Least core：联邦学习中高效稳定的参与者贡献评价机制

林国鹏; 倪郑威

doi:10.11959/j.issn.1000-0801.2026011

您当前的位置：

首页 >

文章列表页 >

Least core：联邦学习中高效稳定的参与者贡献评价机制

研究与开发 | 更新时间：2026-04-08

- Least core：联邦学习中高效稳定的参与者贡献评价机制
- Least core: an efficient and stable participant contribution evaluation mechanism in federated learning
- 电信科学 2026年42卷第3期页码：97-112
- 作者机构：
  
  浙江工商大学信息与电子工程学院，浙江杭州 310018
- 作者简介：
  
  [ "林国鹏（2000- ），男，浙江工商大学信息与电子工程学院硕士生，主要研究方向为联邦学习。" ]
  [ "倪郑威（1989- ），男，博士，浙江工商大学信息与电子工程学院副研究员，主要研究方向为机器学习、物联网、无线通信等。" ]
- 基金信息：
  
  桐乡市通用人工智能研究院项目(TAGI2-B-2024-0017)
- DOI：10.11959/j.issn.1000-0801.2026011
  中图分类号： TN91;TP309
- 收稿：2025-04-16，
  
  修回：2025-06-17，
  
  录用：2025-06-17，
  
  纸质出版：2026-03-20
- 稿件说明：
移动端阅览
林国鹏,倪郑威.Least core：联邦学习中高效稳定的参与者贡献评价机制[J].电信科学,2026,42(03):97-112.

Lin Guopeng,Ni Zhengwei.Least core: an efficient and stable participant contribution evaluation mechanism in federated learning[J].Telecommunications Science,2026,42(03):97-112.
林国鹏,倪郑威.Least core：联邦学习中高效稳定的参与者贡献评价机制[J].电信科学,2026,42(03):97-112. DOI： 10.11959/j.issn.1000-0801.2026011.

Lin Guopeng,Ni Zhengwei.Least core: an efficient and stable participant contribution evaluation mechanism in federated learning[J].Telecommunications Science,2026,42(03):97-112. DOI： 10.11959/j.issn.1000-0801.2026011.

摘要

为了应对数据隐私保护引发的“数据孤岛”问题，联邦学习提供了一种允许多方在不共享原始数据的情况下联合训练模型的技术框架。然而，为了吸引高质量数据拥有者参与并维持系统的长期稳定运行，公平合理的价值评估和收益分配机制至关重要。尽管Shapley值被广泛用于联邦学习中的贡献评估，但其单一性限制了对价值的多维考量。为此，引入最小核心（Least core）方法，通过最小化最大不满度，为价值分配提供新视角。然而，Least core方法的计算复杂性与Shapley值相似，均存在一定的计算限制，需进一步结合多种抽样方式，近似计算分配结果，在显著降低大规模场景下计算成本的同时，确保分配的准确性。实验结果表明，该机制在确保公平性和系统稳定性的同时，有效提升了计算效率，为多样化的联邦学习场景提供了可行的解决方案。

Abstract

To address the “data silo” problem caused by data privacy protection

federated learning offers a technological framework that enables multiple parties to collaboratively train models without sharing raw data. However

to attract high-quality data owners and ensure the long-term stable operation of the system

a fair and reasonable value evaluation and revenue distribution mechanism is essential. Although the Shapley value has been widely used in federated learning for contribution evaluation

relying on a single approach limits the multi-perspective assessment of data contribution and value distribution. For this purpose

the Least core method was introduced

providing a new perspective for value allocation by minimizing the maximum deficit. However

the computational complexity of the Least core method

similar to that of the Shapley value

posed certain computational limitations. By further combining various sampling methods to approximate the allocation results

it significantly reduced the computational cost in large-scale scenarios while ensuring the accuracy of the allocation. Experimental results demonstrate that the mechanism ensures fairness and system stability while effectively improving computational efficiency

providing a feasible solution for diverse federated learning scenarios.

关键词

Keywords

references

Magalhes S T D . The European Union’s general data protection regulation (GDPR) [R ] . 2020 .

Glocker F . Der California consumer privacy act [R ] . 2022 .

中华人民共和国个人信息保护法 [J ] . 中华人民共和国公安部公报 , 2021 ( 4 ): 28 - 36 .

Personal information protection law of the People’s Republic of China [J ] . Gazette of the Ministry of Public Security of the People’s Republic of China , 2021 ( 4 ): 28 – 36 .

Mcmahan H B , Moore E , Ramage D , et al . Communication-efficient learning of deep networks from decentralized data [PP ] . arXiv ( 2016-02-17 )[ 2025-01-16 ] . arXiv: 1602.05629 .

Dayan I , Roth H R , Zhong A X , et al . Federated learning for predicting clinical outcomes in patients with COVID-19 [J ] . Nature Medicine , 2021 , 27 ( 10 ): 1735 - 1743 .

Oualid A , Maleh Y , Moumoun L . Federated learning techniques applied to credit risk management: a systematic literature review [J ] . Edpacs , 2023 , 68 ( 1 ): 42 - 56 .

Kairouz P , McMahan H B , Avent B , et al . Advances and open problems in federated learning [C ] // Proceedings of the Foundations and Trends® in Machine Learning . New York : ACM Press , 2021 : 1 - 210 .

王勇 , 李国良 , 李开宇 . 联邦学习贡献评估综述 [J ] . 软件学报 , 2023 , 34 ( 3 ): 1168 - 1192 .

Wang Y , Li G L , Li K Y . Survey on contribution evaluation for federated learning [J ] . Journal of Software , 2023 , 34 ( 3 ): 1168 - 1192 .

Kumar I E , Venkatasubramanian S , Scheidegger C , et al . Problems with Shapley-value-based explanations as feature importance measures [PP ] . arXiv ( 2020-02-25 )[ 2025-01-16 ] . arXiv: 2002.11097 .

Yan T , Procaccia A D . If you like shapley then you’ll love the core [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2021 , 35 ( 6 ): 5751 - 5759 .

Liu Z L , Chen Y Y , Yu H , et al . GTG-shapley: efficient and accurate participant contribution evaluation in federated learning [J ] . ACM Transactions on Intelligent Systems and Technology , 2022 , 13 ( 4 ): 1 - 21 .

Yang C Y , Liu J , Sun H , et al . WTDP-shapley: efficient and effective incentive mechanism in federated learning for intelligent safety inspection [J ] . IEEE Transactions on Big Data , 2024 , 10 ( 6 ): 1028 - 1037 .

Deng X T , Papadimitriou C H . On the complexity of cooperative solution concepts [J ] . Mathematics of Operations Research , 1994 , 19 ( 2 ): 257 - 266 .

Conitzer V , Sandholm T . Complexity of constructing solutions in the core based on synergies among coalitions [J ] . Artificial Intelligence , 2006 , 170 ( 6/7 ): 607 - 619 .

Bachrach Y , Rosenschein J S . Coalitional skill games [J ] . Proceedings of the Third International Joint Conference on Autonomous Agents and Multiagent Systems - Volume 1 , 2008 : 1023 - 1030 .

Elkind E , Pasechnik D . Computing the nucleolus of weighted voting games [C ] // Proceedings of the Twentieth Annual ACM-SIAM Symposium on Discrete Algorithms . Society for Industrial and Applied Mathematics , 2009 : 327 - 335 .

Balcan M F , Procaccia A D , Zick Y . Learning cooperative games [C ] // Proceedings of the 24th International Conference on Artificial Intelligence . New York : ACM Press , 2015 : 475 - 481 .

Balkanski E , Syed U , Vassilvitskii S . Statistical cost sharing [PP ] . arXiv ( 2017-03-09 )[ 2025-01-17 ] . arXiv: 1703.03111 .

Yang M W , Jarin I , Buyukates B , et al . Maverick-aware shapley valuation for client selection in federated learning [PP ] . arXiv ( 2024-05-21 )[ 2025-01-16 ] . arXiv: 2405.12590 .

Chen K Y , Xu Z M . Federated learning for data market: shapley-UCB for seller selection and incentives [PP ] . arXiv ( 2024-10-10 )[ 2025-01-16 ] . arXiv: 2410.09107 .

Liu Y Q , Chang S , Liu Y , et al . FairFed: improving fairness and efficiency of contribution evaluation in federated learning via cooperative shapley value [C ] // Proceedings of the IEEE INFOCOM 2024 - IEEE Conference on Computer Communications . Piscataway : IEEE Press , 2024 : 621 - 630 .

Wei S Y , Tong Y X , Zhou Z M , et al . Efficient data valuation approximation in federated learning: a sampling-based approach [PP ] . arXiv ( 2025-04-23 )[ 2025-05-10 ] . arXiv: 2504.16668 .

GEIMER A , FIZ B , STATE R . On the volatility of shapley-based contribution metrics in federated learning [PP ] . arXiv ( 2024-05-13 )[ 2025-01-16 ] . arXiv: 2405.08044 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

SWTA-Shapley：一种高效的联邦学习贡献评估方法

异构数据分布下的跨域联邦入侵检测模型

无人机联邦学习场景下动态选择性同态加密隐私保护研究

基于模型分割的联邦学习数据隐私保护方法