神经网络自定义（了解原理）

数据挖掘课讲到了神经网络，让自己写一个简单的例子，笔者觉得非常好，特别有助于加强对神经网络的认识，所以来做一个小小的分享，让我们一起共同进步吧。

1.问题描述：

2.代码展示

dxdy.py

'''
coding:utf-8
@author: Li Sentan
@time:2021.10.24
@file:dxdy.py
'''

from layer_naive import *

x = eval(input("Please input the value of x:"))
y = eval(input("Please input the value of y:"))
x = x*math.pi

threex = MulLayer()
sin = sinLayer()
netwoy = MulLayer()
e = eLayer()
twoadd = AddLayer()
div = divLayer()

# forward
threex1 = threex.forward(3,x)
sin1 = sin.forward(threex1)
netwoy1 = netwoy.forward(-2,y)
e1 =e.forward((netwoy1))
twoadd1 = twoadd.forward(sin1,e1)
div1 = div.forward(twoadd1,8)
result = min(5,div1)

# backward
dresult = 1
dtwoadd,dy= div.backward(dout = dresult)
dsin,de = twoadd.backward(dtwoadd)
dnetwoy = e.backward(de)
dx2,dy = netwoy.backward(dnetwoy)
dthreex = sin.backward(dsin)
dy2,dx  = threex.backward(dthreex)


print("result:", result)
print("dx:", int(round(dx,0)))
print("dy:", int(round(dy,0)))

 layer_naive.py

'''
coding:utf-8
@author: Li Sentan
@time:2021.10.24
@file:layer_naive.py
'''
import math

class sinLayer:
    def __init__(self):
        pass

    def forward(self,x):
        self.x = x
        out = math.sin(x)
        return out

    def backward(self, dout):
        dx = dout * math.cos(self.x)
        return dx


class divLayer:
    def __init__(self):
        pass

    def forward(self, x, y):
        self.x = x
        self.y = y
        out = y/x
        return out

    def backward(self, dout):
        dx = dout * (-self.y)/(self.x**2)
        dy = dout * (1/self.x)
        return dx, dy


class eLayer:
    def __init__(self):
        pass

    def forward(self, x):
        self.x = x
        out = math.exp(x)
        return out

    def backward(self, dout):
        dx = dout * math.exp(self.x)
        return dx


class MulLayer:
    def __init__(self):
        self.x = None
        self.y = None

    def forward(self, x, y):
        self.x = x
        self.y = y                
        out = x * y
        return out

    def backward(self, dout):
        dx = dout * self.y
        dy = dout * self.x
        return dx, dy


class AddLayer:
    def __init__(self):
        pass

    def forward(self, x, y):
        out = x + y
        return out

    def backward(self, dout):
        dx = dout * 1
        dy = dout * 1
        return dx, dy

3.结果展示：

结果跟计算一致。

从中我们可以感受到，通俗点概述，我们就是将一个复合函数分成一个个简单函数，forward就是向一个个简单函数按顺序代入值， backward就是反顺序对一个个简单函数进行求导，然后进行乘积，跟我们平时的对复合函数的求导思想一致。哈哈哈，放个好玩的图，深度学习 so easy!