Code:
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)
operation result:
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
Code:
# size of each batch
batch_size = 100
# Calculate how many batches there are
n_batch = mnist.train.num_examples // batch_size
#Parameter summary
def variable_summaries(var):
with tf.name_scope('summaries'):
mean = tf.reduce_mean(var)
tf.summary.scalar('mean', mean)#average
with tf.name_scope('stddev'):
stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
tf.summary.scalar('stddev', stddev)#标准差
tf.summary.scalar('max', tf.reduce_max(var))#maximum
tf.summary.scalar('min', tf.reduce_min(var))#minimum
tf.summary.histogram('histogram', var)#histogram
#initialize weights
def weight_variable(shape,name):
initial = tf.truncated_normal(shape,stddev=0.1)#Generate a truncated normal distribution
return tf.Variable(initial,name=name)
#initialize bias
def bias_variable(shape,name):
initial = tf.constant(0.1,shape=shape)
return tf.Variable(initial,name=name)
# convolutional layer
def conv2d(x,W):
#x input tensor of shape `[batch, in_height, in_width, in_channels]`
#W filter / kernel tensor of shape [filter_height, filter_width, in_channels, out_channels]
#`strides[0] = strides[3] = 1`. strides[1] represents the step size in the x direction, and strides[2] represents the step size in the y direction
#padding: A `string` from: `"SAME", "VALID"`
return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding='SAME')
#pooling layer
def max_pool_2x2(x):
#ksize [1,x,y,1]
return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')
#Namespaces
with tf.name_scope('input'):
#Define two placeholders
x = tf.placeholder(tf.float32,[None,784],name='x-input')
y = tf.placeholder(tf.float32,[None,10],name='y-input')
with tf.name_scope('x_image'):
#Change the format of x to a 4D vector [batch, in_height, in_width, in_channels]`
x_image = tf.reshape(x,[-1,28,28,1],name='x_image')
with tf.name_scope('Conv1'):
#Initialize the weights and biases of the first convolutional layer
with tf.name_scope('W_conv1'):
W_conv1 = weight_variable([5,5,1,32],name='W_conv1')#5*5 sampling window, 32 convolution kernels extract features from 1 plane
with tf.name_scope('b_conv1'):
b_conv1 = bias_variable([32],name='b_conv1')#A bias value for each convolution kernel
#Convolve x_image with the weight vector, add the bias value, and then apply it to the relu activation function
with tf.name_scope('conv2d_1'):
conv2d_1 = conv2d(x_image,W_conv1) + b_conv1
with tf.name_scope('relu'):
h_conv1 = tf.nn.relu(conv2d_1)
with tf.name_scope('h_pool1'):
h_pool1 = max_pool_2x2(h_conv1)#进行max-pooling
with tf.name_scope('Conv2'):
#Initialize the weights and biases of the second convolutional layer
with tf.name_scope('W_conv2'):
W_conv2 = weight_variable([5,5,32,64],name='W_conv2')#5*5 sampling window, 64 convolution kernels extract features from 32 planes
with tf.name_scope('b_conv2'):
b_conv2 = bias_variable([64],name='b_conv2')#A bias value for each convolution kernel
#Convolve h_pool1 with the weight vector, add the bias value, and then apply it to the relu activation function
with tf.name_scope('conv2d_2'):
conv2d_2 = conv2d(h_pool1,W_conv2) + b_conv2
with tf.name_scope('relu'):
h_conv2 = tf.nn.relu(conv2d_2)
with tf.name_scope('h_pool2'):
h_pool2 = max_pool_2x2(h_conv2)# for max-pooling
The image of #28*28 is still 28*28 after the first convolution, and becomes 14*14 after the first pooling
#After the second convolution, it is 14*14, and after the second pooling, it becomes 7*7
#After entering the above operation, get 64 7*7 planes
with tf.name_scope('fc1'):
#Initialize the weights of the first fully connected layer
with tf.name_scope('W_fc1'):
W_fc1 = weight_variable([7*7*64,1024],name='W_fc1')#The last field has 7*7*64 neurons, and the fully connected layer has 1024 neurons
with tf.name_scope('b_fc1'):
b_fc1 = bias_variable([1024],name='b_fc1')#1024 nodes
# Flatten the output of pooling layer 2 to 1 dimension
with tf.name_scope('h_pool2_flat'):
h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64],name='h_pool2_flat')
# Find the output of the first fully connected layer
with tf.name_scope('wx_plus_b1'):
wx_plus_b1 = tf.matmul(h_pool2_flat,W_fc1) + b_fc1
with tf.name_scope('relu'):
h_fc1 = tf.nn.relu(wx_plus_b1)
#keep_prob is used to represent the output probability of the neuron
with tf.name_scope('keep_prob'):
keep_prob = tf.placeholder(tf.float32,name='keep_prob')
with tf.name_scope('h_fc1_drop'):
h_fc1_drop = tf.nn.dropout(h_fc1,keep_prob,name='h_fc1_drop')
with tf.name_scope('fc2'):
#Initialize the second fully connected layer
with tf.name_scope('W_fc2'):
W_fc2 = weight_variable([1024,10],name='W_fc2')
with tf.name_scope('b_fc2'):
b_fc2 = bias_variable([10],name='b_fc2')
with tf.name_scope('wx_plus_b2'):
wx_plus_b2 = tf.matmul(h_fc1_drop,W_fc2) + b_fc2
with tf.name_scope('softmax'):
#Calculate output
prediction = tf.nn.softmax(wx_plus_b2)
#Cross entropy cost function
with tf.name_scope('cross_entropy'):
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction),name='cross_entropy')
tf.summary.scalar('cross_entropy',cross_entropy)
#Optimize with AdamOptimizer
with tf.name_scope('train'):
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
# find the accuracy
with tf.name_scope('accuracy'):
with tf.name_scope('correct_prediction'):
#The result is stored in a boolean list
correct_prediction = tf.equal(tf.argmax(prediction,1),tf.argmax(y,1))#argmax returns the position of the largest value in the one-dimensional tensor
with tf.name_scope('accuracy'):
# find the accuracy
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
tf.summary.scalar('accuracy',accuracy)
#Merge all summary
merged = tf.summary.merge_all()
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
train_writer = tf.summary.FileWriter('logs/train',sess.graph)
test_writer = tf.summary.FileWriter('logs/test',sess.graph)
for i in range(1001):
#train model
batch_xs,batch_ys = mnist.train.next_batch(batch_size)
sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys,keep_prob:0.5})
#Record the parameters of the training set calculation
summary = sess.run(merged,feed_dict={x:batch_xs,y:batch_ys,keep_prob:1.0})
train_writer.add_summary(summary,i)
#Record the parameters of the test set calculation
batch_xs,batch_ys = mnist.test.next_batch(batch_size)
summary = sess.run(merged,feed_dict={x:batch_xs,y:batch_ys,keep_prob:1.0})
test_writer.add_summary(summary,i)
if i%100==0:
test_acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels,keep_prob:1.0})
train_acc = sess.run(accuracy,feed_dict={x:mnist.train.images[:10000],y:mnist.train.labels[:10000],keep_prob:1.0})
print ("Iter " + str(i) + ", Testing Accuracy= " + str(test_acc) + ", Training Accuracy= " + str(train_acc))
operation result:
Iter 0, Testing Accuracy= 0.0958, Training Accuracy= 0.0985 Iter 100, Testing Accuracy= 0.5912, Training Accuracy= 0.5964 Iter 200, Testing Accuracy= 0.8075, Training Accuracy= 0.8032 Iter 300, Testing Accuracy= 0.8582, Training Accuracy= 0.853 Iter 400, Testing Accuracy= 0.9402, Training Accuracy= 0.9359 Iter 500, Testing Accuracy= 0.9457, Training Accuracy= 0.9429 Iter 600, Testing Accuracy= 0.9537, Training Accuracy= 0.9525 Iter 700, Testing Accuracy= 0.9591, Training Accuracy= 0.9596 Iter 800, Testing Accuracy= 0.9607, Training Accuracy= 0.9578 Iter 900, Testing Accuracy= 0.964, Training Accuracy= 0.9615 Iter 1000, Testing Accuracy= 0.9667, Training Accuracy= 0.9652
tensorboard:
