Cross-Domain Federated Semantic Communication with Global Representation Alignment and Domain-Aware Aggregation

2025-11-29

AI Summary

To address the significant performance degradation caused by heterogeneous data distributions across clients in cross-domain federated semantic communication, this paper pioneers the integration of domain shift modeling into the federated training framework for semantic communication systems. We propose a dual mechanism: (1) global representation alignment, which employs contrastive learning to project local features into a shared semantic space, ensuring cross-domain semantic consistency; and (2) domain-aware aggregation, which dynamically weights domain-specific gradients during model aggregation to mitigate bias induced by dominant domains with large sample sizes. Our approach synergistically integrates deep joint source-channel coding, federated learning, representation alignment, and domain-aware aggregation. Experiments on a three-domain setting at 1 dB SNR demonstrate that our method improves PSNR by 0.5 dB over MOON; moreover, this gain widens as channel quality improves.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Agent Communication
• Application Domains: Mobility, Driving & Flight

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Semantic communication can significantly improve bandwidth utilization in wireless systems by exploiting the meaning behind raw data. However, the advancements achieved through semantic communication are closely dependent on the development of deep learning (DL) models for joint source-channel coding (JSCC) encoder/decoder techniques, which require a large amount of data for training. To address this data-intensive nature of DL models, federated learning (FL) has been proposed to train a model in a distributed manner, where the server broadcasts the DL model to clients in the network for training with their local data. However, the conventional FL approaches suffer from catastrophic degradation when client data are from different domains. In contrast, in this paper, a novel FL framework is proposed to address this domain shift by constructing the global representation, which aligns with the local features of the clients to preserve the semantics of different data domains. In addition, the dominance problem of client domains with a large number of samples is identified and, then, addressed with a domain-aware aggregation approach. This work is the first to consider the domain shift in training the semantic communication system for the image reconstruction task. Finally, simulation results demonstrate that the proposed approach outperforms the model-contrastive FL (MOON) framework by 0.5 for PSNR values under three domains at an SNR of 1 dB, and this gap continues to widen as the channel quality improves.

Problem

• Addresses domain shift in federated semantic communication systems
• Aligns global representation with local features across domains
• Mitigates dominance of clients with large sample sizes

Innovation

• Global representation alignment for domain shift
• Domain-aware aggregation to prevent sample dominance
• Cross-domain federated semantic communication framework