ch04-手写体识别/02-MNIST-softmax.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 加载MNIST数据集"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import mnist\n",
    "(X_train, y_train),(X_test,y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28) (60000,) <class 'numpy.ndarray'>\n",
      "(10000, 28, 28) (10000,) <class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, y_train.shape, type(X_train))\n",
    "print(X_test.shape, y_test.shape, type(X_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 数据处理：规范化"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 784) <class 'numpy.ndarray'>\n",
      "(10000, 784) <class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "# 将图像本身从(28,28)转换成(784,)\n",
    "X_train = X_train.reshape(60000, 784)\n",
    "X_test = X_test.reshape(10000, 784)\n",
    "print(X_train.shape, type(X_train))\n",
    "print(X_test.shape, type(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 将数据类型转换成float32\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "# 数据归一化，不转float32除255都是0\n",
    "X_train /= 255\n",
    "X_test /= 255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 统计训练数据中各标签数量"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0 1 2 3 4 5 6 7 8 9] [5923 6742 5958 6131 5842 5421 5918 6265 5851 5949]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "label, count = np.unique(y_train, return_counts=True)\n",
    "print(label, count)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x137b33f28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "plt.bar(label, count, width = 0.7, align='center')\n",
    "plt.title(\"Label Distribution\")\n",
    "plt.xlabel(\"Label\")\n",
    "plt.ylabel(\"Count\")\n",
    "plt.xticks(label)\n",
    "plt.ylim(0,7500)\n",
    "\n",
    "for a,b in zip(label, count):\n",
    "    plt.text(a, b, '%d' % b, ha='center', va='bottom',fontsize=10)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## one-hot 编码"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape before one-hot encoding:  (60000,)\n",
      "Shape after one-hot encoding:  (60000, 10)\n"
     ]
    }
   ],
   "source": [
    "from keras.utils import np_utils\n",
    "\n",
    "n_classes = 10\n",
    "print(\"Shape before one-hot encoding: \", y_train.shape)\n",
    "Y_train = np_utils.to_categorical(y_train, n_classes)\n",
    "print(\"Shape after one-hot encoding: \", Y_train.shape)\n",
    "Y_test = np_utils.to_categorical(y_test, n_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5\n",
      "[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n"
     ]
    }
   ],
   "source": [
    "print(y_train[0])\n",
    "print(Y_train[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 使用 Keras sequential model 定义神经网络\n",
    "\n",
    "### softmax 网络层\n",
    "\n",
    "![](https://developers.google.com/machine-learning/crash-course/images/SoftmaxLayer.svg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Activation\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(512, input_shape=(784,))) #输入\n",
    "model.add(Activation('relu'))             #激活层                        \n",
    "\n",
    "model.add(Dense(512))                     #激活层输出作为下一层的输入\n",
    "model.add(Activation('relu'))\n",
    "\n",
    "model.add(Dense(10))                      #神经元为10的全连接层\n",
    "model.add(Activation('softmax'))          #输出10个概率，0-9的10个标签"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 编译模型\n",
    "\n",
    "[model.compile()](https://keras.io/models/sequential/#compile)\n",
    "\n",
    "```python\n",
    "compile(optimizer, loss=None, metrics=None, loss_weights=None, sample_weight_mode=None, weighted_metrics=None, target_tensors=None)\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 训练模型，并将指标保存到 history 中\n",
    "\n",
    "[model.fit()](https://keras.io/models/sequential/#fit)\n",
    "\n",
    "```python\n",
    "fit(x=None, y=None, batch_size=None, epochs=1, verbose=1, callbacks=None, validation_split=0.0, validation_data=None, shuffle=True, class_weight=None, sample_weight=None, initial_epoch=0, steps_per_epoch=None, validation_steps=None)\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/5\n",
      " - 7s - loss: 0.2188 - acc: 0.9362 - val_loss: 0.1035 - val_acc: 0.9683\n",
      "Epoch 2/5\n",
      " - 6s - loss: 0.0799 - acc: 0.9747 - val_loss: 0.0704 - val_acc: 0.9781\n",
      "Epoch 3/5\n",
      " - 6s - loss: 0.0521 - acc: 0.9837 - val_loss: 0.0798 - val_acc: 0.9749\n",
      "Epoch 4/5\n",
      " - 6s - loss: 0.0361 - acc: 0.9886 - val_loss: 0.0699 - val_acc: 0.9787\n",
      "Epoch 5/5\n",
      " - 6s - loss: 0.0275 - acc: 0.9909 - val_loss: 0.0810 - val_acc: 0.9762\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train,\n",
    "                    Y_train,\n",
    "                    batch_size=128,\n",
    "                    epochs=5,\n",
    "                    verbose=2,\n",
    "                    validation_data=(X_test, Y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 可视化指标"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x128b499e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "plt.subplot(2,1,1)\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('Model Accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='lower right')\n",
    "\n",
    "plt.subplot(2,1,2)\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', 'test'], loc='upper right')\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 保存模型\n",
    "\n",
    "[model.save()](https://keras.io/getting-started/faq/#how-can-i-save-a-keras-model)\n",
    "\n",
    "You can use `model.save(filepath)` to save a Keras model into a single **HDF5 file** which will contain:\n",
    "\n",
    "- the architecture of the model, allowing to re-create the model\n",
    "- the weights of the model\n",
    "- the training configuration (loss, optimizer)\n",
    "- the state of the optimizer, allowing to resume training exactly where you left off.\n",
    "\n",
    "You can then use `keras.models.load_model(filepath)` to reinstantiate your model. load_model will also take care of compiling the model using the saved training configuration (unless the model was never compiled in the first place)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saved trained model at ./mnist/model/keras_mnist.h5 \n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import tensorflow.gfile as gfile\n",
    "\n",
    "save_dir = \"./mnist/model/\"\n",
    "\n",
    "if gfile.Exists(save_dir):\n",
    "    gfile.DeleteRecursively(save_dir)\n",
    "gfile.MakeDirs(save_dir)\n",
    "\n",
    "model_name = 'keras_mnist.h5'\n",
    "model_path = os.path.join(save_dir, model_name)\n",
    "model.save(model_path)\n",
    "print('Saved trained model at %s ' % model_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 加载模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "\n",
    "mnist_model = load_model(model_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 统计模型在测试集上的分类结果"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Loss: 0.08101196186965971\n",
      "Test Accuracy: 97.61999999999999%\n",
      "Classified correctly count: 9762\n",
      "Classified incorrectly count: 238\n"
     ]
    }
   ],
   "source": [
    "loss_and_metrics = mnist_model.evaluate(X_test, Y_test, verbose=2)\n",
    "    \n",
    "print(\"Test Loss: {}\".format(loss_and_metrics[0]))\n",
    "print(\"Test Accuracy: {}%\".format(loss_and_metrics[1]*100))\n",
    "\n",
    "predicted_classes = mnist_model.predict_classes(X_test)\n",
    "\n",
    "correct_indices = np.nonzero(predicted_classes == y_test)[0]\n",
    "incorrect_indices = np.nonzero(predicted_classes != y_test)[0]\n",
    "print(\"Classified correctly count: {}\".format(len(correct_indices)))\n",
    "print(\"Classified incorrectly count: {}\".format(len(incorrect_indices)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7, 2, 1, 0, 4, 1, 4, 9, 5, 9])"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predicted_classes[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7, 2, 1, 0, 4, 1, 4, 9, 5, 9], dtype=uint8)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 92, 115, 149, 247, 259, 321, 339, 340, 381, 445])"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "incorrect_indices[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "92"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "incorrect_indices[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test[92]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 绘制错误的手写体数字，其中A为正确标签"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a3610b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig  =plt.figure()\n",
    "for i in range(15):\n",
    "    plt.subplot(3, 5, i+1)\n",
    "    plt.tight_layout() #自动适配子图尺寸，间距\n",
    "    incor_index = incorrect_indices[i]\n",
    "    plt.imshow(X_test[incor_index].reshape(28,28), cmap='Greys') #灰色显示像素灰度值\n",
    "    #设置对应标签为子图标题\n",
    "    plt.title(\"P*{} A={}\".format(predicted_classes[incor_index], y_test[incor_index]))\n",
    "    plt.xticks([]) #删除x轴\n",
    "    plt.yticks([])"
   ]
  }
 ],
 "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": "3.6.4rc1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}