神经网络

一、手工搭建神经网络

 1 import numpy as np
 2 import scipy.special
 3 import matplotlib.pyplot as plt
 4 import pylab
 5 
 6 
 7 class NeuralNetwork():
 8     def __init__(self,inputnodes,hiddennodes,outputnodes,learningrate):
 9         self.inodes = inputnodes
10         self.hnodes = hiddennodes
11         self.onodes = outputnodes
12         self.lr = learningrate
13         self.wih = np.random.normal(0.0,pow(self.hnodes,-0.5),(self.hnodes,self.inodes))
14         self.who = np.random.normal(0.0,pow(self.onodes,-0.5),(self.onodes,self.hnodes))
15         self.activation_function = lambda x:scipy.special.expit(x)
16         pass
17     def train(self,input_list,target_list):
18         inputs = np.array(input_list,ndmin=2).T
19         targets = np.array(target_list,ndmin=2).T
20         hidden_inputs = np.dot(self.wih,inputs)
21         hidden_outputs = self.activation_function(hidden_inputs)
22         final_inputs = np.dot(self.who,hidden_outputs)
23         final_outputs = self.activation_function(final_inputs)
24         output_errors = targets-final_outputs
25         hidden_errors = np.dot(self.who.T,output_errors)
26         self.who += self.lr*np.dot((output_errors*final_outputs*(1.0-final_outputs)),np.transpose(hidden_outputs))
27         self.wih += self.lr*np.dot((hidden_errors*hidden_outputs*(1.0-hidden_outputs)),np.transpose(inputs))
28         pass
29     def query(self,input_list):
30         inputs = np.array(input_list,ndmin=2).T
31         hidden_inputs = np.dot(self.wih,inputs)
32         hidden_outputs = self.activation_function(hidden_inputs)
33         final_inputs = np.dot(self.who,hidden_outputs)
34         final_outputs = self.activation_function(final_inputs)
35         return final_outputs
36         print('n')
37 
38 
39 input_nodes = 2
40 hidden_nodes = 100
41 output_nodes = 2
42 learning_rate = 0.3
43 n = NeuralNetwork(input_nodes,hidden_nodes,output_nodes,learning_rate)
44 training_data_file = open(r"E:\人工智能\train.csv","r")
45 training_data_list = training_data_file.readlines()
46 training_data_file.close()
47 print(training_data_list[0])
48 
49 
50 epochs = 10
51 for e in range(epochs):
52     for record in training_data_list:
53         all_values = record.split(",")
54         inputs = (np.asfarray(all_values[1:])/255.0*0.99)+0.01
55         targets = np.zeros(output_nodes)+0.01
56         targets[int(float(all_values[0]))]=0.99
57         n.train(inputs,targets)
58         pass
59     pass
60 
61 
62 test_data_file = open(r"E:\人工智能\train.csv","r")
63 test_data_list = test_data_file.readlines()
64 test_data_file.close()
65 all_values = record.split(',')
66 
67 scorecard=[]
68 for record in test_data_list:
69     all_values = record.split(',')
70     correct_label = int(float(all_values[0]))
71     inputs = (np.asfarray(all_values[1:])/255.0*0.99)+0.01
72     outputs = n.query(inputs)
73     label = np.argmax(outputs)
74     if(label == correct_label):
75         scorecard.append(1)
76     else:
77         scorecard.append(0)
78         pass
79     pass
80 print(outputs)
81 
82 scorecard_array = np.asarray(scorecard)
83 print("正确率是：",(scorecard_array.sum()/scorecard_array.size)*100,'%')

 

 

 

二、BP神经网络

 1 def sigmoid(x):  #映射函数
 2     return 1/(1+math.exp(-x))
 3 import math
 4 import numpy as np
 5 import pandas as pd
 6 from pandas import DataFrame,Series
 7 
 8 #中间层神经元输入和输出层神经元输入
 9 Net_in = DataFrame(0.6,index=['input1','input2','theata'],columns=['a'])
10 Out_in = DataFrame(0,index=['input1','input2','input3','input4','theata'],columns=['a'])
11 Net_in.iloc[2,0] = -1
12 Out_in.iloc[4,0] = -1
13 real = Net_in.iloc[0,0]**2+Net_in.iloc[1,0]**2
14 print(Out_in)
15 
16 #中间层和输出层神经元权值
17 W_mid=DataFrame(0.5,index=['input1','input2','theata'],columns=['mid1','mid2','mid3','mid4'])
18 W_out=DataFrame(0.5,index=['input1','input2','input3','input4','theata'],columns=['a'])
19 W_mid_delta=DataFrame(0,index=['input1','input2','theata'],columns=['mid1','mid2','mid3','mid4'])
20 W_out_delta=DataFrame(0,index=['input1','input2','input3','input4','theata'],columns=['a'])
21 print(W_out)
22 
23 #中间层的输出
24 for i in range(0,4):
25     Out_in.iloc[i,0] = sigmoid(sum(W_mid.iloc[:,i]*Net_in.iloc[:,0]))
26 #输出层的输出/网络输出
27 res = sigmoid(sum(Out_in.iloc[:,0]*W_out.iloc[:,0]))
28 error = abs(res-real)
29 print(error)
30 
31 yita=0.6
32 #输出层权值变化量
33 W_out_delta.iloc[:,0] = yita*res*(1-res)*(real-res)*Out_in.iloc[:,0]
34 W_out_delta.iloc[4,0] = -(yita*res*(1-res)*(real-res))
35 W_out = W_out + W_out_delta #输出层权值更新
36 print(W_out)
37 
38 #中间层权值变化量
39 for i in range(0,4):
40     W_mid_delta.iloc[:,i] = yita*Out_in.iloc[i,0]*(1-Out_in.iloc[i,0])*W_out.iloc[i,0]*res*(1-res)*(real-res)*Net_in.iloc[:,0]
41     W_mid_delta.iloc[2,i] = -(yita*Out_in.iloc[i,0]*(1-Out_in.iloc[i,0])*W_out.iloc[i,0]*res*(1-res)*(real-res))
42 W_mid = W_mid + W_mid_delta #中间层权值更新
43 print(W_mid)

 

 

posted @ 2022-03-18 17:11  熊能能  阅读(53)  评论(0)    收藏  举报