n",
    "model.save(model_path)\n",
    "print('Saved trained model at %s ' % model_path)\n",
    "\n",
    "# Score trained model.\n",
    "scores = model.evaluate(x_test, y_test, verbose=1)\n",
    "print('Test loss:', scores[0])\n",
    "print('Test accuracy:', scores[1])\n",
    "\n",
    "\n",
    "#    第六步：輸出\n",
    "import numpy \n",
    "print ( \" test set \" )\n",
    "scores = model.evaluate(x_test,y_test,batch_size=200,verbose= 0)\n",
    "print ( \"\" )\n",
    "#print ( \" The test loss is %f \" % scores)\n",
    "print ( \" The test loss is %f \", scores)\n",
    "\n",
    "\n",
    "result = model.predict(x_test,batch_size=200,verbose= 0)\n",
    "\n",
    "result_max = numpy.argmax(result, axis = 1 )\n",
    "test_max = numpy.argmax(y_test, axis = 1 )\n",
    "\n",
    "result_bool = numpy.equal(result_max, test_max)\n",
    "true_num = numpy.sum(result_bool)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "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",
    "%matplotlib inline\n",
    "\n",
    "# Plot training & validation accuracy values\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', 'Valiidation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Valiidation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "採用Rmsprop的成果比SGD的更差，完全無法收斂；總體而言還是以Adam優化器為佳"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}