神经网络BP算法

每层激活函数均为sigmoid函数。

import numpy as np

w=[0,0.1,0.15,0.2,0.25,0.3,0.35,0.4,0.45,0.5,0.55,0.6,0.65] #数组下标从0开始
b=[0,0.35,0.65]
l=[0,5,10]

def sigmoid(z):
    return 1.0/(1+np.exp(-z))

def bp(w,b,l):
    # 1.前向传播，计算预测值与误差
    h1 = sigmoid(w[1] * l[1] + w[2] * l[2] + b[1])
    h2 = sigmoid(w[3] * l[1] + w[4] * l[2] + b[1])
    h3 = sigmoid(w[5] * l[1] + w[6] * l[2] + b[1])

    o1 = sigmoid(w[7] * h1 + w[9] * h2 + w[11] * h3 + b[2])
    o2 = sigmoid(w[8] * h1 + w[10] * h2 + w[12] * h3 + b[2])

    e = np.square(0.01 - o1) / 2,0 + np.square(0.99 - o2) /2.0

    # 2.后向传播，更新权重w
    t1 = -(0.01 - o1) * o1 * (1 - o1)
    t2 = -(0.99 - o2) * o2 * (1 - o2)

    w[7] = w[7] - 0.05 * t1 * h1
    w[9] = w[9] - 0.05 * t1 * h2
    w[11] = w[11] - 0.05 * t1 * h3

    w[8] = w[8] - 0.05 * t2 * h1
    w[10] = w[10] - 0.05 * t2 * h2
    w[12] = w[12] - 0.05 * t2 * h3

    w[1] = w[1] - 0.5 * (t1 * w[7] + t2 * w[8]) * h1 * (1 - h1) * l[1]
    w[2] = w[2] - 0.5 * (t1 * w[7] + t2 * w[8]) * h1 * (1 - h1) * l[2]
    w[3] = w[3] - 0.5 * (t1 * w[9] + t2 * w[10]) * h2 * (1 - h2) * l[1]
    w[4] = w[4] - 0.5 * (t1 * w[9] + t2 * w[10]) * h2 * (1 - h2) * l[2]
    w[5] = w[5] - 0.5 * (t1 * w[11] + t2 * w[12]) * h3 * (1 - h3) * l[1]
    w[6] = w[6] - 0.5 * (t1 * w[11] + t2 * w[12]) * h3 * (1 - h3) * l[2]

    return o1,o2,w,e

for i in range(1001):
    o1,o2,w,e=bp(w,b,l)
    print('第{}次迭代:预测值为({},{}),总误差:{},权重系数:{}'.format(i,o1,o2,e,w))