28-深度学习之原理及mnist分类.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "比如对于 MNIST 数字，全局模式就是涉及所有像素的模式，这个参数太多了，随便进行一次反向传播计算量都是巨大的，从计算资源和调参的角度都不建议用传统的神经网络，参数过多还会导致出现过拟合情况），而卷积层学到的是局部模式，也就是参数会少很多！！！！那什么是卷积核呢"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div  align=\"center\">    \n",
    "    <img src = './img/16.jpg', width=600, height=200/>\n",
    "    <img src = './img/17.jpg', width=600, height=200/>\n",
    "    <img src = './img/18.jpg', width=600, height=200/>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Administrator\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 26, 26, 32)        320       \n",
      "=================================================================\n",
      "Total params: 320\n",
      "Trainable params: 320\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 现在我们来验证一下\n",
    "from keras import layers \n",
    "from keras import models\n",
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "确实是（26, 26, 32）"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "最大池化"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div  align=\"center\">    \n",
    "    <img src = './img/19.jpg', width=600, height=200/>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_2 (Conv2D)            (None, 26, 26, 32)        320       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 13, 13, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 11, 11, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 5, 5, 64)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 3, 3, 64)          36928     \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 576)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 64)                36928     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                650       \n",
      "=================================================================\n",
      "Total params: 93,322\n",
      "Trainable params: 93,322\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 现在我们来验证一下\n",
    "from keras import layers \n",
    "from keras import models\n",
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu')) \n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(10, activation='softmax'))  # 多分类用softmax， 二分类可以用sigmoid\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "在次模型的基础上训练一下我们的手写收据集"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "# 加载数据集\n",
    "path = './files/mnist.npz'\n",
    "f = np.load(path)\n",
    "train_x, train_y = f['x_train'], f['y_train']\n",
    "test_x, test_y = f['x_test'], f['y_test']\n",
    "f.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10000, 28, 28, 1)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_x.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 10, 2)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 10, 2, 2)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 因为我们的模型输入需要是3D，所以对我们数据的维度进行reshape一下\n",
    "train_x = train_x.reshape((60000, 28, 28, 1))\n",
    "train_x = train_x.astype('float64') / 255    # 相当于进行数据的归一化，因为像素是0-255之间的\n",
    "test_x = test_x.reshape((10000, 28, 28, 1))\n",
    "test_x = test_x.astype('float64') / 255\n",
    "from keras.utils import to_categorical\n",
    "train_y = to_categorical(train_y)  # 化成独热码的形式[1, 0, 0, 0, 0, 0, 0, 0, 0, 0] 代表的是1\n",
    "test_y = to_categorical(test_y)    #                 [0, 1, 0, 0, 0, 0, 0, 0, 0, 0] 代表的是2\n",
    "\n",
    "train_y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='rmsprop',\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "model.fit(train_x, train_y, epochs=5, batch_size=64)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "END"
   ]
  }
 ],
 "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}