NPD-SG：一种网络上抗噪声的原对偶随机梯度扩散算法

IF 3 3区计算机科学 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

IEEE Transactions on Signal and Information Processing over Networks Pub Date : 2025-08-20 DOI:10.1109/TSIPN.2025.3600760

Jiacheng Wu;Zhengchun Zhou;Sheng Zhang;Hongyu Han

{"title":"NPD-SG：一种网络上抗噪声的原对偶随机梯度扩散算法","authors":"Jiacheng Wu;Zhengchun Zhou;Sheng Zhang;Hongyu Han","doi":"10.1109/TSIPN.2025.3600760","DOIUrl":null,"url":null,"abstract":"In this paper, we develop a noise-resistant primal-dual stochastic gradient-based diffusion algorithm (named NPD-SG) designed to operate effectively in scenarios with link noise. The mean-square analysis indicates that, with enough small step-size <inline-formula><tex-math>$\\mu$</tex-math></inline-formula> and forgetting factor <inline-formula><tex-math>$\\gamma$</tex-math></inline-formula> in (0, 1), the strategy is stable in terms of mean-square error; by reducing the value of <inline-formula><tex-math>$\\gamma$</tex-math></inline-formula>, it is possible to maintain a low level of estimation error. Then, we modify the update step for dual variables to address the numerical accumulation problem, resulting in an improved NPD-SG (INPD-SG) algorithm. The theoretical analysis also reveals the impact of this modification on algorithm performance. Finally, several simulations demonstrate the theoretical findings and the effectiveness of the proposed approaches.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"1087-1099"},"PeriodicalIF":3.0000,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"NPD-SG: A Noise-Resistant Primal-Dual Stochastic Gradient Diffusion Algorithm Over Networks\",\"authors\":\"Jiacheng Wu;Zhengchun Zhou;Sheng Zhang;Hongyu Han\",\"doi\":\"10.1109/TSIPN.2025.3600760\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In this paper, we develop a noise-resistant primal-dual stochastic gradient-based diffusion algorithm (named NPD-SG) designed to operate effectively in scenarios with link noise. The mean-square analysis indicates that, with enough small step-size <inline-formula><tex-math>$\\\\mu$</tex-math></inline-formula> and forgetting factor <inline-formula><tex-math>$\\\\gamma$</tex-math></inline-formula> in (0, 1), the strategy is stable in terms of mean-square error; by reducing the value of <inline-formula><tex-math>$\\\\gamma$</tex-math></inline-formula>, it is possible to maintain a low level of estimation error. Then, we modify the update step for dual variables to address the numerical accumulation problem, resulting in an improved NPD-SG (INPD-SG) algorithm. The theoretical analysis also reveals the impact of this modification on algorithm performance. Finally, several simulations demonstrate the theoretical findings and the effectiveness of the proposed approaches.\",\"PeriodicalId\":56268,\"journal\":{\"name\":\"IEEE Transactions on Signal and Information Processing over Networks\",\"volume\":\"11 \",\"pages\":\"1087-1099\"},\"PeriodicalIF\":3.0000,\"publicationDate\":\"2025-08-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE Transactions on Signal and Information Processing over Networks\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://ieeexplore.ieee.org/document/11130908/\",\"RegionNum\":3,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"ENGINEERING, ELECTRICAL & ELECTRONIC\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Signal and Information Processing over Networks","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/11130908/","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}

引用次数: 0

摘要

在本文中，我们开发了一种抗噪声的基于原始-对偶随机梯度的扩散算法（命名为NPD-SG），旨在有效地在有链路噪声的情况下运行。均方分析表明，当（0,1）中的步长$\mu$和遗忘因子$\gamma$足够小时，该策略在均方误差方面是稳定的；通过减小$\gamma$的值，可以保持较低的估计误差水平。然后，我们修改了对偶变量的更新步骤，以解决数值积累问题，从而得到改进的NPD-SG （INPD-SG）算法。理论分析也揭示了这种修改对算法性能的影响。最后，通过仿真验证了理论结果和所提方法的有效性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

NPD-SG: A Noise-Resistant Primal-Dual Stochastic Gradient Diffusion Algorithm Over Networks

In this paper, we develop a noise-resistant primal-dual stochastic gradient-based diffusion algorithm (named NPD-SG) designed to operate effectively in scenarios with link noise. The mean-square analysis indicates that, with enough small step-size

$\mu$

and forgetting factor

$\gamma$

in (0, 1), the strategy is stable in terms of mean-square error; by reducing the value of

$\gamma$

, it is possible to maintain a low level of estimation error. Then, we modify the update step for dual variables to address the numerical accumulation problem, resulting in an improved NPD-SG (INPD-SG) algorithm. The theoretical analysis also reveals the impact of this modification on algorithm performance. Finally, several simulations demonstrate the theoretical findings and the effectiveness of the proposed approaches.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

IEEE Transactions on Signal and Information Processing over Networks Computer Science-Computer Networks and Communications

CiteScore

5.80

自引率

12.50%

发文量

期刊介绍： The IEEE Transactions on Signal and Information Processing over Networks publishes high-quality papers that extend the classical notions of processing of signals defined over vector spaces (e.g. time and space) to processing of signals and information (data) defined over networks, potentially dynamically varying. In signal processing over networks, the topology of the network may define structural relationships in the data, or may constrain processing of the data. Topics include distributed algorithms for filtering, detection, estimation, adaptation and learning, model selection, data fusion, and diffusion or evolution of information over such networks, and applications of distributed signal processing.