运用神经网络LeNet-5实现对手写数字的分类。
原文标题：Gradient-Based Learning Applied to Document Recognition
采用反向传播算法训练的多层神经网络是一种成功的基于梯度的学习的最佳例子。给定一个适当的网络结构，基于梯度的学习算法可以用来合成一个复杂的决策面，该决策面只需很少的预处理就可以对高维模式（如手写字符）进行分类。
代码：
from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf
import numpy as np
import sys
import os
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
logs_train_dir = './Model'
def weight_variable(shape):
# 产生正态分布，标准差为0.1
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial)
def bias_variable(shape):
# 创建一个结构为shape矩阵也可以说是数组shape声明其行列，初始化所有值为0.1
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial)
def conv2d(x,W):
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
def inference(images):
# 第1层卷积
with tf.variable_scope('conv1'):
W_conv1 = tf.Variable(weight_variable([5, 5, 1, 6]), name="weight")
b_conv1 = tf.Variable(bias_variable([6]), name="bias")
h_conv1 = tf.nn.relu(conv2d(images, W_conv1) + b_conv1)
# print(np.shape(h_conv1))
# 第2层池化
with tf.variable_scope('pool2'):
h_pool2 = max_pool_2x2(h_conv1)
# print(np.shape(h_pool2))
# 第3层卷积
with tf.variable_scope('conv3'):
W_conv3 = tf.Variable(weight_variable([5, 5, 6, 16]), name="weight")
b_conv3 = tf.Variable(bias_variable([16]), name="bias")
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
# print(np.shape(h_conv3))
# 第4层池化
with tf.variable_scope('pool4'):
h_pool4 = max_pool_2x2(h_conv3)
# print(np.shape(h_pool4))
h_pool4_flat = tf.reshape(h_pool4, [-1, 7 * 7 * 16])
# 5\6\7全连接层
with tf.variable_scope('fc5'):
W_fc5 = weight_variable([7 * 7 * 16, 120])
b_fc5 = bias_variable([120])
h_fc5 = tf.nn.relu(tf.matmul(h_pool4_flat, W_fc5) + b_fc5)
with tf.variable_scope('fc6'):
W_fc6 = weight_variable([120, 84])
b_fc6 = bias_variable([84])
h_fc6 = tf.nn.relu(tf.matmul(h_fc5, W_fc6) + b_fc6)
with tf.variable_scope('out'):
W_out = weight_variable([84, 10])
b_out = bias_variable([10])
h_out = tf.nn.softmax(tf.matmul(h_fc6, W_out) + b_out)
# h_out_drop = tf.nn.dropout(h_out, 0.5)
return h_out
x = tf.placeholder(tf.float32, [None, 28*28])
y_ = tf.placeholder(tf.float32, [None, 10])
x_image = tf.reshape(x, [-1, 28, 28, 1])
y_conv = inference(x_image)
# 定义损失函数和学习步骤
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices=[1])) # 交叉熵
train_step = tf.train.AdamOptimizer(1e-3).minimize(cross_entropy) # 用Adam优化器
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
summary_op = tf.summary.merge_all()
sess = tf.InteractiveSession()
saver = tf.train.Saver()
sess.run(tf.global_variables_initializer())
train_writer = tf.summary.FileWriter(logs_train_dir, sess.graph)
for i in range(2000):
batch = mnist.train.next_batch(50)
if i % 100 == 0:
train_accuracy = accuracy.eval(feed_dict={x: batch[0], y_: batch[1]})
print("step %d, training accuracy %g" % (i, train_accuracy))
# summary_str = sess.run(summary_op)
# train_writer.add_summary(summary_str, i)
train_step.run(feed_dict={x: batch[0], y_: batch[1]})
checkpoint_path = os.path.join(logs_train_dir, 'thing.ckpt')
saver.save(sess, checkpoint_path)
print("test accuracy %g" % accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels}))
网友评论