softmax和卷积神经网络训练Minist
https://www.cnblogs.com/zf-blog/p/6075286.html
##代码
#input-->conv--->activation--->pool---->fc--->logits--->softmax
import tensorflow as tf
import os
import csv
import argparse
from tensorflow.examples.tutorials.mnist import input_data
import sys
learning_rate_init = 0.001
train_epochs = 1
batch_size =100
display_step = 10
n_input =784
n_classes = 10
FLAGS = None
def weightVariable(shape, name_str,stddev=0.1):
inital = tf.random_normal(shape= shape,stddev=stddev,dtype=tf.float32)
return tf.Variable(inital,dtype=tf.float32,name=name_str)
def biasVariable(shape,name_str,stddev=0.00001):
inital = tf.random_normal(shape=shape,stddev=stddev,name=name_str)
return inital
def conv2d(x,W,b,stride=1,padding='SAME'):
with tf.name_scope('conv_wx_b'):
y = tf.nn.conv2d(x,W,strides=[1,stride,stride,1],padding=padding)
y= tf.nn.bias_add(y,b)
return y
def Activation(x,activation=tf.nn.relu,name='relu'):
with tf.name_scope(name):
y = activation(x)
return y
def pool2d(x,pool=tf.nn.max_pool,k=2,stride=2):
return pool(x,ksize=[1,k,k,1],strides=[1,stride,stride,1],padding='VALID')
def full_connected(x,W,b,activate=tf.nn.relu,act_name='relu'):
with tf.name_scope('Wx_b'):
y=tf.matmul(x,W)
y=tf.add(y,b)
with tf.name_scope(act_name):
y=activate(y)
return y
# def EvaluateModeOnData(sess,images,labels):
# n_samples = images.shape[0]
# per_batch_size = batch_size
# loss = 0
# acc = 0
#
# if(n_samples<=per_batch_size):
# batch_count = 1
# loss,acc = sess.run([cross_entropy_loss,accuracy],feed_dict={X_origin:images,Y_true:labels,
# learning_rate:learning_rate_init})
# else:
# batch_count = int(n_samples/per_batch_size)
# batch_start = 0
# for idx in range(batch_count):
# batch_loss,batch_acc = sess.run([cross_entropy_loss,accuracy],feed_dict={X_origin:images[batch_start:batch_start+per_batch_size,:],
# Y_true:labels[batch_start:batch_start+per_batch_size,:],
# learning_rate:learning_rate_init})
# batch_start += per_batch_size
# loss += batch_loss
# acc += batch_acc
# return loss/batch_count,acc/batch_count
def main(_):
mnist = input_data.read_data_sets(FLAGS.data_dir, one_hot=True)
with tf.Graph().as_default():
with tf.name_scope('inputs'):
X_origin = tf.placeholder(tf.float32,[None,n_input],name='x_o')
Y_true = tf.placeholder(tf.float32,[None,n_classes],name='y_t')
#将一维数据转换为二维的784-->28*28
X_image = tf.reshape(X_origin,[-1,28,28,1])
with tf.name_scope('inference'):
with tf.name_scope('conv2d'):
weights = weightVariable(shape=[5,5,1,24],name_str='weights')
bias = biasVariable(shape=[24],name_str='bias')
conv_out = conv2d(X_image,weights,bias,stride=1,padding='VALID')
with tf.name_scope('activate'):
activate_out = Activation(conv_out,activation=tf.nn.relu,name='relu')
with tf.name_scope('pool2d'):
pool_out = pool2d(activate_out,pool=tf.nn.max_pool,k=2,stride=2)
with tf.name_scope('featurereshape'):
feature = tf.reshape(pool_out,[-1,12*12*24])
with tf.name_scope('FC_Liear'):
weights = weightVariable(shape=[12*12*24,n_classes],name_str='weights')
bias = biasVariable(shape=[n_classes],name_str='biases')
#tf.identity 恒等映射
Ypred_logits = full_connected(feature,weights,bias,activate=tf.identity,act_name='identity')
Y_P_max = tf.argmax(Ypred_logits,1)
Y_T_max = tf.argmax(Y_true,1)
with tf.name_scope('loss'):
cross_entropy_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=Y_true,logits=Ypred_logits))
with tf.name_scope('train'):
learning_rate = tf.placeholder(tf.float32)
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
trainer = optimizer.minimize(cross_entropy_loss)
with tf.name_scope('Evaluate'):
correct_pred = tf.equal(tf.argmax(Ypred_logits,1),tf.argmax(Y_true,1))
accuracy = tf.reduce_mean(tf.cast(correct_pred,tf.float32))
init = tf.global_variables_initializer();
print('把节点写入事件文件')
summary_writer = tf.summary.FileWriter(logdir='logs_3',graph=tf.get_default_graph())
summary_writer.close()
results_list = list()
training_step = 0
results_list.append(['learning_rate',learning_rate,
'train_epochs',train_epochs,
'batch_size',batch_size,
'display_step',display_step])
with tf.Session() as sess:
sess.run(init)
# total_batches = int(mnist.train.num_examples/batch_size)
# print('per batch size:',batch_size)
# print('train sample count:',mnist.train.num_examples)
# print('tatal batches count:',total_batches)
# train_step = 0
#
# for epoch in range(train_epochs):
# for batch_idx in range(total_batches):
# batch_x,batch_y = mnist.train.next_batch(batch_size)
# sess.run(trainer,feed_dict={X_origin:batch_x,Y_true:batch_y,learning_rate:learning_rate_init})
# training_step += 1
#
# if train_step%display_step == 0:
# start_idx = max(0,(batch_idx-display_step)*batch_size)
# end_idx = batch_idx*batch_size
# train_loss,train_acc = EvaluateModeOnData(sess,mnist.train.images[start_idx:end_idx,:],
# mnist.train.labels[start_idx:end_idx,:])
# print("train step:",str(train_step)," Train loss:","{:.6f}".format(train_loss)," train_acc",
# "{:.6f}".format(train_acc))
for step in range(1000):
batch_xs,batch_ys = mnist.train.next_batch(1000)
_,train_loss = sess.run([trainer,cross_entropy_loss],feed_dict={X_origin:batch_xs,Y_true:batch_ys,learning_rate:learning_rate_init})
print("train step:",step," train_loss:",train_loss)
ypm = sess.run(Y_P_max,feed_dict={X_origin:mnist.test.images})
ytm = sess.run(Y_T_max,feed_dict={Y_true:mnist.test.labels})
for i in range(len(ypm)):
print ("P 序号：%s 值：%s" % (i + 1, ypm[i]))
print ("T 序号：%s 值：%s" % (i + 1, ytm[i]))
#print("max pre Y : ",ypm)
#print("max true Y: ",ytm)
accuracy_score = sess.run(accuracy,feed_dict={X_origin:mnist.test.images,Y_true:mnist.test.labels})
print("accuracy_score:",accuracy_score)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--data_dir', type=str, default='D:/tensorflow/pys/test1/tmp/tensorflow/mnist/input_data',
help='Directory for storing input data')
FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)