Tensorflow卷积神经网络

'''
##卷积神经网络，两个卷积层：32和64个特征平面，两个全连接层1024和10个神经元
'''
#加载数据，设定batch
mnist = input_data.read_data_sets('MNIST_data/',one_hot=True)
batch_size = 100
n_batch = mnist.train.num_examples // batch_size

#初始化权值
def weight_var(shape):
    return tf.Variable(tf.truncated_normal(shape,stddev=0.1))

#初始化偏移值
def bias_var(shape):
    return tf.Variable(tf.constant(0.1,shape=shape))

#卷积操作
def conv2d(x,W):
    return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding='SAME')

#池化操作
def max_pool_2x2(x):
    return tf.nn.max_pool(x,strides=[1,2,2,1],ksize=[1,2,2,1],padding='SAME')

#定义三个占位符，数据，标签和dropout
x = tf.placeholder(tf.float32,shape=[None,784])
y = tf.placeholder(tf.float32,shape=[None,10])
keep_prob = tf.placeholder(tf.float32)

#把x变成一个4d向量，其第2、第3维对应图片的宽、高，最后一维代表图片的颜色通道数
x_image = tf.reshape(x,[-1,28,28,1])

#卷积层1，32个特征平面，卷积操作后[-1，28,28,32]，池化操作后[-1，14,14,32]
W_conv1 = weight_var([5,5,1,32])
b_conv1 = bias_var([32])
h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1)+b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

#卷积层2，64个特征平面，卷积操作后[-1，14,14,64]，池化操作后[-1，7,7,64]
W_conv2 = weight_var([5,5,32,64])
b_conv2 = bias_var([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1,W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)

#全连接层1，1024个神经元，先将卷积层2的输出扁平化处理
h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64])
W_fc1 = weight_var([7*7*64,1024])
b_fc1 = bias_var([1024])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1) + b_fc1)
h_fc1 = tf.nn.dropout(h_fc1,keep_prob)

#全连接层2,10个神经元
W_fc2 = weight_var([1024,10])
b_fc2 = bias_var([10])
y_ = tf.nn.softmax(tf.matmul(h_fc1,W_fc2)+b_fc2)

#交叉熵损失函数，Adam优化器
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=y_))
train = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)

#准确率
correct = tf.equal(tf.argmax(y,1),tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct,tf.float32))

#变量初始化
init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)
    for iteration in range(21):
        for batch in range(n_batch):
            train_xs,train_ys = mnist.train.next_batch(batch_size)
            sess.run(train,feed_dict={x:train_xs,y:train_ys,keep_prob:0.5})
            
        print('iter: ',iteration,'accuracy: ',sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels,keep_prob:1}))