梯度法

最新推荐文章于 2023-02-24 01:00:00 发布

caojinpei123

最新推荐文章于 2023-02-24 01:00:00 发布

阅读量441

点赞数

本文链接：https://blog.csdn.net/caojinpei123/article/details/102874498

版权

# -*- coding: utf-8 -*-
"""
Spyder Editor

This is a temporary script file.
"""

import numpy as np

def function(x):
    return x[0]**2+x[1]**2

def numerical_gradient(x,f):
    h = 0.0001
    grad = np.zeros_like(x)
    for index in range(x.size):#x.size()
        temp_x = x[index]
        x[index] = temp_x + h
        fxh1= f(x)
        
        x[index] = temp_x - h
        fxh2 = f(x)
        grad[index] = (fxh1 - fxh2)/(2*h) 
        x[index] = temp_x
    return grad
        
        


        
    
        

def gradient_descent(init_x,y, lr = 0.1, step_num=100):
    x = init_x
    for i in range(step_num):
        grad = numerical_gradient(x, y)
        x = x - lr *grad
    
    return x

if __name__ == "__main__":
    int_x = np.array([-3.0, 4.0])
    c = gradient_descent(int_x,function, lr = 0.01, step_num=100)
    print(c)