LeNet-5分析理解
文章目录
概要
LeNet-5：是Yann LeCun在1998年设计的用于手写体识别的经典卷积神经网络（CNN），是早期卷积神经网络中最有代表性的实验系统之一。
LeNet-5共有7层，主要有两个卷积层、两个下采样（池化层）、三个全连接层。
各层概况
（注：卷积后数据计算公式）
o
u
t
p
u
t
=
(
N
+
2
×
p
a
d
−
F
)
stride
+
1
output =\frac{(N+2 \times p a d-F)}{\text {stride}}+1
output=stride(N+2×pad−F)+1
LeNet-5第一层：卷积层C1
该层由6个5X5卷积核进行卷积，生成六个卷积图谱（feature map）。输入图像大小为32X32，则卷积后的特征图谱大小为[(32-5)/1+1]X[(32-5)/1+1]=28X28。由于卷积核大小为5X5则加上一个偏置参数（bias），共有5X5+1=26个参数，则C1层有26X6=156个训练参数。
with tf.variable_scope("C1-conv",reuse=resuse):
# 这行代码指定了第一个卷积层作用域为C1-conv，在这个作用域下有两个变量conv1_weights和conv1_biases
conv1_weights = tf.get_variable("weight", [5, 5, 1, 6],
initializer=tf.truncated_normal_initializer(stddev=0.1))
# tf.get_variable共享变量 stddev正太分布的标准差
conv1_biases = tf.get_variable("bias", [32], initializer=tf.constant_initializer(0.0))
# tf.constant_initializer初始化为常数
conv1 = tf.nn.conv2d(input_tensor, conv1_weights, strides=[1, 1, 1, 1], padding="SAME")
relu1 = tf.nn.relu(tf.nn.bias_add(conv1, conv1_biases))
LeNet-5第二层：池化层S2
对C1层的六个28x28的特征图谱分别进行以2x2为单位的下采样得到6个[（28-2）/2+1]x[（28-2）/2+1]=14x14的特征图谱。
with tf.name_scope("S2-max_pool",):
# tf.name_scope的主要目的是为了更加方便地管理参数命名。
# 与 tf.Variable() 结合使用。简化了命名
pool1 = tf.nn.max_pool(relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
# ksize：池化窗口的大小，取一个四维向量，一般是[1, height, width, 1]
# 因为我们不想在batch和channels上做池化，所以这两个维度设为了1
# strides：窗口在每一个维度上滑动的步长，一般也是[1, stride,stride, 1]
LeNet-5第三层：卷积层C3
使用16个5X5卷积核对S2层进行卷积，生成16个（14-5+10）X（14-5+1）=10X10的特征图谱。则该层共有(5X5+1)X16=416个训练参数。
with tf.variable_scope("C3-conv",reuse=resuse):
conv2_weights = tf.get_variable("weight", [5, 5, 32, 64],
initializer=tf.truncated_normal_initializer(stddev=0.1))
conv2_biases = tf.get_variable("bias", [64], initializer=tf.constant_initializer(0.0))
conv2 = tf.nn.conv2d(pool1, conv2_weights, strides=[1, 1, 1, 1], padding="VALID")
relu2 = tf.nn.relu(tf.nn.bias_add(conv2, conv2_biases))
LeNet-5第四层：池化层S4
对C3层的16个10X10的特征图谱分别进行以2x2为单位的下采样得到16个[（10-2）/2+1]x[（10-2）/2+1]=5x5的特征图谱。
with tf.name_scope("S4-max_pool",):
pool2 = tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
shape = pool2.get_shape().as_list()
# get_shape()函数可以得到这一层维度信息，由于每一层网络的输入输出都是一个batch的矩阵
nodes = shape[1] * shape[2] * shape[3]
# shape[1]是长度方向，shape[2]是宽度方向，shape[3]是深度方向
reshaped = tf.reshape(pool2, [shape[0], nodes])
# shape[0]是一个batch中数据的个数，reshape()函数原型reshape(tensor,shape,name)
LeNet-5第五层：卷积层C5
使用120个5X5卷积核对S4层进行卷积，由于S4的特征图谱大小与卷积核大小相同，会形成120个卷积结果，每个结果都与上一层的16个特征图谱相连。
with tf.variable_scope("layer5-full1",reuse=resuse):
Full_connection1_weights = tf.get_variable("weight", [nodes, 120],
initializer=tf.truncated_normal_initializer(stddev=0.1))
# if regularizer != None:
tf.add_to_collection("losses", regularizer(Full_connection1_weights))
Full_connection1_biases = tf.get_variable("bias", [120],
initializer=tf.constant_initializer(0.1))
if avg_class ==None:
Full_1 = tf.nn.relu(tf.matmul(reshaped, Full_connection1_weights) + \
Full_connection1_biases)
else:
Full_1 = tf.nn.relu(tf.matmul(reshaped, avg_class.average(Full_connection1_weights))
+ avg_class.average(Full_connection1_biases))
LeNet-5第六层：全连接层F6
该层为全连接层，有84个节点，采用了正切函数，计算公式为：
x
i
=
f
(
a
i
)
=
A
tanh
(
S
a
i
)
x_{i}=f\left(\mathrm{a}_{i}\right)=A \tanh \left(\mathrm{Sa}_{i}\right)
xi=f(ai)=Atanh(Sai)
with tf.variable_scope("layer6-full2",reuse=resuse):
Full_connection2_weights = tf.get_variable("weight", [120, 10],
initializer=tf.truncated_normal_initializer(stddev=0.1))
# if regularizer != None:
tf.add_to_collection("losses", regularizer(Full_connection2_weights))
Full_connection2_biases = tf.get_variable("bias", [10],
initializer=tf.constant_initializer(0.1))
if avg_class == None:
result = tf.matmul(Full_1, Full_connection2_weights) + Full_connection2_biases
else:
result = tf.matmul(Full_1, avg_class.average(Full_connection2_weights)) + \
avg_class.average(Full_connection2_biases)
LeNet-5第七层：全连接层output
该层为全连接层，有10个节点，分别对应0~9，采用的是径向基函数（RBF）的网络连接方式，计算公式为：
y
i
=
∑
j
(
x
j
−
w
i
j
)
2
y_{i}=\sum_{j}\left(x_{j}-w_{i j}\right)^{2}
yi=j∑(xj−wij)2
代码整合
import tensorflow as tf
import numpy as np
from tensorflow.examples.tutorials.mnist import input_data
# 下载数据集到当前目录，并且新建文件夹MNIST_data
# input_data.read_data_sets函数生成的类会自动将MNIST数据集划分为train, validation和test三个数据集
mnist = input_data.read_data_sets("./MNIST_data", one_hot=True)
batch_size = 100
learning_rate = 0.01
learning_rate_decay = 0.99
max_steps = 30000
def hidden_layer(input_tensor,regularizer,avg_class,resuse):
with tf.variable_scope("C1-conv",reuse=resuse):
# 这行代码指定了第一个卷积层作用域为C1-conv，在这个作用域下有两个变量conv1_weights和conv1_biases
conv1_weights = tf.get_variable("weight", [5, 5, 1, 6],
initializer=tf.truncated_normal_initializer(stddev=0.1))
# tf.get_variable共享变量 stddev正太分布的标准差
conv1_biases = tf.get_variable("bias", [32], initializer=tf.constant_initializer(0.0))
# tf.constant_initializer初始化为常数
conv1 = tf.nn.conv2d(input_tensor, conv1_weights, strides=[1, 1, 1, 1], padding="VALID")
relu1 = tf.nn.relu(tf.nn.bias_add(conv1, conv1_biases))
with tf.name_scope("S2-max_pool",):
# tf.name_scope的主要目的是为了更加方便地管理参数命名。
# 与 tf.Variable() 结合使用。简化了命名
pool1 = tf.nn.max_pool(relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
# ksize：池化窗口的大小，取一个四维向量，一般是[1, height, width, 1]
# 因为我们不想在batch和channels上做池化，所以这两个维度设为了1
# strides：窗口在每一个维度上滑动的步长，一般也是[1, stride,stride, 1]
with tf.variable_scope("C3-conv",reuse=resuse):
conv2_weights = tf.get_variable("weight", [5, 5, 32, 64],
initializer=tf.truncated_normal_initializer(stddev=0.1))
conv2_biases = tf.get_variable("bias", [64], initializer=tf.constant_initializer(0.0))
conv2 = tf.nn.conv2d(pool1, conv2_weights, strides=[1, 1, 1, 1], padding="VALID")
relu2 = tf.nn.relu(tf.nn.bias_add(conv2, conv2_biases))
with tf.name_scope("S4-max_pool",):
pool2 = tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
shape = pool2.get_shape().as_list()
# get_shape()函数可以得到这一层维度信息，由于每一层网络的输入输出都是一个batch的矩阵
nodes = shape[1] * shape[2] * shape[3]
# shape[1]是长度方向，shape[2]是宽度方向，shape[3]是深度方向
reshaped = tf.reshape(pool2, [shape[0], nodes])
# shape[0]是一个batch中数据的个数，reshape()函数原型reshape(tensor,shape,name)
with tf.variable_scope("layer5-full1",reuse=resuse):
Full_connection1_weights = tf.get_variable("weight", [nodes, 120],
initializer=tf.truncated_normal_initializer(stddev=0.1))
# if regularizer != None:
tf.add_to_collection("losses", regularizer(Full_connection1_weights))
Full_connection1_biases = tf.get_variable("bias", [120],
initializer=tf.constant_initializer(0.1))
if avg_class ==None:
Full_1 = tf.nn.relu(tf.matmul(reshaped, Full_connection1_weights) + \
Full_connection1_biases)
else:
Full_1 = tf.nn.relu(tf.matmul(reshaped, avg_class.average(Full_connection1_weights))
+ avg_class.average(Full_connection1_biases))
with tf.variable_scope("layer6-full2",reuse=resuse):
Full_connection2_weights = tf.get_variable("weight", [120, 10],
initializer=tf.truncated_normal_initializer(stddev=0.1))
# if regularizer != None:
tf.add_to_collection("losses", regularizer(Full_connection2_weights))
Full_connection2_biases = tf.get_variable("bias", [10],
initializer=tf.constant_initializer(0.1))
if avg_class == None:
result = tf.matmul(Full_1, Full_connection2_weights) + Full_connection2_biases
else:
result = tf.matmul(Full_1, avg_class.average(Full_connection2_weights)) + \
avg_class.average(Full_connection2_biases)
return result
# tf.placeholder(dtype, shape=None, name=None)
x = tf.placeholder(tf.float32, [batch_size ,28,28,1],name="x-input")
y_ = tf.placeholder(tf.float32, [None, 10], name="y-input")
# L2正则化是一种减少过拟合的方法
regularizer = tf.contrib.layers.l2_regularizer(0.0001)
# 调用定义的CNN的函数
y = hidden_layer(x,regularizer,avg_class=None,resuse=False)
# 定义存储训练轮数的变量
training_step = tf.Variable(0, trainable=False)
# tf.train.ExponentialMovingAverage是指数加权平均的求法
# 可以加快训练早期变量的更新速度。
variable_averages = tf.train.ExponentialMovingAverage(0.99, training_step)
variables_averages_op = variable_averages.apply(tf.trainable_variables())
average_y = hidden_layer(x,regularizer,variable_averages,resuse=True)
# 使用交叉熵作为损失函数。这里使用
# sparse_softmax_cross_entropy_with_logits函数来计算交叉熵。因为手写体是一个长度为
# 10的一维数组，而该函数需要提供的是一个正确答案的数字，所以需要使用tf.argmax函数来
# 得到正确答案对应的类别编号
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1))
# 计算在当前batch中所有样例的交叉熵平均值
cross_entropy_mean = tf.reduce_mean(cross_entropy)
# 总损失等于交叉熵损失和正则化损失的和
loss = cross_entropy_mean + tf.add_n(tf.get_collection('losses'))
# 设置指数衰减的学习率
learning_rate = tf.train.exponential_decay(learning_rate,# 基础的学习率，随着迭代的进行，更新变量时使用的学习率
training_step, mnist.train.num_examples /batch_size ,
learning_rate_decay, staircase=True)
# 使用tf.train.GradientDescentOptimizer优化算法来优化损失函数
train_step = tf.train.GradientDescentOptimizer(learning_rate). \
minimize(loss, global_step=training_step)
with tf.control_dependencies([train_step, variables_averages_op]):
train_op = tf.no_op(name='train')
crorent_predicition = tf.equal(tf.arg_max(average_y,1),tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(crorent_predicition,tf.float32))
# 初始化会话并开始训练过程
with tf.Session() as sess:
tf.global_variables_initializer().run()
for i in range(max_steps):
if i %1000==0:
x_val, y_val = mnist.validation.next_batch(batch_size)
reshaped_x2 = np.reshape(x_val, (batch_size,28,28, 1))
validate_feed = {x: reshaped_x2, y_: y_val}
validate_accuracy = sess.run(accuracy, feed_dict=validate_feed)
print("After %d trainging step(s) ,validation accuracy"
"using average model is %g%%" % (i, validate_accuracy * 100))
x_train, y_train = mnist.train.next_batch(batch_size)
reshaped_xs = np.reshape(x_train, (batch_size ,28,28,1))
sess.run(train_op, feed_dict={x: reshaped_xs, y_: y_train})
扫一扫
2万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
