Federated Joint Learning for Domain and Class Generalization

2026-01-18

AI Summary

Existing federated learning approaches typically address unseen classes or unseen domains in isolation, lacking a unified framework for joint generalization across both dimensions. This work proposes FedDCG, the first method to achieve simultaneous class and domain generalization in federated learning. FedDCG employs a domain grouping strategy to train a learnable class generalization network, incorporates a knowledge disentanglement mechanism to separate domain-invariant and domain-specific features, and fuses predictions at inference time based on domain similarity. Extensive experiments demonstrate that FedDCG significantly outperforms current state-of-the-art methods across multiple benchmark datasets, achieving superior performance in both accuracy and robustness.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Learning & Optimization for CV
• Application Domains: Misinformation & Fake News

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

Efficient fine-tuning of visual-language models like CLIP has become crucial due to their large-scale parameter size and extensive pretraining requirements. Existing methods typically address either the issue of unseen classes or unseen domains in isolation, without considering a joint framework for both. In this paper, we propose \textbf{Fed}erated Joint Learning for \textbf{D}omain and \textbf{C}lass \textbf{G}eneralization, termed \textbf{FedDCG}, a novel approach that addresses both class and domain generalization in federated learning settings. Our method introduces a domain grouping strategy where class-generalized networks are trained within each group to prevent decision boundary confusion. During inference, we aggregate class-generalized results based on domain similarity, effectively integrating knowledge from both class and domain generalization. Specifically, a learnable network is employed to enhance class generalization capabilities, and a decoupling mechanism separates general and domain-specific knowledge, improving generalization to unseen domains. Extensive experiments across various datasets show that \textbf{FedDCG} outperforms state-of-the-art baselines in terms of accuracy and robustness.

Problem

• federated learning
• domain generalization
• class generalization
• visual-language models
• unseen domains

Innovation

• Federated Learning
• Domain Generalization
• Class Generalization
• Knowledge Decoupling
• Visual-Language Models