Probabilistic robust regression with adaptive weights-a case study on face recognition
作者机构：Department of Computer ScienceShanghai Jiao Tong UniversityShanghai200240China
出 版 物：《Frontiers of Computer Science》 (中国计算机科学前沿（英文版）)
年 卷 期：2020年第14卷第5期
页 面：123-134页
核心收录：
学科分类：08[工学] 080203[工学-机械设计及理论] 0802[工学-机械工程]
基 金：We thank our anonymous reviewers for their feedback and suggestions.This work was partially sponsored by the National Basic Research 973 Program of China(2015CB352403) the National Natural Science Foundation of China(NSFC)(Grant Nos.61702328,61602301,61632017)
主 题：robust regression nonconvex loss face recognition
摘 要：Robust regression plays an important role in many machine learning problems.A primal approach relies on the use of Huber loss and an iteratively reweightedℓ2 ***,because the Huber loss is not smooth and its corresponding distribution cannot be represented as a Gaussian scale mixture,such an approach is extremely difficult to handle using a probabilistic *** address those limitations,this paper proposes two novel losses and the corresponding probability *** is called Soft Huber,which is well suited for modeling non-Gaussian *** is Nonconvex Huber,which can help produce much sparser results when imposed as a prior on regression *** can represent anyℓq loss(1/2≤q2)with tuning parameters,which makes the regression model more *** also show that both distributions have an elegant form,which is a Gaussian scale mixture with a generalized inverse Gaussian mixing *** enables us to devise an expectation maximization(EM)algorithm for solving the regression *** can obtain an adaptive weight through EM,which is very useful to remove noise data or irrelevant features in regression *** apply our model to the face recognition problem and show that it not only reduces the impact of noise pixels but also removes more irrelevant face *** experiments demonstrate the promising results on two datasets.
维普期刊数据库