BP网络搭建和例题测试

最新推荐文章于 2023-08-10 19:09:42 发布

L-Y-B

最新推荐文章于 2023-08-10 19:09:42 发布

阅读量1.7k

点赞数

文章标签： python 人工智能深度学习神经网络

本文链接：https://blog.csdn.net/weixin_53639312/article/details/123578461

版权

本文介绍了一种优化后的代码实现，通过使用数组替代DataFrame，显著提高了训练速度。作者详细展示了如何使用sigmoid函数、梯度下降和BP算法训练一个多层感知器，并在训练集和测试集上计算误差。最后，文章着重于提升预测阶段的性能和效率。

摘要由CSDN通过智能技术生成

1.代码如下：

# -*- coding: utf-8 -*-
"""
Created on Mon Mar 14 18:58:05 2022

@author: L
"""
import math
import pandas as pd
import numpy  as np
import matplotlib.pyplot as plt
from pandas import DataFrame,Series
#定义函数：
def sigmoid(x):
    return  1/(1+math.exp(-x))
#读取数据
data_tr=pd.read_csv('D:\\text_class\\data_tr.txt')
data_te=pd.read_csv('D:\\text_class\\data_te.txt')
#设置学习量
yita=0.3
#设置结构
Err = []
Net_in=DataFrame(0.6,index=['input1','input2','theata'],columns=['a'])
Out_in=DataFrame(0,index=['input1','input2','input3','input4','theata'],columns=['a'])
Net_in.iloc[2,0]=-1 
Out_in.iloc[4,0]=-1
W_mid=DataFrame(0.5,index=['input1','input2','theata'],columns=['mid1','mid2','mid3','mid4'])
W_out=DataFrame(0.5,index=['input1','input2','input3','input4','theata'],columns=['a'])
W_mid_delta=DataFrame(0,index=['input1','input2','theata'],columns=['mid1','mid2','mid3','mid4'])
W_out_delta=DataFrame(0,index=['input1','input2','input3','input4','theata'],columns=['a'])
#设置训练
n=len(data_tr)
for j in range(1000):
    error = []
    for i in range(n):
        Net_in.iloc[0,0]=data_tr.iloc[i,0]
        Net_in.iloc[1,0]=data_tr.iloc[i,1]
        real=data_tr.iloc[i,2]
        for t in range(0,4):
            Out_in.iloc[t,0]=sigmoid(sum(W_mid.iloc[:,t]*Net_in.iloc[:,0]))
        res=sigmoid(sum(Out_in.iloc[:,0]*W_out.iloc[:,0]))
        error.append(abs(res-real))
        print(i,':',res,real)
        W_out_delta.iloc[:,0]=yita*res*(1-res)*(real-res)*Out_in.iloc[:,0]
        W_out_delta.iloc[4,0]=-(yita*res*(1-res)*(real-res))
        W_out=W_out+W_out_delta
        for g in range(0,4):
            W_mid_delta.iloc[:,g]=yita*Out_in.iloc[g,0]*(1-Out_in.iloc[g,0])*W_out.iloc[g,0]*res*(1-res)*(real-res)*Net_in.iloc[:,0]
            W_mid_delta.iloc[2,g]=-(yita*Out_in.iloc[g,0]*(1-Out_in.iloc[g,0])*W_out.iloc[g,0]*res*(1-res)*(real-res))
        W_mid=W_mid+W_mid_delta
    Err.append(np.mean(error))
print(W_mid,W_out)
plt.plot(Err)#训练集上每一轮的平均误差
plt.show()
plt.close()
#测试部分
error_te=[]
k=len(data_te)
Net_in.iloc[2,0]=-1 
Out_in.iloc[4,0]=-1
for i in range(k):
    Net_in.iloc[0,0]=data_te.iloc[i,0]
    Net_in.iloc[1,0]=data_te.iloc[i,1]
    real=data_te.iloc[i,2]
    for t in range(0,4):
        Out_in.iloc[t,0]=sigmoid(sum(W_mid.iloc[:,t]*Net_in.iloc[:,0]))
    res=sigmoid(sum(Out_in.iloc[:,0]*W_out.iloc[:,0]))
    error_te.append(abs(res-real))
plt.plot(error_te)#训练集上每一轮的平均误差
plt.show()
plt.close()

该代码使用了DataFrame模块，运行速度会很慢，训练1000次需要30多分钟，可用数组代替，效率会快很多

运行结果如下：

2.求出预测值

题如下：

代码如下：

# -*- coding: utf-8 -*-
"""
Created on Sun Mar 13 20:56:46 2022

@author: L
"""

import math
import pandas as pd
import numpy  as np
import matplotlib.pyplot as plt
from pandas import DataFrame,Series
#定义函数：
def sigmoid(x):
    return  1/(1+math.exp(-x))
#读取数据
data_tr=pd.read_csv('D:\\text_class\\data_tr_else.txt')
data_te=pd.read_csv('D:\\text_class\\data_te_else.txt')
#设置学习量
yita=0.3
#设置结构
Err = []
Net_in=DataFrame(0.6,index=['input1','input2','theata'],columns=['a'])
Out_in=DataFrame(0,index=['input1','input2','input3','input4','theata'],columns=['a'])
Net_in.iloc[2,0]=-1 
Out_in.iloc[4,0]=-1
W_mid=DataFrame(0.5,index=['input1','input2','theata'],columns=['mid1','mid2','mid3','mid4'])
W_out=DataFrame(0.5,index=['input1','input2','input3','input4','theata'],columns=['a'])
W_mid_delta=DataFrame(0,index=['input1','input2','theata'],columns=['mid1','mid2','mid3','mid4'])
W_out_delta=DataFrame(0,index=['input1','input2','input3','input4','theata'],columns=['a'])
#设置训练
n=len(data_tr)
for j in range(1000):
    error = []
    for i in range(n):
        Net_in.iloc[0,0]=data_tr.iloc[i,0]
        Net_in.iloc[1,0]=data_tr.iloc[i,1]
        real=data_tr.iloc[i,2]
        for t in range(0,4):
            Out_in.iloc[t,0]=sigmoid(sum(W_mid.iloc[:,t]*Net_in.iloc[:,0]))
        res=sigmoid(sum(Out_in.iloc[:,0]*W_out.iloc[:,0]))
        error.append(abs(res-real))
        print(i,':',res,real)
        W_out_delta.iloc[:,0]=yita*res*(1-res)*(real-res)*Out_in.iloc[:,0]
        W_out_delta.iloc[4,0]=-(yita*res*(1-res)*(real-res))
        W_out=W_out+W_out_delta
        for g in range(0,4):
            W_mid_delta.iloc[:,g]=yita*Out_in.iloc[g,0]*(1-Out_in.iloc[g,0])*W_out.iloc[g,0]*res*(1-res)*(real-res)*Net_in.iloc[:,0]
            W_mid_delta.iloc[2,g]=-(yita*Out_in.iloc[g,0]*(1-Out_in.iloc[g,0])*W_out.iloc[g,0]*res*(1-res)*(real-res))
        W_mid=W_mid+W_mid_delta
    Err.append(np.mean(error))
print(W_mid,W_out)
plt.plot(Err)#训练集上每一轮的平均误差
plt.show()
plt.close()
#测试部分


k=len(data_te)
Net_in.iloc[2,0]=-1 
Out_in.iloc[4,0]=-1
for i in range(k):
    Net_in.iloc[0,0]=data_te.iloc[i,0]
    Net_in.iloc[1,0]=data_te.iloc[i,1]
    for t in range(0,4):
        Out_in.iloc[t,0]=sigmoid(sum(W_mid.iloc[:,t]*Net_in.iloc[:,0]))
    res=sigmoid(sum(Out_in.iloc[:,0]*W_out.iloc[:,0]))
    print('第'+str(i)+'个预测值为: '+str(res))

运行结果如下：