发表机构
State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications; National Engineering Research Center of Disaster Backup and Recovery, Beijing University of Posts and Telecommunications; JIUTIAN Research, China Mobile; School of Cyberspace Security, Beijing University of Posts and Telecommunications; University of Leicester（北京邮电大学网络与交换技术国家重点实验室; 北京邮电大学国家灾备工程技术研究中心; 中国移动九天研究院; 北京邮电大学网络空间安全学院; 莱斯特大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对联邦学习中跨客户端数据异质性导致的全局知识遗忘问题，提出FedADB框架，通过类别锚与双分支机制平衡全局一致性与本地优化，在多数据集上提升了准确率与收敛速度。
AI 中文摘要
异构设备收集的多模态数据用于协同训练，联邦学习（FL）是实现数据隐私保护的分布式建模关键范式。然而，在跨客户端数据异质性下，本地训练会遗忘先前学习的全局知识，导致性能和收敛速度显著下降。多数现有研究依赖全局对齐策略保留全局知识，这会阻碍本地优化，且对缺失类的监督不足；部分研究引入代理数据集补充缺失类的监督，但在无代理数据集时，平衡类别级全局一致性与本地优化目标仍是挑战。本研究提出FedADB，即类别锚驱动双分支联邦学习框架。具体而言，服务器生成在可微分输入空间中优化的类别锚，供所有客户端共享，这些类别锚作为全局参考，在本地训练时为缺失类提供监督；客户端设计双分支协同训练机制，其中基于锚的全局分支聚焦于全局一致性学习，通过类别锚平衡采样实现全局知识对齐；本地校准分支聚焦于学习判别性本地特征，缓解过度全局对齐导致的本地表征退化。在多个医疗和自然数据集上的大量实验表明，FedADB在准确率和收敛速度上均实现了显著提升。
英文摘要
Multimodal data collected by heterogeneous devices are used for collaborative training, where federated learning (FL) serves as a key paradigm for effective distributed modeling with data privacy preservation. However, local training suffers from the forgetting of previously learned global knowledge under cross-client data heterogeneity, which leads to significant declines in both performance and convergence speed. Most previous studies rely on global alignment strategies to retain global knowledge, which hinder local optimization and lead to inadequate supervision of missing classes. Some studies introduce proxy datasets to supplement supervision for missing classes. However, it remains a challenge to balance class-wise global consistency and local optimization objectives without proxy datasets. In this work, we propose FedADB, a Class Anchor-Driven Dual-Branch FL framework. Specifically, the server generates class anchors optimized in a differentiable input space, which are shared across clients. These class anchors serve as global references that provide supervision for missing classes during local training. A dual-branch collaborative training mechanism is designed for clients. In this mechanism, the anchor-based global branch focuses on learning with global consistency, achieving global knowledge alignment by class-anchor balanced sampling. The local calibration branch focuses on learning discriminative local features, mitigating the degradation of local representations caused by excessive global alignment. Extensive experiments across multiple medical and natural datasets demonstrate that FedADB achieves significant improvements in both accuracy and convergence speed.
CommentsAccepted to appear in Proceedings of the 34th ACM International Conference on Multimedia (MM '26)
