Federated semantic communication allows user equipment (UEs) to participate in training a shared model by contributing either model updates or semantic features. This hybrid approach is flexible but introduces challenges around the varying costs of different roles and the interaction between model updates and semantic features.
The paper, "Adaptive Role Assignment and Controlled Feature Reconstruction for Federated Image Semantic Communication," addresses these issues. It proposes an adaptive role assignment mechanism to manage heterogeneous role costs, and a controlled feature reconstruction method to handle the interaction between model and feature components.
The work is presented on arXiv and targets image semantic communication, though the abstract does not provide experimental results. The contribution is a framework-level solution to the role-assignment and feature-reconstruction problems in hybrid federated settings.