Globally-convergent Iteratively Reweighted Least Squares for Robust Regression Problems

被引:0
|
作者
Mukhoty, Bhaskar [1 ]
Gopakumar, Govind [1 ]
Jain, Prateek [2 ]
Kar, Purushottam [1 ]
机构
[1] IIT Kanpur, Kanpur, Uttar Pradesh, India
[2] Microsoft Res India, Bengaluru, India
关键词
MINIMIZATION;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We provide the first global model recovery results for the IRLS (iteratively reweighted least squares) heuristic for robust regression problems. IRLS is known to offer excellent performance, despite bad initializations and data corruption, for several parameter estimation problems. Existing analyses of IRLS frequently require careful initialization, thus offering only local convergence guarantees. We remedy this by proposing augmentations to the basic IRLS routine that not only offer guaranteed global recovery, but in practice also outperform state-of-the-art algorithms for robust regression. Our routines are more immune to hyperparameter misspecification in basic regression tasks, as well as applied tasks such as linear-armed bandit problems. Our theoretical analyses rely on a novel extension of the notions of strong convexity and smoothness to weighted strong convexity and smoothness, and establishing that sub-Gaussian designs offer bounded weighted condition numbers. These notions may be useful in analyzing other algorithms as well.
引用
收藏
页码:313 / 322
页数:10
相关论文
共 50 条
  • [31] INFLUENCE FUNCTIONS OF ITERATIVELY REWEIGHTED LEAST-SQUARES ESTIMATORS
    DOLLINGER, MB
    STAUDTE, RG
    [J]. JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1991, 86 (415) : 709 - 716
  • [32] ROBUST REGRESSION USING ITERATIVELY RE-WEIGHTED LEAST-SQUARES
    HOLLAND, PW
    WELSCH, RE
    [J]. COMMUNICATIONS IN STATISTICS PART A-THEORY AND METHODS, 1977, 6 (09): : 813 - 827
  • [33] Detection and correction of poorly converged optimizations by iteratively reweighted least squares
    Kim, HM
    Papila, M
    Mason, WH
    Haftka, RT
    Watson, LT
    Grossman, B
    [J]. COLLECTION OF THE 41ST AIAA/ASME/ASCE/AHS/ASC STRUCTURES, STRUCTURAL DYNAMICS, AND MATERIALS CONFERENCE AND EXHIBIT, VOL 1 PTS 1-3, 2000, : 1159 - 1176
  • [34] Preconditioning for Accelerated Iteratively Reweighted Least Squares in Structured Sparsity Reconstruction
    Chen, Chen
    Huang, Junzhou
    He, Lei
    Li, Hongsheng
    [J]. 2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, : 2713 - 2720
  • [35] Improved iteratively reweighted least squares algorithms for sparse recovery problem
    Liu, Yufeng
    Zhu, Zhibin
    Zhang, Benxin
    [J]. IET IMAGE PROCESSING, 2022, 16 (05) : 1324 - 1340
  • [36] Baseline correction using adaptive iteratively reweighted penalized least squares
    Zhang, Zhi-Min
    Chen, Shan
    Liang, Yi-Zeng
    [J]. ANALYST, 2010, 135 (05) : 1138 - 1146
  • [37] Fast General Norm Approximation via Iteratively Reweighted Least Squares
    Samejima, Masaki
    Matsushita, Yasuyuki
    [J]. COMPUTER VISION - ACCV 2016 WORKSHOPS, PT II, 2017, 10117 : 207 - 221
  • [38] Impulsive noise elimination using polynomial iteratively reweighted least squares
    Kuruoglu, EE
    Rayner, PJW
    Fitzgerald, WJ
    [J]. 1996 IEEE DIGITAL SIGNAL PROCESSING WORKSHOP, PROCEEDINGS, 1996, : 347 - 350
  • [40] Robust Low-Rank Matrix Factorization via Block Iteratively Reweighted Least-Squares
    Li, Nanxi
    He, Zhen-Qing
    Zhu, Jianchi
    She, Xiaoming
    Chen, Peng
    Liang, Ying-Chang
    [J]. 2022 14TH INTERNATIONAL CONFERENCE ON WIRELESS COMMUNICATIONS AND SIGNAL PROCESSING, WCSP, 2022, : 725 - 730