tensoflow练习2：利用Recurrent Neural Network 进行分类

快一周没写博客了，总觉得缺少点什么。最近忙着写论文、改论文也没什么时间写。好了，不废话了，直接上干货，利用RNN进行分类。RNN主要用于序列数据的处理，在图像、语音、文本等领域有着广泛的应用。这里使用RNN对手写体进行识别（0-9共10类）。
代码如下：

#coding=utf-8
import tensorflow as tf
import numpy as numpy
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data/',one_hot=True)#下载数据，有点慢

#一张图片是28*28，FNN一次将数据输入到网络，RNN将它分成快
chunk_size = 28
chunk_n = 28

rnn_size = 256#隐层尺寸（维度）
#输出层
n_output_layer = 10 #输出层
X = tf.placeholder('float',[None,chunk_n,chunk_size])
Y = tf.placeholder('float')

def recurrent_neural_network(data):
    layer = {'w_':tf.Variable(tf.random_normal([rnn_size,n_output_layer])), 'b_':tf.Variable(tf.random_normal([n_output_layer]))}
    lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(rnn_size)
    data = tf.transpose(data,[1,0,2])
    data = tf.reshape(data,[-1,chunk_size])
    data = tf.split(0,chunk_n,data)
    outputs, status = tf.nn.rnn(lstm_cell,data,dtype=tf.float32)
    output = tf.add(tf.matmul(outputs[-1],layer['w_']),layer['b_'])
    return output

#每一次100条数据
batch_size = 100
def train_neural_network(X,Y):
    predict = recurrent_neural_network(X)
    cost_func = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(predict,Y))
    optimizer = tf.train.AdamOptimizer().minimize(cost_func)

    epochs = 13
    with tf.Session() as session:
        session.run(tf.global_variables_initializer())
        epoch_loss =0
        print('training begins:')
        for epoch in range(epochs):
            for i in range( int(mnist.train.num_examples/batch_size)):
                x,y = mnist.train.next_batch(batch_size)
                x = x.reshape([batch_size,chunk_n,chunk_size])
                _,c =session.run([optimizer,cost_func],feed_dict={X:x,Y:y})
                epoch_loss += c
            print(epoch,':',epoch_loss)
        correct = tf.equal(tf.argmax(predict,1),tf.argmax(Y,1))
        accuracy = tf.reduce_mean(tf.cast(correct,'float'))
        print('精确率:',accuracy.eval({X:mnist.test.images.reshape(-1,chunk_n,chunk_size),Y:mnist.test.labels}))

train_neural_network(X,Y)

接下来，我对上述代码进行分析；
数据准备，就是最上面的下载代码。
（1）定义数据与标签类型；

X = tf.placeholder('float',[None,chunk_n,chunk_size])
Y = tf.placeholder('float')
n_output_layer = 10 #输出类数

其中，None表示不确定，可随机自定。
（2）定义rnn:

def recurrent_neural_network(data):
    layer = {'w_':tf.Variable(tf.random_normal([rnn_size,n_output_layer])), 'b_':tf.Variable(tf.random_normal([n_output_layer]))}
    lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(rnn_size)#定义神经元LSTM
    data = tf.transpose(data,[1,0,2])
    data = tf.reshape(data,[-1,chunk_size])
    data = tf.split(0,chunk_n,data)
    outputs, status = tf.nn.rnn(lstm_cell,data,dtype=tf.float32)
    output = tf.add(tf.matmul(outputs[-1],layer['w_']),layer['b_'])
    return output

layer:是对输出进行映射的参数。
比较难理解的就是三行data了。其实不用纠结。这三行的目的是将3维张量转换成chunk_n个形状为batch_size * chunk_size的张量。
outpus[-1]表示最后的隐层表示。

（3）定义网络训练

def train_neural_network(X,Y):
    predict = recurrent_neural_network(X)#预测
    cost_func = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(predict,Y))#损失函数
    optimizer = tf.train.AdamOptimizer().minimize(cost_func)#优化器

    epochs = 13
    with tf.Session() as session:
        session.run(tf.global_variables_initializer())
        epoch_loss =0
        print('training begins:')
        for epoch in range(epochs):
            for i in range( int(mnist.train.num_examples/batch_size)):#一共多少个batch
                x,y = mnist.train.next_batch(batch_size)
                x = x.reshape([batch_size,chunk_n,chunk_size])#重新塑形
                _,c =session.run([optimizer,cost_func],feed_dict={X:x,Y:y})
                epoch_loss += c
            print(epoch,':',epoch_loss)
        correct = tf.equal(tf.argmax(predict,1),tf.argmax(Y,1))
        accuracy = tf.reduce_mean(tf.cast(correct,'float'))
        print('精确率:',accuracy.eval({X:mnist.test.images.reshape(-1,chunk_n,chunk_size),Y:mnist.test.labels}))

X,Y相当于盒子，容器，不要纠结太多。loss为累加损失

（4）输出结果：