Python BP反向传播神经网络

Code:

import numpy as np
"""Sigmod激活函数"""
def sigmod(x):
    return 1.0/(1.0 + np.exp(-x))

def BackpropCE(W1,W2,X,D):
    alpha = 0.9 #学习率
    
    N=4 #4层网络
    for k in range(N): 
        x = X[k,:].T  #对数据每行转置
        d = D[k]      #每行对应结果（标签）
        
        v1 = np.matmul(W1, x)
        y1 = sigmod(v1)
        v = np.matmul(W2,y1)
        y = sigmod(v)
        
        e = d - y
        delta = e*y*(1-y)
        
        e1 = np.matmul(W2.T,delta)
        delta1 = y1*(1-y1)*e1
        
        dW1 = (alpha*delta1).reshape(4,1)
        W1 += dW1
        
        dW2 = alpha * delta *y1
        W2 += dW2
        
    return W1,W2  

def TestBackpropCE():
    X = np.array([[0,0,1],
                 [0,1,1],
                 [1,0,1],
                 [1,1,1]])
    D = np.array([[0],[1],[1],[0]])
    
    W1 = 2*np.random.random((4,3))-1  
    W2 = 2*np.random.random((1,4))-1
    
    for epoch in range(10000):
        W1,W2 = BackpropCE(W1, W2, X, D)
        
    N = 4
    for k in range(N):
        x = X[k,:].T
        v1 = np.matmul(W1,x)
        y1 = sigmod(v1)
        v = np.matmul(W2,y1)
        y = sigmod(v)
        print(y)
        
if __name__ == "__main__":
TestBackpropCE()

标签：sigmod,Python,matmul,y1,神经网络,BP,W2,W1,np
来源： https://blog.csdn.net/gezongbo/article/details/121074037