import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
#载入数据集
mnist = input_data.read_data_sets("F:TensorflowProjectMNIST_data",one_hot=True)
# 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]))
#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
#lstm_cell = tf.contrib.rnn.core_rnn_cell.BasicLSTMCell(lstm_size)
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
#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()
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})
test_acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels})
print ("Iter " + str(epoch) + ", Testing Accuracy= " + str(test_acc ))
#############运行结果
Extracting F:TensorflowProjectMNIST_data rain-images-idx3-ubyte.gz Extracting F:TensorflowProjectMNIST_data rain-labels-idx1-ubyte.gz Extracting F:TensorflowProjectMNIST_data 10k-images-idx3-ubyte.gz Extracting F:TensorflowProjectMNIST_data 10k-labels-idx1-ubyte.gz WARNING:tensorflow:From <ipython-input-5-efe314a2932e>:40: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version. Instructions for updating: Future major versions of TensorFlow will allow gradients to flow into the labels input on backprop by default. See tf.nn.softmax_cross_entropy_with_logits_v2. Iter 0, Testing Accuracy= 0.7372 Iter 1, Testing Accuracy= 0.8334 Iter 2, Testing Accuracy= 0.908 Iter 3, Testing Accuracy= 0.9182 Iter 4, Testing Accuracy= 0.9172 Iter 5, Testing Accuracy= 0.9266
##############第二次运行
Iter 0 ,Testing Accuracy= 0.744 Iter 1 ,Testing Accuracy= 0.8028 Iter 2 ,Testing Accuracy= 0.8803 Iter 3 ,Testing Accuracy= 0.9115 Iter 4 ,Testing Accuracy= 0.9262 Iter 5 ,Testing Accuracy= 0.9303