{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 机器学习与社会科学应用"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 第七章 神经网络"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 第二节 卷积神经网络"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"郭峰 \n",
" 教授、博士生导师 \n",
"上海财经大学公共经济与管理学院 \n",
"上海财经大学数实融合与智能治理实验室 \n",
" 邮箱：guofengsfi@163.com "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"本节目录： \n",
"2.1.导入模块 \n",
"2.2.加载数据 \n",
"2.3.预处理数据 \n",
"2.4.定义模型结构 \n",
"2.5.编译 \n",
"2.6.训练 \n",
"2.7.评估模型 "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.1.导入模块"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from keras.models import Sequential #from keras.models import Sequential\n",
"#接下来，我们导入各种层(各种形状各异积木)\n",
"from keras.layers import Conv2D, MaxPool2D\n",
"from keras.layers import Dense, Flatten\n",
"\n",
"#最后，我们导入to_categorical函数，以便之后对数据进行转换\n",
"#from keras.utils import to_categorical\n",
"\n",
"import tensorflow as tf"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.2.加载数据"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#MNIST是一个非常有名的手写数字数据集，我们可以使用Keras轻松加载它\n",
"from keras.datasets import mnist\n",
"(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
"print(x_train.shape)# (60000, 28, 28) 60000个样本，它们都是28像素x28像素的"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#看一下这些手写数字长什么样\n",
"import matplotlib.pyplot as plt\n",
"%matplotlib inline\n",
"plt.imshow(x_train[0])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.3.预处理数据"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#使用Keras是必须显式声明输入图像深度的尺寸。例如，具有所有3个RGB通道的全色图像的深度为3。\n",
"#我们的MNIST图像的深度为1，但我们必须明确声明。\n",
"#也就是说，我们希望将数据集从形状(n,rows,cols)转换为(n,rows,cols,channels)。\n",
"img_x, img_y = 28, 28\n",
"x_train = x_train.reshape(x_train.shape[0], img_x, img_y, 1)\n",
"x_test = x_test.reshape(x_test.shape[0], img_x, img_y, 1)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#除此之外，我们将数据标准化一下：\n",
"x_train = x_train.astype('float32')\n",
"x_test = x_test.astype('float32')\n",
"x_train /= 255\n",
"x_test /= 255"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#将标记值(y_train, y_test)转换为One-Hot Encode的形式\n",
"y_train = tf.keras.utils.to_categorical(y_train, 10)\n",
"y_test = tf.keras.utils.to_categorical(y_test, 10)\n",
"print(y_train.shape) # (60000, 10)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.4.定义模型结构"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"model = Sequential()\n",
"model.add(Conv2D(32, kernel_size=(5,5), activation='relu', input_shape=(img_x, img_y, 1))) #一层卷积层，包含了32个卷积核，大小为5*5\n",
"model.add(MaxPool2D(pool_size=(2,2), strides=(2,2))) #一个最大池化层，池化大小为2*2\n",
"model.add(Conv2D(64, kernel_size=(5,5), activation='relu')) #添加一个卷积层，包含64个卷积和，每个卷积核仍未5*5\n",
"model.add(MaxPool2D(pool_size=(2,2), strides=(2,2))) #一个最大池化层，池化大小为2*2\n",
"model.add(Flatten()) #压平层\n",
"model.add(Dense(1000, activation='relu')) #来一个全连接层\n",
"model.add(Dense(10, activation='softmax')) #最后为分类层"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#另一个定义\n",
"# build the neural net 建模型(卷积—relu-卷积-relu-池化-relu-卷积-relu-池化-全连接)\n",
"#model = Sequential()\n",
"#model.add(Conv2D(32, kernel_size=(3, 3),\n",
"# activation='relu',\n",
"# input_shape=input_shape)) # 32个过滤器，过滤器大小是3×3，32×26×26\n",
"#model.add(Conv2D(64, (3, 3), activation='relu')) #64×24×24\n",
"#model.add(MaxPooling2D(pool_size=(2, 2)))# 向下取样\n",
"#model.add(Dropout(0.25))\n",
"#model.add(Flatten()) #降维：将64×12×12降为1维（即把他们相乘起来）#\n",
"#model.add(Dense(128, activation='relu'))\n",
"#model.add(Dropout(0.5))\n",
"#model.add(Dense(num_classes, activation='softmax')) #全连接2层\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.5.编译"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#现在，只需要编译模型，就可以开始训练了。当编译模型时，我们声明了损失函数和优化器（SGD，Adam等）。\n",
"model.compile(optimizer='adam',\n",
" loss='categorical_crossentropy',\n",
" metrics=['accuracy'])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.6.训练"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"model.fit(x_train, y_train, batch_size=128, epochs=10)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2.7.评估模型"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"score = model.evaluate(x_test, y_test)\n",
"print('acc', score[1]) # acc 0.9926"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"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": "3.9.12"
}
},
"nbformat": 4,
"nbformat_minor": 2
}