import numpy as npdef tanh(x):    #定义tan函数return np.tanh(x)def tanh_deriv(x): #定义tan函数的导数return 1.0-np.tan(x)*np.tan(x)def logistic(x):#定义逻辑回归函数return 1/(1+np.exp(-x))def logistic_deriv(x):#定义逻辑回归函数的导数return logistic(x)*(1-logistic(x))#定义一个名为神经网络的类
class NeuralNetwork:#init类似于构造函数 初始化一个类对象是默认调用def __init__(self,layers,activation='tanh'):#选择该神经网络所使用的损耗函数if activation == 'logistic':self.activation = logisticself.activation_deriv = logistic_derivelif activation == 'tanh':self.activation = tanhself.activation_deriv = tanh_derivself.weights = [] #初始化一个weigt权重容器#初始化权重 是随机的 还要加上偏差biasfor i in range(0,len(layers)-1):self.weights.append((2*np.random.random((layers[i]+1,layers[i+1]+1))-1)*0.25)#参数说明def fit(self,x,y,learn_rate = 0.2,epochs = 100000):x = np.atleast_2d(x) #升到2维以上temp = np.ones([x.shape[0],x.shape[1]+1])  #增加的一类用于处理Bias 因此设置为1temp[:,0:-1] = x  #左闭右开  把X的值赋值给新的数组x = tempy = np.array(y)for k in range(epochs):  #采用抽样的方法进行训练i =  np.random.randint(x.shape[0])a = [x[i]]for l in range(len(self.weights)):a.append(self.activation(np.dot(a[l],self.weights[l])))  #矩阵求和 然后在用损耗函数处理 再放在数组后面  整个过程处理完毕
error = y[i] - a[-1] #计算结果误差deltas = [ error*self.activation_deriv(a[-1])]  #计算输出层误差#现在开始反向走for i in range(len(a)-2,0,-1): #
                deltas.append(deltas[-1].dot(self.weights[i].T)*self.activation_deriv(a[i]))                deltas.reverse()for i in range(len(self.weights)):layer = np.atleast_2d(a[i])delta = np.atleast_2d(deltas[i])self.weights[i]+= learn_rate*layer.T.dot(delta)def predict(self,x):x = np.array(x)temp = np.ones(x.shape[0]+1)temp[0:-1] = xa = tempfor i in range(0,len(self.weights)):a = self.activation(np.dot(a,self.weights[i]))return a[0]

from NeuralNetwork import NeuralNetwork
import numpy as np nn = NeuralNetwork([2,2,1],'tanh')
#建立两层神经网络 输入层为2维  隐藏层为2维 输出层为1维
x = np.array([[0,0],[0,1],[1,0],[1,1]])
y = np.array([0,1,1,0])
nn.fit(x,y)for i in [[0,0],[0,1],[1,0],[1,1]]:print(nn.predict(i))