神经网络实现Mnist手写数字识别笔记1.Mnist手写数字识别介绍2.神经网络架构介绍3.代码实现4.运行结果5.代码中部分方法介绍

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
from tensorflow.examples.tutorials.mnist import input_data
#获取手写字体识别数据
mnist = input_data.read_data_sets('data/', one_hot=True)

#网络架构定义
#第一个隐层有256个神经元
n_hidden_1 = 256 
#第二个隐层有128个神经元
n_hidden_2 = 128
#输入的样本有784个特征 
n_input    = 784
#手写字分为0 1 2....9这10个类别
n_classes  = 10  

# 输入和输出
x = tf.placeholder("float", [None, n_input])
y = tf.placeholder("float", [None, n_classes])
    
# 各层网络参数定义
stddev = 0.1
weights = {
    'w1': tf.Variable(tf.random_normal([n_input, n_hidden_1], stddev=stddev)),
    'w2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2], stddev=stddev)),
    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes], stddev=stddev))
}
biases = {
    'b1': tf.Variable(tf.random_normal([n_hidden_1])),
    'b2': tf.Variable(tf.random_normal([n_hidden_2])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}

#前向传播计算
def multilayer_perceptron(_X, _weights, _biases):
    #根据输入_X计算第一层
    layer_1 = tf.nn.sigmoid(tf.add(tf.matmul(_X, _weights['w1']), _biases['b1']))
    #以第一层layer_1作为输入计算第二层
    layer_2 = tf.nn.sigmoid(tf.add(tf.matmul(layer_1, _weights['w2']), _biases['b2']))
    #以第二层layer_2作为输入计算第三层，也就是计算输出层
    return (tf.matmul(layer_2, _weights['out']) + _biases['out'])

# 前向传播进行预测
pred = multilayer_perceptron(x, weights, biases)

#求损失函数
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))
#使用随机梯度下降进行优化
optm = tf.train.GradientDescentOptimizer(learning_rate=0.001).minimize(cost)
#预测值和真实标签进行对比，tf.argmax(pred, 1)返回的是pred中最大值的索引号
#tf.equal将预测的值和标签值进行比较，相同为True,不相同为False
corr = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
#tf.cast为类型转换函数，将corr转换为float类型数据 
accr = tf.reduce_mean(tf.cast(corr, "float"))

# INITIALIZER
init = tf.global_variables_initializer()

#训练200个epoch
training_epochs = 200
#每一个batch取100个样本
batch_size      = 100
display_step    = 4
# LAUNCH THE GRAPH
sess = tf.Session()
sess.run(init)
# 遍历epoch
for epoch in range(training_epochs):
    avg_cost = 0.
    #计算batch次数
    total_batch = int(mnist.train.num_examples/batch_size)
    # 遍历batch
    for i in range(total_batch):
        batch_xs, batch_ys = mnist.train.next_batch(batch_size)
        feeds = {x: batch_xs, y: batch_ys}
        sess.run(optm, feed_dict=feeds)
        avg_cost += sess.run(cost, feed_dict=feeds)
    avg_cost = avg_cost / total_batch
    # DISPLAY
    if (epoch+1) % display_step == 0:
        print ("Epoch: %03d/%03d cost: %.9f" % (epoch, training_epochs, avg_cost))
        feeds = {x: batch_xs, y: batch_ys}
        train_acc = sess.run(accr, feed_dict=feeds)
        print ("TRAIN ACCURACY: %.3f" % (train_acc))
        feeds = {x: mnist.test.images, y: mnist.test.labels}
        test_acc = sess.run(accr, feed_dict=feeds)
        print ("TEST ACCURACY: %.3f" % (test_acc))
print ("OPTIMIZATION FINISHED")

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Epoch: 003/200 cost: 2.278398191
TRAIN ACCURACY: 0.260
TEST ACCURACY: 0.198
Epoch: 007/200 cost: 2.245890754
.......
Epoch: 191/200 cost: 0.457598143
TRAIN ACCURACY: 0.920
TEST ACCURACY: 0.883
Epoch: 195/200 cost: 0.452329346
TRAIN ACCURACY: 0.860
TEST ACCURACY: 0.883
Epoch: 199/200 cost: 0.447292094
TRAIN ACCURACY: 0.910
TEST ACCURACY: 0.884
OPTIMIZATION FINISHED