用神经网络实现异或运算

import numpy as np

class Logistic():

    def __init__(self):

        pass

    def sigmoid(self, z):
        '''激活函数'''
        return 1 / (1 + np.exp(-z))

    def logistic(self, X, theta):
        '''一层神经网络进行简单的逻辑运算'''
        h = self.sigmoid(X*theta.T)
        for i in range(int(h.shape[1])):
            h[0, i] = 1 if h[0, i] >= 0.5 else 0
        return h


X = np.matrix(np.array([1, 0, 1]))
theta1 = np.matrix([[-30, 20, 20], [10, -20, -20]])     #第一层网络的权重
theta2 = np.matrix([-10, 20, -20])                      #第二层网络的权重


log = Logistic()       #实例化

a1 = log.logistic(X, theta1)         #第一层
a1 = np.c_[1, a1]                    #添加偏置单元
a2 = log.logistic(a1, theta2)        #第二层
print(a2)