深度学习框架TensorFlow学习与应用（七）——循环神经网络（RNN）应用于MNIST数据集分类

关于RNN的介绍可以参考：RNN介绍

下面将RNN用于前文所提到的MNIST手写数字识别中。

1.获取数据集

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("D:\BaiDu\MNIST_data",one_hot=True)

2.参数定义

# 输入图片是28*28
n_inputs = 28 #输入一行，一行有28个数据
max_time = 28 #一共28行
lstm_size = 100 #隐层单元
n_classes = 10 # 10个分类
batch_size = 50 #每批次50个样本
n_batch = mnist.train.num_examples // batch_size #计算一共有多少个批次

#这里的none表示第一个维度可以是任意的长度
x = tf.placeholder(tf.float32,[None,784])
#正确的标签
y = tf.placeholder(tf.float32,[None,10])

#初始化权值
weights = tf.Variable(tf.truncated_normal([lstm_size, n_classes], stddev=0.1))
#初始化偏置值
biases = tf.Variable(tf.constant(0.1, shape=[n_classes]))

3.定义RNN函数

#定义RNN网络
def RNN(X,weights,biases):
    # inputs=[batch_size, max_time, n_inputs]
    inputs = tf.reshape(X,[-1,max_time,n_inputs])
    #定义LSTM基本CELL
    from tensorflow.contrib import rnn 
    lstm_cell = rnn.BasicLSTMCell(lstm_size) 
    # final_state[0]是cell state
    # final_state[1]是hidden_state
    outputs,final_state = tf.nn.dynamic_rnn(lstm_cell,inputs,dtype=tf.float32)
    results = tf.nn.softmax(tf.matmul(final_state[1],weights) + biases)
    return results

4.优化与训练

#计算RNN的返回结果
prediction= RNN(x, weights, biases)  
#损失函数
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,labels=y))
#使用AdamOptimizer进行优化
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#结果存放在一个布尔型列表中
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))#argmax返回一维张量中最大的值所在的位置
#求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))#把correct_prediction变为float32类型
#初始化
init = tf.global_variables_initializer()

5.执行会话

with tf.Session() as sess:
    sess.run(init)
    for epoch in range(6):
        for batch in range(n_batch):
            batch_xs,batch_ys =  mnist.train.next_batch(batch_size)
            sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys})
        
        acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels})
        print ("Iter " + str(epoch) + ", Testing Accuracy= " + str(acc))

6.执行结果

Iter 0, Testing Accuracy= 0.733
Iter 1, Testing Accuracy= 0.8607
Iter 2, Testing Accuracy= 0.9006
Iter 3, Testing Accuracy= 0.9168
Iter 4, Testing Accuracy= 0.929
Iter 5, Testing Accuracy= 0.9374