基于一致遍历马氏链样本的分布式Huber回归算法的泛化性

doi:10.13438/j.cnki.jdzk.2026.04.003

吉首大学学报（自然科学版） ›› 2026, Vol. 47 ›› Issue (4): 11-18.DOI: 10.13438/j.cnki.jdzk.2026.04.003

基于一致遍历马氏链样本的分布式Huber回归算法的泛化性

李珂，姜宏伟

(沈阳工业大学理学院，辽宁沈阳 110870)

出版日期:2026-07-25 发布日期:2026-08-06
作者简介:李珂（2001—），女，河南濮阳人，沈阳工业大学理学院硕士研究生，主要从事统计学习理论研究
基金资助:
国家自然科学基金资助项目(42171351)

Generalization of Distributed Huber Regression Algorithms with Uniformly Ergodic Markov Chain Samples

LI Ke,JIANG Hongwei

(School of Science,Shenyang University of Technology,Shenyang 110870,China)

Online:2026-07-25 Published:2026-08-06

摘要/Abstract

摘要：讨论了基于一致遍历马氏链样本的分布式Huber回归算法的泛化性.建立了基于一致遍历马氏链样本的分布式Huber回归算法，并采用统计学习理论方法，推导出该算法的收敛速率和泛化界，证明了在非独立同分布样本的假设下算法具有较快的收敛速率.

关键词： 分布式, Huber回归算法, 一致遍历马氏链样本, 收敛速率

Abstract: This paper investigates the generalization performance of distributed Huber regression with samples generated from uniformly ergodic Markov chains.We first construct a distributed Huber regression algorithm applicable to uniformly ergodic Markov chain samples.Relying on statistical learning theory,we further derive the convergence rates and generalization bounds for the proposed distributed Huber regression with Markov chain sampling.Theoretical proofs demonstrate that the distributed Huber regression algorithm achieves a fast convergence rate even in the non-independent and identically distributed (non-i.i.d.) sample setting.

Key words: distributed, huber regression algorithm, uniformly ergodic Markov chain samples, convergence rates

李珂, 姜宏伟. 基于一致遍历马氏链样本的分布式Huber回归算法的泛化性[J]. 吉首大学学报（自然科学版）, 2026, 47(4): 11-18.

LI Ke, JIANG Hongwei. Generalization of Distributed Huber Regression Algorithms with Uniformly Ergodic Markov Chain Samples[J]. Journal of Jishou University(Natural Sciences Edition), 2026, 47(4): 11-18.

参考文献

[1] Davies P L.Aspects of Robust Linear Regression[J].The Annals of Statistics,1993,21(4):1843-1899.
[2] Huber P J.Robust Estimation of a Location Parameter[J].The Annals of Mathematical Statistics,1964,35(1):73-101.
[3] Gong Tieliang,Dong Yuxin,Chen Hong,et al.Markov Subsampling Based on Huber Criterion[J].IEEE Transactions on Neural Networks and Learning Systems,2022,35(2):2250-2262.
[4] 雷云祥，蒋其海，田杰，等.分布式架构下权限控制系统的设计与优化[J].吉首大学学报(自然科学版)，2022，43(6)：26-31.
[5] Yang Fan,Li Jinfeng,Cheng J.Husky:Towards a More Efficient and Expressive Distributed Computing Framework[J].Proceedings of the VLDB Endowment,2016,9(5):420-431.
[6] Chen Xueying,Xie Minge.A Split-and-Conquer Approach for Analysis of Extraordinarily Large Data[J].Statistica Sinica,2014,24(4):1655-1684.
[7] Lin Nan,Xi Ruibin.Aggregated Estimating Equation Estimation[J].Statistics and Its Interface,2011,4(1):73-83.
[8] Pu Shi,Nedic' A.Distributed Stochastic Gradient Tracking Methods[J].Mathematical Programming,2021,187(1):409-457.
[9] Kar P,Sriperumbudur B K,Jain P,et al.On the Generalization Ability of Online Learning Algorithms for Pairwise Loss Functions[J].Eprint Arxiv,2013,10(1):441-449.
[10] Ying Yiming,Zhou Dingxuan.Online Pairwise Learning Algorithms[J].Neural Computation,2016,28(4):743-777.
[11] Zou Bin,Tang Yuanyan,Xu Zongben,et al.The Generalization Performance of Regularized Regression Algorithms Based on Markov Sampling[J].IEEE Transactions on Cybernetics,2014,44(9):1497-1507.
[12] Wang Zhan,Liang Ting,Zou Bin,et al.Incremental Fisher Linear Discriminant Based on Data Denoising[J].Knowledge-Based Systems,2022,237:107799.
[13] Vapnik V N.Statistical Learning Theory[M].New York:John Wiley,1998:219-245.
[14] Wu Qiang,Ying Yiming,Zhou Dingxuan.Learning Rates of Least-Square Regularized Regression[J].Foundations of Computational Mathematics,2006,6:171-192.
[15] Feng Yunlong,Wu Qiang.A Statistical Learning Assessment of Huber Regression[J].Journal of Approximation Theory,2022,273:105660.

基于一致遍历马氏链样本的分布式Huber回归算法的泛化性

Generalization of Distributed Huber Regression Algorithms with Uniformly Ergodic Markov Chain Samples

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics

本文评价