基于拍卖的联合学习的成本意识效用最大化投标策略。

IF 10.2 1区 计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
Xiaoli Tang, Han Yu
{"title":"基于拍卖的联合学习的成本意识效用最大化投标策略。","authors":"Xiaoli Tang, Han Yu","doi":"10.1109/TNNLS.2024.3474102","DOIUrl":null,"url":null,"abstract":"<p><p>Auction-based federated learning (AFL) has emerged as an efficient and fair approach to incentivize data owners (DOs) to contribute to federated model training, garnering extensive interest. However, the important problem of helping data consumers (DCs) bid for DOs in competitive AFL settings remains open. Existing work simply treats that the actual cost paid by a winning DC (i.e., the bid cost) is equal to the bid price offered by that DC itself. However, this assumption is inconsistent with the widely adopted generalized second-price (GSP) auction mechanism used in AFL, including in these existing works. Under a GSP auction, the winning DC does not pay its own proposed bid price. Instead, the bid cost for the winner is determined by the second-highest bid price among all participating DCs. To address this limitation, we propose a first-of-its-kind federated cost-aware bidding strategy () to help DCs maximize their utility under GSP auction-based federated learning (FL). It enables DCs to efficiently bid for DOs in competitive AFL markets, maximizing their utility and improving the resulting FL model accuracy. We first formulate the optimal bidding function under the GSP auction setting, and then demonstrate that it depends on utility estimation and market price modeling, which are interrelated. Based on this analysis, jointly optimizes in a novel end-to-end framework, and then executes the proposed return on investment (ROI)-based method to determine the optimal bid price for each piece of the data resource. Through extensive experiments on six commonly adopted benchmark datasets, we show that outperforms eight state-of-the-art methods, beating the best baseline by 4.39%, 4.56%, 1.33%, and 5.43% on average in terms of the total amount of data obtained, number of data samples per unit cost, total utility, and FL model accuracy, respectively.</p>","PeriodicalId":13303,"journal":{"name":"IEEE transactions on neural networks and learning systems","volume":"PP ","pages":""},"PeriodicalIF":10.2000,"publicationDate":"2024-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"A Cost-Aware Utility-Maximizing Bidding Strategy for Auction-Based Federated Learning.\",\"authors\":\"Xiaoli Tang, Han Yu\",\"doi\":\"10.1109/TNNLS.2024.3474102\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Auction-based federated learning (AFL) has emerged as an efficient and fair approach to incentivize data owners (DOs) to contribute to federated model training, garnering extensive interest. However, the important problem of helping data consumers (DCs) bid for DOs in competitive AFL settings remains open. Existing work simply treats that the actual cost paid by a winning DC (i.e., the bid cost) is equal to the bid price offered by that DC itself. However, this assumption is inconsistent with the widely adopted generalized second-price (GSP) auction mechanism used in AFL, including in these existing works. Under a GSP auction, the winning DC does not pay its own proposed bid price. Instead, the bid cost for the winner is determined by the second-highest bid price among all participating DCs. To address this limitation, we propose a first-of-its-kind federated cost-aware bidding strategy () to help DCs maximize their utility under GSP auction-based federated learning (FL). It enables DCs to efficiently bid for DOs in competitive AFL markets, maximizing their utility and improving the resulting FL model accuracy. We first formulate the optimal bidding function under the GSP auction setting, and then demonstrate that it depends on utility estimation and market price modeling, which are interrelated. Based on this analysis, jointly optimizes in a novel end-to-end framework, and then executes the proposed return on investment (ROI)-based method to determine the optimal bid price for each piece of the data resource. Through extensive experiments on six commonly adopted benchmark datasets, we show that outperforms eight state-of-the-art methods, beating the best baseline by 4.39%, 4.56%, 1.33%, and 5.43% on average in terms of the total amount of data obtained, number of data samples per unit cost, total utility, and FL model accuracy, respectively.</p>\",\"PeriodicalId\":13303,\"journal\":{\"name\":\"IEEE transactions on neural networks and learning systems\",\"volume\":\"PP \",\"pages\":\"\"},\"PeriodicalIF\":10.2000,\"publicationDate\":\"2024-11-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE transactions on neural networks and learning systems\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://doi.org/10.1109/TNNLS.2024.3474102\",\"RegionNum\":1,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE transactions on neural networks and learning systems","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1109/TNNLS.2024.3474102","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

摘要

基于拍卖的联合学习(AFL)是激励数据所有者(DOs)为联合模型训练做出贡献的一种高效、公平的方法,因而受到广泛关注。然而,帮助数据消费者(DCs)在竞争性 AFL 环境中竞标 DOs 这一重要问题仍未解决。现有工作简单地认为,中标数据消费者支付的实际成本(即投标成本)等于该数据消费者自己提出的投标价格。然而,这一假设与 AFL 中广泛采用的通用第二价格(GSP)拍卖机制不一致,包括在这些现有著作中。在 GSP 拍卖机制下,获胜的区议会并不支付自己提出的出价。相反,获胜者的投标成本由所有参与竞拍的 DC 中第二高的投标价格决定。为解决这一局限性,我们首次提出了一种联合成本感知投标策略(),帮助区委在基于 GSP 拍卖的联合学习(FL)中实现效用最大化。该策略使区委能够在竞争激烈的联合学习市场上有效地竞标指定经营者,从而实现其效用最大化,并提高联合学习模型的准确性。我们首先提出了 GSP 拍卖设置下的最优投标函数,然后证明它取决于效用估计和市场价格建模,而这两者是相互关联的。在此分析的基础上,我们在一个新颖的端到端框架中进行了联合优化,然后执行了所提出的基于投资回报率(ROI)的方法,以确定每块数据资源的最优投标价格。通过在六个常用的基准数据集上进行大量实验,我们发现该方法优于八种最先进的方法,在获得的数据总量、单位成本的数据样本数量、总效用和 FL 模型准确性方面分别比最佳基准方法平均高出 4.39%、4.56%、1.33% 和 5.43%。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
A Cost-Aware Utility-Maximizing Bidding Strategy for Auction-Based Federated Learning.

Auction-based federated learning (AFL) has emerged as an efficient and fair approach to incentivize data owners (DOs) to contribute to federated model training, garnering extensive interest. However, the important problem of helping data consumers (DCs) bid for DOs in competitive AFL settings remains open. Existing work simply treats that the actual cost paid by a winning DC (i.e., the bid cost) is equal to the bid price offered by that DC itself. However, this assumption is inconsistent with the widely adopted generalized second-price (GSP) auction mechanism used in AFL, including in these existing works. Under a GSP auction, the winning DC does not pay its own proposed bid price. Instead, the bid cost for the winner is determined by the second-highest bid price among all participating DCs. To address this limitation, we propose a first-of-its-kind federated cost-aware bidding strategy () to help DCs maximize their utility under GSP auction-based federated learning (FL). It enables DCs to efficiently bid for DOs in competitive AFL markets, maximizing their utility and improving the resulting FL model accuracy. We first formulate the optimal bidding function under the GSP auction setting, and then demonstrate that it depends on utility estimation and market price modeling, which are interrelated. Based on this analysis, jointly optimizes in a novel end-to-end framework, and then executes the proposed return on investment (ROI)-based method to determine the optimal bid price for each piece of the data resource. Through extensive experiments on six commonly adopted benchmark datasets, we show that outperforms eight state-of-the-art methods, beating the best baseline by 4.39%, 4.56%, 1.33%, and 5.43% on average in terms of the total amount of data obtained, number of data samples per unit cost, total utility, and FL model accuracy, respectively.

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 neural networks and learning systems
IEEE transactions on neural networks and learning systems COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE-COMPUTER SCIENCE, HARDWARE & ARCHITECTURE
CiteScore
23.80
自引率
9.60%
发文量
2102
审稿时长
3-8 weeks
期刊介绍: The focus of IEEE Transactions on Neural Networks and Learning Systems is to present scholarly articles discussing the theory, design, and applications of neural networks as well as other learning systems. The journal primarily highlights technical and scientific research in this domain.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信