当前位置:网站首页>Huber Loss
Huber Loss
2022-07-05 08:51:00 【Wanderer001】
Huber Loss 是一个用于回归问题的带参损失函数, 优点是能增强平方误差损失函数(MSE, mean square error)对离群点的鲁棒性。
当预测偏差小于 δ 时,它采用平方误差,当预测偏差大于 δ 时,采用的线性误差。
相比于最小二乘的线性回归,HuberLoss降低了对离群点的惩罚程度,所以 HuberLoss 是一种常用的鲁棒的回归损失函数。
Huber Loss 定义如下
参数 a 通常表示 residuals,写作 y−f(x),当 a = y−f(x) 时,Huber loss 定义为:
δ 是 HuberLoss 的参数,y是真实值,f(x)是模型的预测值, 且由定义可知 Huber Loss 处处可导。
边栏推荐
- JS asynchronous error handling
- C#图像差异对比:图像相减(指针法、高速)
- Use arm neon operation to improve memory copy speed
- 微信H5公众号获取openid爬坑记
- 【日常训练--腾讯精选50】557. 反转字符串中的单词 III
- Blue Bridge Cup provincial match simulation question 9 (MST)
- Pytorch entry record
- 2020 "Lenovo Cup" National College programming online Invitational Competition and the third Shanghai University of technology programming competition
- 使用arm Neon操作,提高内存拷贝速度
- [牛客网刷题 Day4] JZ35 复杂链表的复制
猜你喜欢
Count of C # LINQ source code analysis
Install the CPU version of tensorflow+cuda+cudnn (ultra detailed)
319. Bulb switch
整形的分类:short in long longlong
Solutions of ordinary differential equations (2) examples
Halcon blob analysis (ball.hdev)
2020 "Lenovo Cup" National College programming online Invitational Competition and the third Shanghai University of technology programming competition
Halcon: check of blob analysis_ Blister capsule detection
Guess riddles (6)
Guess riddles (5)
随机推荐
Pearson correlation coefficient
Basic number theory - factors
特征工程
Programming implementation of ROS learning 5-client node
2011-11-21 training record personal training (III)
Bit operation related operations
Beautiful soup parsing and extracting data
我从技术到产品经理的几点体会
C语言标准函数scanf不安全的原因
C [essential skills] use of configurationmanager class (use of file app.config)
[daiy4] jz32 print binary tree from top to bottom
[matlab] matlab reads and writes Excel
Infix expression evaluation
Codeforces Round #648 (Div. 2) D. Solve The Maze
Typescript hands-on tutorial, easy to understand
Meta标签详解
Codeworks round 681 (Div. 2) supplement
12. Dynamic link library, DLL
Business modeling of software model | vision
多元线性回归(梯度下降法)