首先加载必要模块和数据
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
设置一些常量
INPUT_NODE = 784 #输入层的节点数。对于MNIST数据集，这个就等于图片的像素。
OUTPUT_NODE = 10 #输出层的节点数。这个等于类别的数目。因为在MNIST数据集中需要区分的事0-9，所以这里输出层的节点数为10。
LAYER1_NODE = 500 #只有一个带有500个节点的隐藏层
BATCH_SIZE = 100 #定义batch的大小
LEARNING_RATE_BASE = 0.8 #基础的学习率
LEARNING_RATE_DECAY = 0.99 #学习率的衰减率
REGULARIZATION_RATE = 0.0001 #描述模型复杂度的正则化项在损失函数中的系数
TRAINING_STEPS = 30000 #训练轮数
MOVING_AVERAGE_DECAY = 0.99 #滑动平均距离
一个辅助训练的函数
def inference(input_tensor,avg_class,reuse = False):
#当没有提供滑动平均类时，直接使用参数当前的取值
#这里实际含义是： avg_class == None 时，是训练时的前向传播过程，else时是为了在测试时计算准确里用的
#在复用之前训练的好的模型时，可以直接使reuse为True
if avg_class == None:
with tf.variable_scope('layer1',reuse = reuse):
weights = tf.get_variable(name = 'weights',\
initializer = tf.truncated_normal(stddev=0.1,shape = [INPUT_NODE,LAYER1_NODE]))
biases = tf.get_variable(name = 'biases',shape=[LAYER1_NODE],\
initializer = tf.constant_initializer(0.0))
layer1 = tf.nn.relu(tf.matmul(input_tensor,weights)+biases)
with tf.variable_scope('layer2',reuse = reuse):
weights = tf.get_variable(name = 'weights',\
initializer = tf.truncated_normal(stddev=0.1,shape = [LAYER1_NODE,OUTPUT_NODE]))
biases = tf.get_variable(name = 'biases',shape=[OUTPUT_NODE],\
initializer = tf.constant_initializer(0.0))
return tf.matmul(layer1,weights)+biases
else:
with tf.variable_scope('layer1',reuse = reuse):
weights = tf.get_variable(name = 'weights',\
initializer = tf.truncated_normal(stddev=0.1,shape = [INPUT_NODE,LAYER1_NODE]))
biases = tf.get_variable(name = 'biases',shape=[LAYER1_NODE],\
initializer = tf.constant_initializer(0.0))
layer1 = tf.nn.relu(tf.matmul(input_tensor,avg_class.average(weights))+avg_class.average(biases))
with tf.variable_scope('layer2',reuse = reuse):
weights = tf.get_variable(name = 'weights',\
initializer = tf.truncated_normal(stddev=0.1,shape = [LAYER1_NODE,OUTPUT_NODE]))
biases = tf.get_variable(name = 'biases',shape=[OUTPUT_NODE],\
initializer = tf.constant_initializer(0.0))
return tf.matmul(layer1,avg_class.average(weights))+biases
定义训练过程
def train(mnist):
x = tf.placeholder(tf.float32,[None,INPUT_NODE],name='x-input')
y_ = tf.placeholder(tf.float32,[None,OUTPUT_NODE],name='y-input')
#truncated_normal生成正太分布值
#隐藏层参数
#weights1 = tf.Variable(tf.truncated_normal([INPUT_NODE,LAYER1_NODE],stddev=0.1))
#biases1 = tf.Variable(tf.constant(0.1,shape=[LAYER1_NODE]))
#输出层参数
#weights2 = tf.Variable(tf.truncated_normal([LAYER1_NODE,OUTPUT_NODE],stddev=0.1))
#biases2 = tf.Variable(tf.constant(0.1,shape=[OUTPUT_NODE]))
#计算未使用滑动平均一次前向传播结果
y = inference(x,None)
#定义当前步数，移动平均时会用到，自动更新+1
global_step = tf.Variable(0,trainable = False)
#计算使用滑动平均的前向传播结果
variable_averages = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY,global_step)
variables_averages_op = variable_averages.apply(tf.trainable_variables())
average_y = inference(x,variable_averages,True)
test_y = inference(x,None,True)
#在前向传播过后计算交叉熵
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=test_y,labels=tf.argmax(y_,1))
#交叉熵平均值
cross_entropy_mean = tf.reduce_mean(cross_entropy)
#正则项
regularizer=tf.contrib.layers.l2_regularizer(REGULARIZATION_RATE)
with tf.variable_scope("",reuse = True):
regularization = regularizer(tf.get_variable("layer1/weights"))+regularizer(tf.get_variable("layer2/weights"))
#损失等于交叉商加上正则项
loss = cross_entropy_mean + regularization
#定义学习率
learning_rate = tf.train.exponential_decay(LEARNING_RATE_BASE,#基础学习速率
global_step, #当前迭代轮数
mnist.train.num_examples/BATCH_SIZE, #总共需要的迭代次数
LEARNING_RATE_DECAY) #学习率衰减速率
#训练过程
train_step = tf.train.GradientDescentOptimizer(learning_rate)\
.minimize(loss,global_step = global_step)
#反向传播和滑动平均更新参数，这里直接实现了前向及逆向传播过程，在利用滑动平均更新参数的一整个过程
#with tf.control_dependencies([train_step,variables_averages_op]):
#train_op = tf.no_op(name='train')
train_op = tf.group(train_step,variables_averages_op)
correct_prediction = tf.equal(tf.argmax(average_y,1),tf.argmax(y_,1))
#计算出准确度，此处将bool转换成0，1，再用reduce_mean算1占的比例就可以得出准确度，可用一下注释代码验证
#tmp = tf.Variable([True,False,True])
#tmp1 = tf.cast(tmp,dtype=tf.float32)
#with tf.Session() as sess1:
#tf.global_variables_initializer().run()
#print(sess1.run(tf.reduce_mean(sess1.run(tmp1))))
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
with tf.Session() as sess:
#验证数据
tf.global_variables_initializer().run()
validate_feed = {x:mnist.validation.images,y_:mnist.validation.labels}
#测试数据
test_feed = {x:mnist.test.images,y_:mnist.test.labels}
for i in range(TRAINING_STEPS):
if i%1000 == 0:
validate_acc = sess.run(accuracy,feed_dict=validate_feed)
print("after %d training step(s),validation accuracy " "using average model is %g" %(i,validate_acc))
xs,ys = mnist.train.next_batch(BATCH_SIZE)
sess.run(train_op,feed_dict={x:xs,y_:ys})
test_acc = sess.run(accuracy,feed_dict=test_feed)
print("after %d training step(s),validation accuracy " "using average model is %g" %(TRAINING_STEPS,test_acc))
print()
开始执行
mnist = input_data.read_data_sets("/Users/zhouzelun/Documents/python/mnist_data",one_hot=True)
train(mnist)