TF之LoR：基于tensorflow利用邏輯回歸算LoR法實作手寫數字圖檔識别提高準确率

輸出結果

設計代碼

#TF之LoR：基于tensorflow實作手寫數字圖檔識别準确率

import tensorflow as tf

from tensorflow.examples.tutorials.mnist import input_data

import numpy as np  

import matplotlib.pyplot as plt

mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

print(mnist)

#設定超參數

lr=0.001                      #學習率

training_iters=100       #訓練次數

batch_size=100                #每輪訓練資料的大小，如果一次訓練5000張圖檔，電腦會卡死，分批次訓練會更好

display_step=1

#tf Graph的輸入

x=tf.placeholder(tf.float32, [None,784])

y=tf.placeholder(tf.float32, [None, 10])

#設定權重和偏置

w =tf.Variable(tf.zeros([784,10]))

b =tf.Variable(tf.zeros([10]))

#設定運作模式

pred =tf.nn.softmax(tf.matmul(x,w)+b)  #

#設定cost function為cross entropy

cost =tf.reduce_mean(-tf.reduce_sum(y*tf.log(pred),reduction_indices=1))

#GD算法

optimizer=tf.train.GradientDescentOptimizer(lr).minimize(cost)

#初始化權重

init=tf.global_variables_initializer()

#開始訓練

with tf.Session() as sess:

   sess.run(init)

   avg_cost_list=[]

   for epoch in range(training_iters):  #輸入所有訓練資料

       avg_cost=0.

       total_batch=int(mnist.train.num_examples/batch_size)

       for i in range(total_batch): #周遊每個batch

……

       if (epoch+1) % display_step ==0:  #顯示每次疊代日志

           print("疊代次數Epoch:","%04d" % (epoch+1),"下降值cost=","{:.9f}".format(avg_cost))

           avg_cost_list.append(avg_cost)

   print("Optimizer Finished!")

   print(avg_cost_list)

   #測試模型

   correct_prediction=tf.equal(tf.argmax(pred,1),tf.argmax(y,1))

   accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

   print("Accuracy:",accuracy.eval({x:mnist.test.images[:3000],y:mnist.test.labels[:3000]}))

   xdata=np.linspace(0,training_iters,num=len(avg_cost_list))  

   plt.figure()  

   plt.plot(xdata,avg_cost_list,'r')

   plt.xlabel('訓練輪數')

   plt.ylabel('損失函數')

   plt.title('TF之LiR：基于tensorflow實作手寫數字圖檔識别準确率——Jason Niu')    

   plt.show()