![[2505.02515] FedSDAF: Leveraging Source Domain Awareness for Enhanced Federated Domain Generalization](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2505.02515] FedSDAF: Leveraging Source Domain Awareness for Enhanced Federated Domain Generalization
Summary
The paper presents FedSDAF, a novel framework that enhances Federated Domain Generalization by leveraging source domain awareness, demonstrating improved generalization capabilities through a dual-adapter architecture.
Why It Matters
As federated learning becomes increasingly important in machine learning, understanding how to effectively utilize source domain knowledge can lead to significant advancements in model performance across diverse applications. This research addresses a gap in existing methods by proposing a systematic approach that enhances generalization in isolated environments.
Key Takeaways
- FedSDAF utilizes a dual-adapter architecture to separate local expertise from global consensus.
- The framework incorporates a Bidirectional Knowledge Distillation mechanism for effective knowledge exchange.
- Experimental results show significant performance improvements over existing Federated Domain Generalization methods.
Computer Science > Machine Learning arXiv:2505.02515 (cs) [Submitted on 5 May 2025 (v1), last revised 22 Feb 2026 (this version, v4)] Title:FedSDAF: Leveraging Source Domain Awareness for Enhanced Federated Domain Generalization Authors:Hongze Li, Zesheng Zhou, Zhenbiao Cao, Xinhui Li, Wei Chen, Xiaojin Zhang View a PDF of the paper titled FedSDAF: Leveraging Source Domain Awareness for Enhanced Federated Domain Generalization, by Hongze Li and 5 other authors View PDF HTML (experimental) Abstract:Traditional Federated Domain Generalization (FedDG) methods focus on learning domain-invariant features or adapting to unseen target domains, often overlooking the unique knowledge embedded within the source domain, especially in strictly isolated federated learning environments. Through experimentation, we discovered a counterintuitive phenomenon: features learned from a complete source domain have superior generalization capabilities compared to those learned directly from the target domain. This insight leads us to propose the Federated Source Domain Awareness Framework (FedSDAF), the first systematic approach to enhance FedDG by leveraging source domain-aware features. FedSDAF employs a dual-adapter architecture that decouples "local expertise" from "global generalization consensus." A Domain-Aware Adapter, retained locally, extracts and protects the unique discriminative knowledge of each source domain, while a Domain-Invariant Adapter, shared across clients, builds a robust...
