HW-20191117-PM-Keras-MNIST-手寫辨識-Starter-normalized pixel value, square of normalized value, and square root of normalized value.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using CNTK backend\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GPU\n"
     ]
    }
   ],
   "source": [
    "# 如何知道有多少數據\n",
    "from keras import datasets\n",
    "from keras.utils import to_categorical\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "#print(dir(datasets))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((60000, 28, 28), (60000,), (10000, 28, 28), (10000,))"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 數據是28 * 28 = 784\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "X_train.shape, y_train.shape, X_test.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 72x72 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 畫一張\n",
    "plt.figure(figsize=(1,1))\n",
    "plt.imshow(X_train[0], cmap='gray')\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Lab2_LogisticRegression - 講解 - Lab\n",
    "# 增加 特徵數量 可以增加精準度\n",
    "# 試試看  (normalized pixel value, square of normalized value, and square root of normalized value). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# data pre-processing\n",
    "X_train, X_test = X_train.astype(np.float32).reshape(60000, 784)/255, X_test.astype(np.float32).reshape(10000, 784)/255\n",
    "# one_hot encoding\n",
    "y_train_oh, y_test_oh = to_categorical(y_train), to_categorical(y_test) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# y_train[:10] , y_train_oh[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 10)                7850      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                110       \n",
      "=================================================================\n",
      "Total params: 7,960\n",
      "Trainable params: 7,960\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# build model\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras import optimizers\n",
    "model = Sequential()\n",
    "model.add( Dense(units=10, input_shape=(784,) ) )\n",
    "model.add( Dense(units=10, activation='softmax') )\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 54000 samples, validate on 6000 samples\n",
      "Epoch 1/10\n",
      "54000/54000 [==============================] - 3s 58us/step - loss: 0.2585 - acc: 0.9263 - val_loss: 0.2414 - val_acc: 0.9357\n",
      "Epoch 2/10\n",
      "54000/54000 [==============================] - 3s 54us/step - loss: 0.2572 - acc: 0.9281 - val_loss: 0.2533 - val_acc: 0.9317\n",
      "Epoch 3/10\n",
      "54000/54000 [==============================] - 3s 54us/step - loss: 0.2572 - acc: 0.9276 - val_loss: 0.2463 - val_acc: 0.9332\n",
      "Epoch 4/10\n",
      "54000/54000 [==============================] - 3s 55us/step - loss: 0.2566 - acc: 0.9274 - val_loss: 0.2506 - val_acc: 0.9333\n",
      "Epoch 5/10\n",
      "54000/54000 [==============================] - 3s 55us/step - loss: 0.2559 - acc: 0.9278 - val_loss: 0.2523 - val_acc: 0.9293\n",
      "Epoch 6/10\n",
      "54000/54000 [==============================] - 3s 55us/step - loss: 0.2563 - acc: 0.9273 - val_loss: 0.2462 - val_acc: 0.9325\n",
      "Epoch 7/10\n",
      "54000/54000 [==============================] - 3s 56us/step - loss: 0.2562 - acc: 0.9280 - val_loss: 0.2479 - val_acc: 0.9338\n",
      "Epoch 8/10\n",
      "54000/54000 [==============================] - 4s 70us/step - loss: 0.2550 - acc: 0.9279 - val_loss: 0.2581 - val_acc: 0.9300\n",
      "Epoch 9/10\n",
      "54000/54000 [==============================] - 4s 75us/step - loss: 0.2561 - acc: 0.9289 - val_loss: 0.2432 - val_acc: 0.9348\n",
      "Epoch 10/10\n",
      "54000/54000 [==============================] - 3s 63us/step - loss: 0.2547 - acc: 0.9284 - val_loss: 0.2525 - val_acc: 0.9310\n"
     ]
    }
   ],
   "source": [
    "# train model \n",
    "# ValueError: Error when checking target: expected dense_3 to have 2 dimensions\n",
    "# but got array with shape (60000, 10, 2, 2)\n",
    "# ValueError: Error when checking target: expected dense_6 to have shape (10,) but got array with shape (1,)\n",
    "\n",
    "model.compile(optimizer=optimizers.SGA( 0.2), loss='categorical_crossentropy', metrics=['accuracy'] )\n",
    "history = model.fit(X_train, y_train_oh, batch_size=64, validation_split=0.1, epochs=10, verbose=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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": [
    "# Plot\n",
    "plt.title('Accuracy of Keras')\n",
    "plt.plot(history.history['acc'], color='red')\n",
    "plt.plot(history.history['val_acc'], color='blue')\n",
    "plt.grid()\n",
    "plt.show()\n",
    "\n",
    "\n",
    "plt.title('Loss of Keras')\n",
    "plt.plot(history.history['loss'], color='red')\n",
    "plt.plot(history.history['val_loss'], color='blue')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 4s 68us/step\n",
      "10000/10000 [==============================] - 1s 66us/step\n",
      "[0.2529372300858299, 0.9293666666666667] [0.2966662008479238, 0.9199]\n"
     ]
    }
   ],
   "source": [
    "# evaluate model \n",
    "print(model.evaluate(X_train, y_train_oh), model.evaluate(X_test, y_test_oh) )\n",
    "# predict model \n",
    "y_pred = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# top 10 \n",
    "# 請你找一下那些數字是錯誤 slicing \n",
    "index_error = np.argmax(y_pred, axis=1) != np.argmax(y_test_oh, axis=1)\n",
    "\n",
    "X_error      =  X_test[ index_error ]\n",
    "y_error      =  y_test_oh[ index_error ]\n",
    "y_error_pred =  y_pred[ index_error ]\n",
    "\n",
    "print('Total error amount is {}'.format(X_error.shape[0]))\n",
    "\n",
    "# Plot for top 10 [:10]\n",
    "for i in range(10):\n",
    "    plt.figure(figsize=(1,1))\n",
    "    plt.title('GT:{} --> Pred:{}'.format(np.argmax(y_error[i]), np.argmax(y_error_pred[i])) )\n",
    "    plt.imshow(X_error[i].reshape(28,28), cmap='gray' )\n",
    "    plt.axis('off')\n",
    "    plt.show()"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}