Simple Neural Network [Preview]

import math
import random
import string

random.seed(0)

# 生成区间[a, b)内的随机数
def rand(a, b):
    return (b-a)*random.random() + a

# 生成大小 I*J 的矩阵，默认零矩阵 (当然，亦可用 NumPy 提速)
def makeMatrix(I, J, fill=0.0):
    m = []
    for i in range(I):
        m.append([fill]*J)
    return m

# 函数 sigmoid，这里采用 tanh，因为看起来要比标准的 1/(1+e^-x) 漂亮些
def sigmoid(x):
    return math.tanh(x)

# 函数 sigmoid 的派生函数, 为了得到输出 (即：y)
def dsigmoid(y):
    return 1.0 - y**2

class NN:
    ''' 三层反向传播神经网络 '''
    def __init__(self, inputLayer, hiddenLayer, outputLayer):
        # 输入层、隐藏层、输出层的节点（数）
        self.inputLayer = inputLayer + 1 # 增加一个偏差节点
        self.hiddenLayer = hiddenLayer
        self.outputLayer = outputLayer

        # 激活神经网络的所有节点（向量）
        self.inputNodes = [1.0]*self.inputLayer
        self.hiddenNodes = [1.0]*self.hiddenLayer
        self.outputNodes = [1.0]*self.outputLayer

        # 建立权重（矩阵）
        self.weightInput = makeMatrix(self.inputLayer, self.hiddenLayer)
        self.weightOutput = makeMatrix(self.hiddenLayer, self.outputLayer)
        # 设为随机值
        for i in range(self.inputLayer):
            for j in range(self.hiddenLayer):
                self.weightInput[i][j] = rand(-0.2, 0.2)
        for j in range(self.hiddenLayer):
            for k in range(self.outputLayer):
                self.weightOutput[j][k] = rand(-0.2, 0.2)

        # 最后建立动量因子（矩阵）
        self.ci = makeMatrix(self.inputLayer, self.hiddenLayer)
        self.co = makeMatrix(self.hiddenLayer, self.outputLayer)

    def update(self, inputs):
        if len(inputs) != self.inputLayer-1:
            raise ValueError('与输入层节点数不符！')

        # 激活输入层
        for i in range(self.inputLayer-1):
            self.inputNodes[i] = sigmoid(inputs[i]) #High efficiency
            #self.inputNodes[i] = inputs[i] #Low efficiency

        # 激活隐藏层
        for j in range(self.hiddenLayer):
            sum = 0.0
            for i in range(self.inputLayer):
                sum = sum + self.inputNodes[i] * self.weightInput[i][j]
            self.hiddenNodes[j] = sigmoid(sum)

        # 激活输出层
        for k in range(self.outputLayer):
            sum = 0.0
            for j in range(self.hiddenLayer):
                sum = sum + self.hiddenNodes[j] * self.weightOutput[j][k]
            self.outputNodes[k] = sigmoid(sum)

        return self.outputNodes[:]

    def backPropagate(self, targets, N, M):
        ''' 反向传播 '''
        if len(targets) != self.outputLayer:
            raise ValueError('与输出层节点数不符！')

        # 计算输出层的误差
        output_deltas = [0.0] * self.outputLayer
        for k in range(self.outputLayer):
            error = targets[k]-self.outputNodes[k]
            output_deltas[k] = dsigmoid(self.outputNodes[k]) * error

        # 计算隐藏层的误差
        hidden_deltas = [0.0] * self.hiddenLayer
        for j in range(self.hiddenLayer):
            error = 0.0
            for k in range(self.outputLayer):
                error = error + output_deltas[k]*self.weightOutput[j][k]
            hidden_deltas[j] = dsigmoid(self.hiddenNodes[j]) * error

        # 更新输出层权重
        for j in range(self.hiddenLayer):
            for k in range(self.outputLayer):
                change = output_deltas[k]*self.hiddenNodes[j]
                self.weightOutput[j][k] = self.weightOutput[j][k] + N*change + M*self.co[j][k]
                self.co[j][k] = change
                #print(N*change, M*self.co[j][k])

        # 更新输入层权重
        for i in range(self.inputLayer):
            for j in range(self.hiddenLayer):
                change = hidden_deltas[j]*self.inputNodes[i]
                self.weightInput[i][j] = self.weightInput[i][j] + N*change + M*self.ci[i][j]
                self.ci[i][j] = change

        # 计算误差
        error = 0.0
        for k in range(len(targets)):
            error = error + 0.5*(targets[k]-self.outputNodes[k])**2
        return error

    def test(self, patterns):
        for p in patterns:
            print(p[0], '->', self.update(p[0]))

    def weights(self):
        print('输入层权重:')
        for i in range(self.inputLayer):
            print(self.weightInput[i])
        print()
        print('输出层权重:')
        for j in range(self.hiddenLayer):
            print(self.weightOutput[j])

    def train(self, patterns, iterations=1000, N=0.5, M=0.1):
        # N: 学习速率(learning rate)
        # M: 动量因子(momentum factor)
        for i in range(iterations):
            error = 0.0
            for p in patterns:
                inputs = p[0]
                targets = p[1]
                self.update(inputs)
                error = error + self.backPropagate(targets, N, M)
            if i % 100 == 0:
                print('误差 %-.5f' % error)


def demo():
    # 一个演示：教神经网络学习逻辑异或（XOR）------------可以换成你自己的数据试试
    pat = [
        [[0,0], [0]],
        [[0,1], [1]],
        [[1,0], [1]],
        [[1,1], [0]]
    ]

    # 创建一个神经网络：输入层有两个节点、隐藏层有两个节点、输出层有一个节点
    n = NN(2, 2, 1)
    # 用一些模式训练它
    n.train(pat)
    # 测试训练的成果（不要吃惊哦）
    n.test(pat)
    # 看看训练好的权重（当然可以考虑把训练好的权重持久化）
    #n.weights()


if __name__ == '__main__':
    demo()