随机梯度下降法

最新推荐文章于 2024-09-15 22:31:42 发布

W-WEARY

最新推荐文章于 2024-09-15 22:31:42 发布

阅读量696

点赞数

分类专栏：笔记文章标签： python

本文链接：https://blog.csdn.net/qq_52141257/article/details/121281778

版权

笔记专栏收录该内容

7 篇文章 0 订阅

订阅专栏

import random

matrix_A = [[1,4], [2,5], [5,1], [4,2]]
Matrix_y = [19,26,19,20]
theta = [2,5]
#学习速率
leraing_rate = 0.005
loss = 50
iters = 1
Eps = 0.0001

#随机梯度下降
while loss>Eps and iters <1000 :
    loss = 0
    i = random.randint(0, 3)
    h = theta[0]*matrix_A[i][0] + theta[1]*matrix_A[i][1] 
    theta[0] = theta[0] + leraing_rate*(Matrix_y[i]-h)*matrix_A[i][0]
    theta[1] = theta[1] + leraing_rate*(Matrix_y[i]-h)*matrix_A[i][1]
    Error = 0
    Error = theta[0]*matrix_A[i][0] + theta[1]*matrix_A[i][1] - Matrix_y[i]
    Error = Error*Error
    loss = loss +Error
    iters = iters +1
print ('theta=',theta)
print ('iters=',iters)

#梯度下降
while loss>Eps and iters <1000 :
    loss = 0
    for i in range(4):
        h = theta[0]*matrix_A[i][0] + theta[1]*matrix_A[i][1] 
        theta[0] = theta[0] + leraing_rate*(Matrix_y[i]-h)*matrix_A[i][0]
        theta[1] = theta[1] + leraing_rate*(Matrix_y[i]-h)*matrix_A[i][1]
    for i in range(4):
        Error = 0
        Error = theta[0]*matrix_A[i][0] + theta[1]*matrix_A[i][1] - Matrix_y[i]
        Error = Error*Error
        loss = loss +Error
    iters = iters +1
print ('theta=',theta)
print ('iters=',iters)

#批量梯度下降
while loss>Eps and iters <1000 :
    loss = 0
    sampleindex =  random.sample([0,1,2,3],2)
    for i in sampleindex :
        h = theta[0]*matrix_A[i][0] + theta[1]*matrix_A[i][1] 
        theta[0] = theta[0] + leraing_rate*(Matrix_y[i]-h)*matrix_A[i][0]
        theta[1] = theta[1] + leraing_rate*(Matrix_y[i]-h)*matrix_A[i][1]
    for i in sampleindex :
        Error = 0
        Error = theta[0]*matrix_A[i][0] + theta[1]*matrix_A[i][1] - Matrix_y[i]
        Error = Error*Error
        loss = loss +Error
    iters = iters +1
print ('theta=',theta)
print ('iters=',iters)

在这里插入图片描述