人工智能学习（2）MNIST手写识别进阶

接下来开始接触深度学习的内容----卷积神经网络（CNN）
通过加入卷积网络来改善单一softmax回归、随机梯度下降预测的结果

import tensorflow as tf
from tensorflow_core.examples.tutorials.mnist import input_data

def weight_variable(shape):
	#标准差为0.1的随机正态分布
    inital = tf.truncated_normal(shape,stddev=0.1)
    return tf.Variable(inital)
def bias_weight(shape):
	#创建一个常量tensor，按照给出value=0.1来赋值，可以用shape来指定其形状
    inital = tf.constant(0.1,shape=shape)
    return tf.Variable(inital)
def conv2d(x,w):
    return tf.nn.conv2d(x,w,strides=[1,1,1,1],padding='SAME')
def max_pool_2x2(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('mnist_dataset/', one_hot= True)
#灵活构件代码，能够在运行图时，插入计算图
sess = tf.InteractiveSession()
#占位符
x = tf.placeholder('float',[None,784])
y_ = tf.placeholder('float',[None,10])
#变量
w = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))
#第一层卷积、池化
w_conv1 = weight_variable([5,5,1,32])
b_conv1 = bias_weight([32])
#将原来图片改为中间两维表示对应图片的宽高，最后一维代表图片颜色通道数
x_image = tf.reshape(x,[-1,28,28,1])

h_conv1 = tf.nn.relu(conv2d(x_image,w_conv1)+b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

#第二层卷积、池化
w_conv2 = weight_variable([5,5,32,64])
b_conv2 = bias_weight([64])

h_conv2 = tf.nn.relu(conv2d(h_pool1,w_conv2)+b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
#全连接层
#经过[1,2,2,1]两个池化，图片变为28*28*1->14*14*32->7*7*64
w_fc1 = weight_variable([7*7*64,1024])
b_fc1 = bias_weight([1024])

h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,w_fc1)+b_fc1)

#dropout
keep_prob = tf.placeholder('float')
h_fc1_drop = tf.nn.dropout(h_fc1,keep_prob)

#输出层
w_fc2 = weight_variable([1024,10])
b_fc2 = bias_weight([10])
#softmax
y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop,w_fc2)+b_fc2)
#交叉熵
cross_entropy = -tf.reduce_sum(y_*tf.log(y_conv))
#ADAM梯度下降
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y_,1))
# tf.cast 将correct_prediction转化为“float”类型
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))

#变量初始化
sess.run(tf.initialize_all_variables())

#训练
for i in range(20000):
    batch = mnist.train.next_batch(50)
    if i%100 == 0:
        train_accuray = accuracy.eval(feed_dict={x:batch[0],y_:batch[1],keep_prob:1.0})
        print('step %d, training accuracy %g'%(i,train_accuray))
    train_step.run(feed_dict={x:batch[0],y_:batch[1],keep_prob:0.5})

print('test accuray %g'%accuracy.eval(feed_dict={x:mnist.test.images,y_:mnist.test.labels,keep_prob:1.0}))