%matplotlib inline
import tensorflow as tf
import tensorflow.examples.tutorials.mnist.input_data as input_data
mnist=input_data.read_data_sets('MNIST_data/',one_hot=True)
import matplotlib.pyplot as plt
 
def plot_image(image):                #圖片顯示函數(shù)
  plt.imshow(image.reshape(28,28),cmap='binary')
  plt.show()
 
print("訓練集數(shù)量：",mnist.train.num_examples,
   "特征值組成：",mnist.train.images.shape,
   "標簽組成：",mnist.train.labels.shape)
 
batch_images,batch_labels=mnist.train.next_batch(batch_size=10)  #批量讀取數(shù)據(jù)
print(batch_images.shape,batch_labels.shape)
 
print('標簽值：',np.argmax(mnist.train.labels[1000]),end=' ')  #np.argmax()得到實際值
print('獨熱編碼表示：',mnist.train.labels[1000])
plot_image(mnist.train.images[1000])         #顯示數(shù)據(jù)集中第1000張圖片

import numpy as np
import tensorflow as tf
import tensorflow.examples.tutorials.mnist.input_data as input_data
mnist=input_data.read_data_sets('MNIST_data/',one_hot=True)
import matplotlib.pyplot as plt
 
#定義占位符、變量、前向計算
x=tf.placeholder(tf.float32,[None,784],name='x')
y=tf.placeholder(tf.float32,[None,10],name='y')
W=tf.Variable(tf.random_normal([784,10]),name='W')
b=tf.Variable(tf.zeros([10]),name='b')
forward=tf.matmul(x,W)+b
pred=tf.nn.softmax(forward)               #通過softmax將線性結(jié)果分類處理
 
#計算預測值與真實值的匹配個數(shù)
correct_prediction=tf.equal(tf.argmax(pred,1),tf.argmax(y,1))
#將上一步得到的布爾值轉(zhuǎn)換為浮點數(shù)，并求平均值，得到準確率
accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
 
ss=tf.Session()
init=tf.global_variables_initializer()
ss.run(init)
 
#超參數(shù)設(shè)置
train_epochs=50
batch_size=100                        #每個批次的樣本數(shù)
batch_num=int(mnist.train.num_examples/batch_size)      #一輪需要訓練多少批
learning_rate=0.01
 
#定義交叉熵損失函數(shù)、梯度下降優(yōu)化器
loss_function=tf.reduce_mean(-tf.reduce_sum(y*tf.log(pred),reduction_indices=1)) 
optimizer=tf.train.GradientDescentOptimizer(learning_rate).minimize(loss_function)
 
for epoch in range(train_epochs):
  for batch in range(batch_num):              #分批次讀取數(shù)據(jù)進行訓練
    xs,ys=mnist.train.next_batch(batch_size)
    ss.run(optimizer,feed_dict={x:xs,y:ys})
  #每輪訓練結(jié)束后通過帶入驗證集的數(shù)據(jù)，檢測模型的損失與準去率 
  loss,acc=ss.run([loss_function,accuracy],\
          feed_dict={x:mnist.validation.images,y:mnist.validation.labels})
  print('第%2d輪訓練：損失為：%9f，準確率：%.4f'%(epoch+1,loss,acc))

prediction=ss.run(tf.argmax(pred,1),feed_dict={x:mnist.test.images})
 
def show_result(images,labels,prediction,index,num=10):   #繪制圖形顯示預測結(jié)果
  pic=plt.gcf()                      #獲取當前圖像
  pic.set_size_inches(10,12)               #設(shè)置圖片大小
  for i in range(0,num):
    sub_pic=plt.subplot(5,5,i+1)            #獲取第i個子圖
    #將第index個images信息顯示到子圖上
    sub_pic.imshow(np.reshape(images[index],(28,28)),cmap='binary') 
    title="label:"+str(np.argmax(labels[index]))    #設(shè)置子圖的title內(nèi)容
    if len(prediction)>0:
      title+=",predict:"+str(prediction[index])
      
    sub_pic.set_title(title,fontsize=10)
    sub_pic.set_xticks([])               #設(shè)置x、y坐標軸不顯示
    sub_pic.set_yticks([])
    index+=1
  plt.show()
show_result(mnist.test.images,mnist.test.labels,prediction,10)