卷积神经网络

import tensorflow as tfw
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

def compute_accuracy(v_xs,v_ys):
        global prediction
        #修改方法：batch_target = batch_target.astype(np.float32)
        #v_xs = v_xs.astype(np.float32)
        y_pre = sess.run(prediction,feed_dict = {xs:v_xs,keep_prob:1})
        correct_prediction = tfw.equal(tfw.argmax(y_pre,1),tfw.argmax(v_ys,1))
        accuracy = tfw.reduce_mean(tfw.cast(correct_prediction,tfw.float32))
        result = sess.run(accuracy,feed_dict ={xs:v_xs,ys:v_ys,keep_prob:1})
        return result
    
    
def weight_variable(shape):
    inital = tfw.truncated_normal(shape,stddev = 0.1)
    return tfw.Variable(inital)

def bias_variable(shape):
    inital = tfw.constant(0.1,shape =shape)
    return tfw.Variable(inital)

def conv2d(x,W):
    #stride[1,x_movement,y_movement,1]
    #Must have strides[0] = strides[3] = 1
    return tfw.nn.conv2d(x,W,strides = [1,1,1,1],padding = 'SAME')

def max_pool_2x2(x):
    #stride[1,x_movement,y_movement,1]
    return tfw.nn.max_pool(x,ksize = [1,2,2,1],strides = [1,2,2,1],padding = 'SAME')

#define placeholder for inputs to network
xs= tfw.placeholder(tfw.float32,[None,784])  #28*28
ys= tfw.placeholder(tfw.float32,[None,10])  
keep_prob = tfw.placeholder(tfw.float32)
x_image = tfw.reshape(xs,[-1,28,28,1])

##conv1 layer##
W_conv1 = weight_variable([5,5,1,32])#patch 5*5,in size 1,out size 32
b_conv1 = bias_variable([32])
h_conv1 = tfw.nn.relu(conv2d(x_image,W_conv1)+b_conv1)#output size 28*28*32
h_pool1 = max_pool_2x2(h_conv1)                       #output size 14*14*32

##conv2 layer##
W_conv2 = weight_variable([5,5,32,64])#patch 5*5,in size 32,out size 64
b_conv2 = bias_variable([64])
h_conv2 = tfw.nn.relu(conv2d(h_pool1,W_conv2)+b_conv2)#output size 14*14*64
h_pool2 = max_pool_2x2(h_conv2)                       #output size 7*7*64

##func1 layer##
W_fc1 = weight_variable([7*7*64,1024])
b_fc1 = bias_variable([1024])
#[n_samples,7,7,64]->>[n_samples,7*7*64]
h_pool2_flat = tfw.reshape(h_pool2,[-1,7*7*64])
h_fc1 = tfw.nn.relu(tfw.matmul(h_pool2_flat,W_fc1) + b_fc1)
h_fc1_drop = tfw.nn.dropout(h_fc1,keep_prob)

##func2 layer##
W_fc2 = weight_variable([1024,10])
b_fc2 = bias_variable([10])
prediction = tfw.nn.softmax(tfw.matmul(h_fc1_drop,W_fc2) + b_fc2)

#the error between prediction and real data
cross_entropy =tfw.reduce_mean(-tfw.reduce_sum(ys*tfw.log(prediction),reduction_indices = [1]))  #loss
train_step = tfw.train.AdamOptimizer(1e-4).minimize(cross_entropy)                       

sess =tfw.Session()
#important step
sess.run(tfw.initialize_all_variables())

for i in range(1000):
    batch_xs,batch_ys = mnist.train.next_batch(100)
    #batch_target = batch_target.astype(np.float32)
    batch_xs = batch_xs.astype(np.float32)
    batch_ys = batch_ys.astype(np.float32)
    sess.run(train_step,feed_dict = {xs:batch_xs,ys:batch_ys, keep_prob: 0.5})
    if i%50 == 0:
        print(compute_accuracy(mnist.test.images,mnist.test.labels))   

Extracting MNIST_data\train-images-idx3-ubyte.gz
Extracting MNIST_data\train-labels-idx1-ubyte.gz
Extracting MNIST_data\t10k-images-idx3-ubyte.gz
Extracting MNIST_data\t10k-labels-idx1-ubyte.gz
0.1154
0.785
0.8822
0.912
0.9223
0.9353
0.9417
0.9452
0.9496
0.9559
0.9538
0.9574
0.9592
0.9616
0.96
0.9661
0.9672
0.9687
0.9697
0.9685

错误：
InvalidArgumentError: You must feed a value for placeholder tensor ‘Placeholder_2’ with dtype float
[[node Placeholder_2 (defined at :40) ]]
分析：
sess.run(train_step, feed_dict = {xs: batch_xs, ys: batch_ys}) 中 feed_dict 缺少参数
解决方式：
补充 keep_prob:0.5，如下所示sess.run(train_step, feed_dict = {xs: batch_xs, ys: batch_ys, keep_prob: 0.5})