import os
import time
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
import tensorflow.examples.tutorials.mnist.input_data as input_data
plt.rcParams["font.sans-serif"] = ["simhei"]
### 程序执行开始时间
begintime=time.time()
### 读取数据
mnist=tf.keras.datasets.mnist
(train_images,train_labels),(test_images,test_labels)=mnist.load_data()
valid_split=0.2 # 验证集占比
total_num=len(train_images) # 数据集长度
train_num=int((1-valid_split)*total_num) # 训练数据集长度
### 分配数据
train_x=train_images[:train_num]
train_y=train_labels[:train_num]
valid_x=train_images[train_num:]
valid_y=train_labels[train_num:]
test_x =test_images
test_y =test_labels
### 数据拉直
train_x=train_x.reshape(-1,784)
valid_x=valid_x.reshape(-1,784)
test_x = test_x.reshape(-1,784)
### 特征数据归一化
train_x=tf.cast(train_x/255.0,tf.float32)
valid_x=tf.cast(valid_x/255.0,tf.float32)
test_x =tf.cast( test_x/255.0,tf.float32)
### 对标签进行独热编码
train_y=tf.one_hot(train_y,depth=10)
valid_y=tf.one_hot(valid_y,depth=10)
test_y =tf.one_hot( test_y,depth=10)
### 定义模型
def model(x,w,b):
pred=tf.matmul(x,w)+b
return tf.nn.softmax(pred)
### 定义变量
W=tf.Variable(tf.random.normal([784,10],mean=0.0,stddev=1.0,dtype=tf.float32),name="W")
B=tf.Variable(tf.zeros([10]),dtype=tf.float32,name="B")
### 定义交叉熵损失函数
def loss(x,y,w,b):
pred=model(x,w,b)
myloss=tf.keras.losses.categorical_crossentropy(y_true=y,y_pred=pred)
return tf.reduce_mean(myloss)
### 定义参数
epochs_number=1 # 训练了几次
batch_size=200 # 批数据大小
train_epochs=100 # 迭代轮次
learning_rate=0.001 # 学习率
display_step=10 # 显示粒度
total_batch=int(len(train_x)/batch_size)
### 计算梯度
def grad(x,y,w,b):
with tf.GradientTape() as tape:
myloss=loss(x,y,w,b)
return tape.gradient(myloss,[w,b])
### 定义优化器
optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate)
### 定义准确率
def accuracy(x,y,w,b):
pred=model(x,w,b)
correct_prediction=tf.equal(tf.argmax(pred,1),tf.argmax(y,1))
return tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
def plot_images_labels_prediction(images,labels,prediction,index,num=10):
fig=plt.gcf()
fig.set_size_inches(10,12)
if num>25:
num=25
for i in range(num):
ax=plt.subplot(5,5,i+1)
ax.imshow(np.reshape(images[index],(28,28)),cmap="binary")
title=u"标签="+str(np.argmax(labels[index]))
if len(prediction)>0:
title+=",预测值="+str(prediction[index].numpy())
ax.set_title(title,fontsize=10)
ax.set_xticks([])
ax.set_yticks([])
index+=1
plt.show()
### 训练
loss_list_train=[]
loss_list_valid=[]
acc_list_train =[]
acc_list_valid =[]
for epoch in range(train_epochs):
for step in range(total_batch):
xs=train_x[step*batch_size:(step+1)*batch_size]
ys=train_y[step*batch_size:(step+1)*batch_size]
grads=grad(xs,ys,W,B)
optimizer.apply_gradients(zip(grads,[W,B]))
loss_train=loss(train_x,train_y,W,B).numpy()
loss_valid=loss(valid_x,valid_y,W,B).numpy()
acc_train =accuracy(train_x,train_y,W,B).numpy()
acc_valid =accuracy(valid_x,valid_y,W,B).numpy()
loss_list_train.append(loss_train)
loss_list_valid.append(loss_valid)
acc_list_train.append(acc_train)
acc_list_valid.append(acc_valid)
if (epoch+1) % display_step == 0:
print("训练轮次:{:0>5d},训练损失率:{:.10f},训练准确率:{:.5f},验证损失率:{:.10f},验证准确率:{:.5f}".format(epoch+1,loss_train,acc_train,loss_v
alid,acc_valid))
print("程序运行耗时:{:.10f}秒".format(time.time()-begintime))
while True:
os.system("clear")
print("1.........训练数据")
print("2.........识别数字")
print("3.........训练信息")
menu_input=input("请输入选择(Quit/Q退出):")
if menu_input == "1":
epochs_number+=1
### 程序执行开始时间
begintime=time.time()
### 训练
loss_list_train=[]
loss_list_valid=[]
acc_list_train =[]
acc_list_valid =[]
for epoch in range(train_epochs):
for step in range(total_batch):
xs=train_x[step*batch_size:(step+1)*batch_size]
ys=train_y[step*batch_size:(step+1)*batch_size]
grads=grad(xs,ys,W,B)
optimizer.apply_gradients(zip(grads,[W,B]))
loss_train=loss(train_x,train_y,W,B).numpy()
loss_valid=loss(valid_x,valid_y,W,B).numpy()
acc_train =accuracy(train_x,train_y,W,B).numpy()
acc_valid =accuracy(valid_x,valid_y,W,B).numpy()
loss_list_train.append(loss_train)
loss_list_valid.append(loss_valid)
acc_list_train.append(acc_train)
acc_list_valid.append(acc_valid)
if (epoch+1) % display_step == 0:
print("训练轮次:{:0>5d},训练损失率:{:.10f},训练准确率:{:.5f},验证损失率:{:.10f},验证准确率:{:.5f}".format(epoch+1,loss_train,acc_trai
n,loss_valid,acc_valid))
print("程序运行耗时:{:.10f}秒".format(time.time()-begintime))
elif menu_input == "2":
myindex=int(input("请输入索引数字(5-9900之间):"))
mysum=int(input("请输入显示数量(5-25之间):"))
if myindex<=5 or myindex>=9900:
myindex=7425
if mysum<=5 or mysum>=25:
mysum=25
prediction_result=tf.argmax(model(train_x,W,B),1)
plot_images_labels_prediction(train_x,train_y,prediction_result,myindex,mysum)
elif menu_input == "3":
print("训练次数:{:0>5d},训练损失率:{:.10f},训练准确率:{:.5f},验证损失率:{:.10f},验证准确率:{:.5f}".format(epochs_number,loss_train,acc_train,loss_valid,acc_valid))
input("任意键返回")
elif menu_input.upper() == "QUIT" or menu_input.upper() == "Q":
break
训练轮次:00010,训练损失率:0.6595796943,训练准确率:0.85233,验证损失率:0.6406478286,验证准确率:0.85425
训练轮次:00020,训练损失率:0.4611262083,训练准确率:0.88979,验证损失率:0.4707993865,验证准确率:0.88725
训练轮次:00030,训练损失率:0.3869549036,训练准确率:0.90315,验证损失率:0.4105678797,验证准确率:0.89967
训练轮次:00040,训练损失率:0.3459119797,训练准确率:0.91098,验证损失率:0.3793669343,验证准确率:0.90533
训练轮次:00050,训练损失率:0.3192890286,训练准确率:0.91658,验证损失率:0.3597021401,验证准确率:0.91050
训练轮次:00060,训练损失率:0.3004053831,训练准确率:0.92017,验证损失率:0.3460557461,验证准确率:0.91350
训练轮次:00070,训练损失率:0.2863721251,训练准确率:0.92358,验证损失率:0.3360227942,验证准确率:0.91483
训练轮次:00080,训练损失率:0.2755896747,训练准确率:0.92596,验证损失率:0.3285307288,验证准确率:0.91592
训练轮次:00090,训练损失率:0.2669441998,训练准确率:0.92787,验证损失率:0.3226217926,验证准确率:0.91717
…
训练次数:00001,训练损失率:0.2599059641,训练准确率:0.92944,验证损失率:0.3179240823,验证准确率:0.91883