计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (3): 197-206.doi: 10.11896/jsjkx.250100068

• 数据库 & 大数据 & 数据科学 • 上一篇    下一篇

基于多专家协同和信息交互的社会化学习

李林昊1,2,3, 许亚楠1, 董永峰1,2,3, 王振1,2,3   

  1. 1 河北工业大学人工智能与数据科学学院 天津 300401
    2 河北省大数据计算重点实验室(河北工业大学) 天津 300401
    3 河北省数据驱动工业智能工程研究中心(河北工业大学) 天津 300401
  • 收稿日期:2025-01-10 修回日期:2025-05-06 发布日期:2026-03-12
  • 通讯作者: 董永峰(dongyf@hebut.edu.cn)
  • 作者简介:(lilinhao@hebut.edu.cn)
  • 基金资助:
    国家自然科学基金(62306103)

Social Learning Based on Multi-expert Collaboration and Information Interaction

LI Linhao1,2,3, XU Yanan1, DONG Yongfeng1,2,3, WANG Zhen1,2,3   

  1. 1 School of Artificial Intelligence, Hebei University of Technology, Tianjin 300401, China
    2 Hebei Province Key Laboratory of Big Data Computing(Hebei University of Technology), Tianjin 300401, China
    3 Hebei Engineering Research Center of Data-driven Industrial Intelligent(Hebei University of Technology), Tianjin 300401, China
  • Received:2025-01-10 Revised:2025-05-06 Online:2026-03-12
  • About author:LI Linhao,born in 1989,Ph.D,professor,is a member of CCF(No.42701M).His main research interests include machine learning,knowledge inference and computer vision.
    DONG Yongfeng,born in 1977,Ph.D,professor,is a member of CCF(No.28279D).His main research interests include machine learning,knowledge engineering,computer vision and intelligent information processing.
  • Supported by:
    National Natural Science Foundation of China(62306103).

PDF (PC)

104

摘要： 在分布式环境中,数据异质性表现为数据特征差异。专家模型协同存在知识孤立与任务分配不合理的问题,导致专家训练效果参差不齐,难以充分发挥各模型优势,使得整体性能受限。针对这些问题,提出了一种基于多专家协同和信息交互的社会化学习框架(Social Learning Based on Multi-expert Collaboration and Information Interaction,MECII)。该框架结合混合专家模型和社会化学习思想,通过多专家协同、门控网络、自适应信息交互和门控选择约束这四大模块,优化了专家间的知识共享与互补机制,有效解决了分布式学习中的数据异质性和专家协同问题。MECII通过精准的专家选择与任务分配,促进了专家之间的信息流动,使每个专家在处理特定数据时的准确率得到提升,增强了整体模型性能。实验结果表明,MECII在CIFAR-10和CIFAR-100数据集上相比传统的联邦学习基准方法有显著的性能提升,特别是在数据异质性场景下,与先进的FedL2P方法相比,MECII将分类准确率分别提高了6.69个百分点和5.13个百分点,且有效优化了每个专家的准确率。实验结果验证了MECII在促进专家协作和提升个体精度方面具有显著优势。

关键词: 多专家协同, 社会化学习, 数据异质性, 联邦学习, 信息交互, 混合专家模型

Abstract: In distributed environments,data heterogeneity manifests as discrepancies in data features.Expert model collaboration suffers from knowledge isolation and improper task allocation,leading to uneven training results among experts,preventing full exploitation of each model’s advantages,and thus limiting overall performance.To address these challenges,this paper proposes MECII.The framework integrates the mixture of experts(MoE) architecture with social learning(SL) principles,optimizing the knowledge sharing and complementarity among experts through four key components:multi-expert collaboration,gating network,adaptive information interaction,and gating selection constraints.This approach effectively resolves issues related to data heterogeneity and expert collaboration in distributed learning.By ensuring precise expert selection and task allocation,MECII facilitates information flow between experts,enhancing each expert’s accuracy when processing specific data,and consequently enhancing the overall model performance.Experimental results demonstrate that MECII significantly outperforms traditional federated learning(FL) baseline methods on the CIFAR-10 and CIFAR-100 datasets.Particularly in scenarios with data heterogeneity,MECII achieves improvements in classification accuracy of 6.69 percentage points and 5.13 percentage points,respectively,compared to the state-of-the-art FedL2P method.Moreover,individual expert accuracy is effectively optimized,validating the framework’s significant advantages in promoting expert collaboration and improving individual accuracy.

Key words: Multi-expert collaboration, Social learning, Data heterogeneity, Federated learning, Information interaction, Mixture of experts

中图分类号: 

  • TP391
[1]AKERS R L,JENNINGS W G.[M]//The Handbook of Crimi-nological Theory.2015:230-240.
[2]XIE G Q,ZHONG B W,LI Y.Distributed Adaptive Multi-agentRendezvous Control Based on Average Consensus Protocol[J].Computer Science,2024,51(5):242-249.
[3]JACOBS R A,JORDAN M I,NOWLAN S J,et al.Adaptive mixtures of local experts[J].Neural Computation,1991,3(1):79-87.
[4]KONEČKOVINÝ J,MCMAHAN H B,YU F X,et al.Federated Learning:Strategies for Improving Communication Efficiency[J].arXiv:1610.05492,2016.
[5]ZHU G H,QI J H,ZHU Z N,et al.Research on progressive deep ensemble architecture search algorithm[J].Chinese Journal of Computers,2023,46(10):2041-2065.
[6]LIU Q,SHI M L,HUANG Z G,et al.Multi-agent collaborative reinforcement learning method based on dual-perspective mode-ling[J].Chinese Journal of Computers,2024,47(7):1582-1594.
[7]BANDURA A,WALTERS R H.Social learning theory[M].Englewood Cliffs,NJ:Prentice hall,1977.
[8]BADGHISH S,SHAIK A S,SAHORE N,et al.Can transac-tional use of AI-controlled voice assistants for service delivery pickup pace in the near future? A social learning theory(SLT) perspective[J].Technological Forecasting and Social Change,2024,198:122972.
[9]PARKER L E.Distributed Intelligence:Overview of the Field and its Application in Multi-Robot Systems[C]//AAAI Fall Symposium:Regarding the Intelligence in Distributed Intelligent Systems.2007:1-6.
[10]WANG X W,FENG X,YU H Q.Multi-agent Cooperative Algo-rithm for Obstacle Clearance Based on Deep Deterministic Policy Gradient and Attention Critic[J].Computer Science,2024,51(7):319-326.
[11]WANG X,ZHAO C,HUANG T W,et al.Cooperative learning of multi-agent systems via reinforcement learning[J].IEEE Transactions on Signal and Information Processing over Networks,2023,9:13-23.
[12]AHMED I,SYED M A,MAARUF M,et al.Distributed computing in multi-agent systems:a survey of decentralized machine learning approaches[J].Computing,2025,107(1):2.
[13]HONG S,ZHENG X,CHEN J,et al.Metagpt:Meta programming for a multi-agent collaborative framework[J].arXiv:2308.00352,2023.
[14]ZHANG M Y,JIN Z,LIU K.Virtual regret advantage self-playmethod for cooperative-competitive hybrid multi-agent systems[J].Journal of Software,2024,35(2):739-757.
[15]MCMAHAN B,MOORE E,RAMAGE D,et al.Communica-tion-efficient learning of deep networks from decentralized data[C]//Artificial Intelligence and Statistics.PMLR,2017:1273-1282.
[16]KAIROUZ P,MCMAHAN H B,AVENT B,et al.Advances and open problems in federated learning[J].Foundations and Trends© in Machine Learning,2021,14(1/2):1-210.
[17]SATTLER F,WIEDEMANN S,MÜLLER K R,et al.Robust and communication-efficient federated learning from non-iid data[J].IEEE Transactions on Neural Networks and Learning Systems,2019,31(9):3400-3413.
[18]DINH C T,TRAN N,NGUYEN J.Personalized federatedlearning with moreau envelopes[J].Advances in Neural Information Processing Systems,2020,33:21394-21405.
[19]WANG J,LIU Q,LIANG H,et al.Tackling the objective inconsistency problem in heterogeneous federated optimization[J].Advances in Neural Information Processing Systems,2020,33:7611-7623.
[20]LI T,SAHU A K,ZAHEER M,et al.Federated optimization in heterogeneous networks[J].Proceedings of Machine Learning and Systems,2020,2:429-450.
[21]LI Q,HE B,SONG D.Model-contrastive federated learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2021:10713-10722.
[22]QU Z,LI X,DUAN R,et al.Generalized federated learning via sharpness aware minimization[C]//International Conference on Machine Learning.PMLR,2022:18250-18280.
[23]SHI Y,LIANG J,ZHANG W,et al.Towards understanding and mitigating dimensional collapse in heterogeneous federated learning[J].arXiv:2210.00226,2022.
[24]LEE R,KIM M,LI D,et al.Fedl2p:Federated learning to personalize[C]//Advances in Neural Information Processing Systems.2024.
[25]HUANG W,YE M,SHI Z,et al.Federated learning for genera-lization,robustness,fairness:A survey and benchmark[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2024,46(12):9387-9406.
[26]ZHANG R L,DU J H,YIN H.Client selection algorithm incross-device federated learning[J].Journal of Software,2024,35(12):5725-5740.
[27]WANG Y,FU H,KANAGAVELU R,et al.An aggregation-free federated learning for tackling data heterogeneity[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2024:26233-26242.
[28]SHAZZER N,MIRHOSEINI A,MAZIARZ K,et al.Outra-geously large neural networks:The sparsely-gated mixture-of-experts layer[J].arXiv:1701.06538,2017.
[29]YANG Z,DAI Z,SALAKHUTDINOV R,et al.Breaking the softmax bottleneck:A high-rank RNN language model[J].ar-Xiv:1711.03953,2017.
[30]FEDUS W,ZOPH B,SHAZEER N.Switch transformers:Sca-ling to trillion parameter models with simple and efficient sparsity[J].Journal of Machine Learning Research,2022,23(120):1-39.
[31]LI Y,JIANG S,HU B,et al.Uni-moe:Scaling unified multimodal llms with mixture of experts[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2025,47(5):3424-3439.
[32]JORDAN M I,JACOBS R A.Hierarchical mixtures of experts and the EM algorithm[J].Neural Computation,1994,6(2):181-214.
[33]SHAZEER N,STERN M.Adafactor:Adaptive learning rateswith sublinear memory cost[C]//International Conference on Machine Learning.PMLR,2018:4596-4604.
[34]YU J,ZHUGE Y,ZHANG L,et al.Boosting continual learning of vision-language models via mixture-of-experts adapters[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2024:23219-23230.
[35]ROSENBAUM C,KLINGER T,RIEMER M.Routing net-works:Adaptive selection of non-linear functions for multi-task learning[J].arXiv:1711.01239,2017.
[36]DONG L,YANG N,WANG W,et al.Unified language model pre-training for natural language understanding and generation[C]//Advances in Neural Information Processing Systems.2019.
[37]DING X,ZHANG X,HAN J,et al.Scaling up your kernels to 31x31:Revisiting large kernel design in cnns[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:11963-11975.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发