from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf
import os
import tensorflow.contrib.slim as slim
def weight_variable(shape):
return tf.Variable(tf.truncated_normal(shape,stddev=0.1))
def bias_variable(shape):
return tf.Variable(tf.constant(0.0,shape=shape))
def conv2d(x, W):
return tf.nn.conv2d(x, W, strides = [1, 1, 1, 1], padding = 'SAME')
def max_pool_2_2(x):
return tf.nn.max_pool(x, ksize = [1, 2, 2, 1], strides = [1, 2, 2, 1], padding = 'SAME')
mnist = input_data.read_data_sets('/newSoftware/Sublime Text 3/Data/myCode/MNIST-data/',one_hot=True)
#这个函数用于设置默认参数值，对指定列表设置stride和padding的默认值
#这个模块可以精简代码
with slim.arg_scope([slim.conv2d,slim.max_pool2d,slim.avg_pool2d],stride=1,padding='SAME'):
x = tf.placeholder(tf.float32, [None, 784])
y_ = tf.placeholder(tf.float32, [None, 10])
x_image = tf.reshape(x, [-1, 28, 28, 1])
#以下注释是原来做法，现在改为使用slim模块实现，池化层步长为2不是默认值，另外设置一下就好
#W_conv1 = weight_variable([5, 5, 1, 32])
#b_conv1 = bias_variable([32])
#h_conv1=tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)#28*28*32卷积层不改变图片像素大小，只是用一个窗口去做内积
#h_pool1=max_pool_2_2(h_conv1)#池化层根据窗口和步长改变图片大小，但不改变特征维度32.窗口2*2，步长2，即输出14*14*32
h_conv1 = slim.conv2d(x_image,32,[5,5])
h_pool1 = slim.max_pool2d(h_conv1,[2,2],stride=2)
#下面就不一一使用了
W_conv2 = weight_variable([5, 5, 32, 64])#每个卷积核的通道深度32跟输入深度一致
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)#14*14*64同理，卷积层不改变大小，只是调节特征维度
h_pool2 = max_pool_2_2(h_conv2)#进一步缩小尺寸为7*7*64
W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])
h_pool2_flat = tf.reshape(h_pool2, [-1, 7 * 7 * 64])#压缩成二维
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)
#dropout选在第一个全连接后，为何？
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)#随机去掉百分比的训练数据，防止过拟合
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
#此时不用激活函数，softmax转为概率分布
y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)#（batch_size-dropout）*10
#reduction_indices=1表示在行上取最大值，得到batch_size*1，再去这batch_size个交叉熵的平均作为这一batch_size的交叉熵
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices = [1]))
global_step=tf.Variable(0,trainable=False)
#滑动平均，维护影子变量，利于测试验证
variable_averages=tf.train.ExponentialMovingAverage(0.99,global_step)
variable_averages_op=variable_averages.apply(tf.trainable_variables())
#这个优化器一般更好
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy,global_step=global_step)
#控制顺序
with tf.control_dependencies([train_step,variable_averages_op]):
train_op=tf.no_op(name='train')#什么也不做，就一个控制效果
#布尔值再转为浮点值，得到精度
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
#用于保存模型，下次可以直接调出测试或接着训练，这个东西有时候放的位置不对会报错，好像最好放在变量之后
saver=tf.train.Saver()
with tf.Session() as sess:
tf.global_variables_initializer().run()
for i in range(3000):
batch=mnist.train.next_batch(100)
_,entropy_value,train_accuracy,step=sess.run([train_op,cross_entropy,accuracy,global_step],feed_dict={x: batch[0], y_: batch[1],keep_prob:0.5})
if i%1000==0:
print("step %d, cross_entropy %g, training accuracy %g"%(step,entropy_value, train_accuracy))
saver.save(sess,os.path.join('/tmp/','model2.ckpt'),global_step=global_step,write_meta_graph=True)
print("test accuracy %g"%accuracy.eval(feed_dict = {x: mnist.test.images, y_: mnist.test.labels,keep_prob:1.0}))
--------------------------------------------------------------------------------------------------------------------------------