Python实现单隐层神经网络

import random
import math

#---神经网络Model---
class Ann:
#构造函数 初始化模型参数
def __init__(self, i_num, h_num, o_num):
#可调参数
self.learn_rate = 0.1    #学习率
self.num_long = 2        #输出结果位数
self.random_long = 10    #随机种子位数

#输入参数
self.input_num = i_num   #输入层 数量
self.hidden_num = h_num  #隐层 数量
self.output_num = o_num  #输出层 数量

#模型参数
self.input = []          #输入层
self.hidden = []         #隐层
self.output = []         #输出层
self.error = []          #误差
self.expectation = []    #期望
self.weight_ih = self.__ini_weight(self.input_num, self.hidden_num)   #输入层->隐层 连接权
self.weight_ho = self.__ini_weight(self.hidden_num, self.output_num)  #隐层->输出层 连接权
self.threshold_h = self.__ini_threshold(self.hidden_num)              #隐层 阈值
self.threshold_o = self.__ini_threshold(self.output_num)              #输出层 阈值

#初始连接权生成器
def __ini_weight(self, x, y):
result = []
long = math.pow(10, self.random_long)
for i in range(0, x, 1):
res = []
for j in range(0, y, 1):
num = round(random.randint(-1*long,long)/long, self.random_long)
res.insert(j, num)
result.insert(i, res)
return result

#初始阈值生成器
def __ini_threshold(self, n):
result = []
long = pow(10, self.random_long)
for i in range(0, n, 1):
num = round(random.randint(-1*long,long)/long, self.random_long)
result.insert(i, num)
return result

#激励函数 sigma
def excitation(self, value):
sigma = 1/(1+(math.exp(-1*value)))
return sigma

#输入数据
def input_param(self, data, expectation = []):
self.input = []
for value in data:
self.input.append(value)
if(expectation):
self.expectation = []
for value in expectation:
self.expectation.append(value)

#隐层计算
def count_hidden(self):
self.hidden = []
for h in range(0, self.hidden_num, 1):
Hval = 0
for i in range(len(self.input)):
Hval += self.input[i] * self.weight_ih[i][h]
Hval = self.excitation(Hval+self.threshold_h[h])
self.hidden.insert(h, Hval)

#输出层计算
def count_output(self):
self.output = []
for o in range(0, self.output_num, 1):
Oval = 0
for h in range(len(self.hidden)):
Oval += self.hidden[h] * self.weight_ho[h][o]
Oval += self.threshold_o[o]
Oval = round(Oval, self.num_long)
self.output.insert(o, Oval)

#误差计算
def count_error(self):
self.error = []
for key in range(len(self.output)):
self.error.insert(key, self.expectation[key] - self.output[key])

#连接权反馈训练 输入层->隐层
def train_weight_ih(self):
for i in range(len(self.weight_ih)):
for h in range(len(self.weight_ih[i])):
tmp = 0
for o in range(0, self.output_num, 1):
tmp += self.weight_ho[h][o] * self.error[o]
self.weight_ih[i][h] = self.weight_ih[i][h] + self.learn_rate * self.hidden[h] * (1 - self.hidden[h]) * self.input[i] * tmp

#连接权反馈训练 隐层->输出层
def train_weight_ho(self):
for h in range(len(self.weight_ho)):
for o in range(len(self.weight_ho[h])):
self.weight_ho[h][o] = self.weight_ho[h][o] + self.learn_rate * self.hidden[h] * self.error[o]

#阈值反馈训练 隐层
def train_threshold_h(self):
for h in range(len(self.threshold_h)):
tmp = 0
for o in range(0, self.output_num, 1):
tmp += self.weight_ho[h][o] * self.error[o]
self.threshold_h[h] = self.threshold_h[h] + self.learn_rate * self.hidden[h] * (1 - self.hidden[h]) * tmp

#阈值反馈训练 输出层
def train_threshold_o(self):
for o in range(len(self.threshold_o)):
self.threshold_o[o] = self.threshold_o[o] + self.error[o]

#反馈训练
def train(self):
self.train_weight_ih()
self.train_weight_ho()
self.train_threshold_h()
self.train_threshold_o()

#归一化函数
def normal_num(self, max, min, data):
data = (data - min)/(max - min)
return data

#---业务部分(示例)---

#要训练的规则，输入两个值，如果两值相等返回[1,0]，反之返回[0,1]
def testFunc(val):
if(val[0] == val[1]):
return [1,0]
else:
return [0,1]

#构造神经网络模型
ann = Ann(2,3,2)

#生成训练数据，随机生成5000组[0,1][1,0][1,1][0,0]随机数组
data = []
for i in range(0, 10000, 1):
x = random.randint(0,1)
y = random.randint(0,1)
data.append([x,y])

#取得训练数据中的最大值和最小值
for i in range(len(data)):
for j in range(len(data[i])):
if(i == 0 and j == 0):
max = min = data[i][j]
elif(data[i][j] > max):
max = data[i][j]
elif(data[i][j] < min):
min = data[i][j]

#训练数据归一化
dataNormal = []
for i in range(len(data)):
dataNormal.insert(i, [])
for j in range(len(data[i])):
dataNormal[i].append(ann.normal_num(max, min, data[i][j]))

#计算训练数据期望值，并进行反馈训练
for i in range(len(data)):
#计算期望值
exp = testFunc(data[i])
#输入训练数据与期望
ann.input_param(dataNormal[i], exp)
#计算隐层
ann.count_hidden()
#计算输出层
ann.count_output()
#计算误差
ann.count_error()
#反馈训练
ann.train()

#生成测试数据，随机生成20组
testdata = []
for i in range(0, 20, 1):
x = random.randint(0,1)
y = random.randint(0,1)
testdata.append([x,y])

#进行测试，同时输出神经网络预测值与实际期望值
for i in range(len(testdata)):
exp = testFunc(testdata[i])
ann.input_param(testdata[i])
ann.count_hidden()
ann.count_output()
print("Ann:")
print(ann.output)
print("Exp:")
print(exp)
print("\r")