实验简介
??本次实验一方面是熟悉Keras 序列式(Sequential)模型的使用，另一方面是复现早期的经典网络结构来学习神经网络搭建的技巧。数据集采用的是熟知的Minst手写识别，框架采用的是tensorflow、Keras，数据集和框架的导入和安装请点击这里。经典的网络结构已有大量博客进行理论分析，这里只给出代码仅供参考，关于神经网络结构的发展，推荐大家看这篇文章。
DeepDense
??这个是自己定义的名字，也就是深度全连接网络。
# -*- coding: utf-8 -*-
""" Created on Thu Jun 13 11:19:33 2019@author: YLC """from keras.datasets import mnist
from keras.models import Sequential
from keras.layers.core import Dense, Activation, Dropout
from keras.utils import np_utils#数据集导入模块
(X_train, y_train), (X_test, y_test) = mnist.load_data();#参数定义模块
img_rows, img_cols = 28,28# input dimensions
batch_size = 64
num_classes = 10
epochs = 2
img_shape = (img_rows,img_cols,1) #预处理 标准化模块
X_train = X_train.reshape(len(X_train), -1)
X_test = X_test.reshape(len(X_test), -1)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train = (X_train - 127) / 127
X_test = (X_test - 127) / 127#分类转One-Hot模块
y_train = np_utils.to_categorical(y_train, num_classes)
y_test = np_utils.to_categorical(y_test, num_classes)#网络搭建模块
model = Sequential()model.add(Dense(512, input_shape=(784,), kernel_initializer='he_normal'))#全连接层
model.add(Activation('relu'))
model.add(Dropout(0.2)) model.add(Dense(512, kernel_initializer='he_normal'))
model.add(Activation('relu'))
model.add(Dropout(0.2)) model.add(Dense(num_classes))
model.add(Activation('softmax'))
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])#运行调参模块
#epochs是迭代的次数，暂定2次是为了迅速看结果，最后改成5次
model.fit(X_train, y_train, epochs=epochs, batch_size= batch_size, verbose=1, validation_split=0.05)
loss, accuracy = model.evaluate(X_test, y_test)#输出模块
print('Test loss:', loss)
print('Accuracy:', accuracy)
model.summary()
LeNet
# -*- coding: utf-8 -*-
""" Created on Thu Jun 13 11:19:33 2019@author: YLC """import numpy as np
import matplotlib.pyplot as pltfrom keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Activation, Conv2D, MaxPooling2D, Flatten, Dropout
from keras.utils import np_utils#数据集导入模块
(X_train, y_train), (X_test, y_test) = mnist.load_data();#参数定义模块
img_rows, img_cols = 28,28# input dimensions
batch_size = 64
num_classes = 10
epochs = 2
img_shape = (img_rows,img_cols,1) #预处理 标准化模块
X_train = X_train.reshape(-1, img_rows, img_cols, 1) # normalize
X_test = X_test.reshape(-1, img_rows, img_cols, 1) # normalize
X_train = X_train / 255
X_test = X_test / 255#分类转One-Hot模块
y_train = np_utils.to_categorical(y_train, num_classes=num_classes)
y_test = np_utils.to_categorical(y_test, num_classes=num_classes)#网络搭建模块
model = Sequential()model.add(Conv2D(input_shape=img_shape, kernel_size=(5, 5), filters=20, activation='relu'))#卷积层
model.add(MaxPooling2D(pool_size=(2,2), strides=2, padding='same'))#池化层model.add(Conv2D(kernel_size=(5, 5), filters=50, activation='relu', padding='same'))
model.add(MaxPooling2D(pool_size=(2,2), strides=2, padding='same'))model.add(Flatten())
model.add(Dense(500, activation='relu'))
model.add(Dense(num_classes, activation='softmax'))
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])#运行调参模块
#epochs是迭代的次数，暂定2次是为了迅速看结果，最后改成5次
model.fit(X_train, y_train, epochs = epochs, batch_size = batch_size, verbose=1, validation_split=0.05)
loss, accuracy = model.evaluate(X_test, y_test)#输出模块
print('Test loss:', loss)
print('Accuracy:', accuracy)
model.summary()
AlexNet
# -*- coding: utf-8 -*-
""" Created on Thu Jun 13 11:19:33 2019@author: YLC """
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Flatten
from keras.layers.convolutional import Conv2D, MaxPooling2D
from keras.layers.normalization import BatchNormalization
from keras.utils import np_utils
import tensorflow as tf#数据集导入模块
(X_train, y_train), (X_test, y_test) = mnist.load_data();#参数定义模块
img_rows, img_cols = 28,28# input dimensions
batch_size = 64
num_classes = 10
epochs = 5
img_shape = (img_rows,img_cols,1) #预处理 标准化模块
X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)
X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)
X_train = X_train / 255
X_test = X_test / 255#分类转One-Hot模块
y_train = np_utils.to_categorical(y_train, num_classes=10)
y_test = np_utils.to_categorical(y_test, num_classes=10)#网络搭建模块
model = Sequential()model.add(Conv2D(input_shape=img_shape, kernel_size=(11, 11), filters=96, activation='relu',strides= [1,1],padding= 'valid'))#卷积层
model.add(MaxPooling2D(pool_size=(3,3), strides=[2,2]))#池化层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(5,5), filters=256, activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(MaxPooling2D(pool_size=(3,3), strides=[2,2]))#池化层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(3,3), filters=384,activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(3,3), filters=384, activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(3,3), filters=256, activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(MaxPooling2D(pool_size=(3,3), strides=[2,2]))#池化层
model.add(BatchNormalization(axis= 1))model.add(Flatten())
model.add(Dense(4096, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(4096, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])# 训练配置，仅供参考
model.fit(X_train, y_train, epochs=epochs, batch_size= batch_size, verbose=1, validation_split=0.05)
loss, accuracy = model.evaluate(X_test, y_test)#输出模块
print('Test loss:', loss)
print('Accuracy:', accuracy)
model.summary()
ZFNet
# -*- coding: utf-8 -*-
""" Created on Thu Jun 13 11:19:33 2019@author: YLC """
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Flatten
from keras.layers.convolutional import Conv2D, MaxPooling2D
from keras.layers.normalization import BatchNormalization
from keras.utils import np_utils
import tensorflow as tf#数据集导入模块
(X_train, y_train), (X_test, y_test) = mnist.load_data();#参数定义模块
img_rows, img_cols = 28,28# input dimensions
batch_size = 64
num_classes = 10
epochs = 5
img_shape = (img_rows,img_cols,1) #预处理 标准化模块
X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)
X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)#分类转One-Hot模块
y_train = np_utils.to_categorical(y_train, num_classes=10)
y_test = np_utils.to_categorical(y_test, num_classes=10)#网络搭建模块
model = Sequential()model.add(Conv2D(input_shape=img_shape, kernel_size=(7,7), filters=96, activation='relu',strides= [1,1],padding= 'valid'))#卷积层
model.add(MaxPooling2D(pool_size=(3,3), strides=[2,2]))#池化层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(5,5), filters=256, activation='relu',strides= [2,2],padding= 'same'))#卷积层
model.add(MaxPooling2D(pool_size=(3,3), strides=[1,1]))#池化层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(3,3), filters=384,activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(3,3), filters=384, activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(BatchNormalization(axis= 1))model.add(Conv2D(kernel_size=(3,3), filters=256, activation='relu',strides= [1,1],padding= 'same'))#卷积层
model.add(MaxPooling2D(pool_size=(3,3), strides=[1,1]))#池化层
model.add(BatchNormalization(axis= 1))model.add(Flatten())
model.add(Dense(4096, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(4096, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])# 训练配置，仅供参考
model.fit(X_train, y_train, epochs=epochs, batch_size= batch_size, verbose=1, validation_split=0.05)
loss, accuracy = model.evaluate(X_test, y_test)#输出模块
print('Test loss:', loss)
print('Accuracy:', accuracy)
model.summary()