Federated learning improves communication by adapting to noisy channels
FedGenSC: Federated Generative Semantic Communication with Channel-Aware Adaptation
Machine Learning
Summary
Sending meaning through wireless networks can get messed up by noise, and usually, machines try to decode messages by recognizing patterns. The authors found that using AI that generates messages (called GANs) helps keep the meaning clear when the signal is distorted. They created a new method called FedGenSC that fixes problems seen with earlier approaches by making the main AI model shared globally while others work locally and by adjusting to how noisy the connection is. Their tests show this method works better when different devices have different data and connection qualities.
federated learninggenerative adversarial networkssemantic communicationnon-IID datadiscriminatorsignal-to-noise ratioRayleigh fading channelsemantic driftBLEU scoreDirichlet distribution
Authors
Rita Abou Fares, Razan Al Kakoun, Maher Nouiehed, Hadi Sarieddeen
Abstract
Integrating generative adversarial networks (GANs) into federated semantic communication (SemCom) is a natural progression, as generative priors can recover semantic fidelity under channel distortion that discriminative decoders cannot. However, naive GAN federation introduces three failure modes that prior work has, to the best of our knowledge, neither identified nor resolved: discriminator aggregation instability under non-independent and identically distributed (non-IID) data, semantic drift caused by divergent local embedding spaces, and channel-agnostic generation that cannot adapt to heterogeneous link conditions. We propose federated generative semantic communication (FedGenSC), which mitigates all three by employing a global generator with local-only discriminators, providing cross-client semantic information through a semantic prototype bank, and conditioning generation on the instantaneous signal-to-noise ratio (SNR). Experiments on the Europarl dataset over Rayleigh fading channels (K=10 clients, Dirichlet α=0.5) show that FedGenSC under non-IID data outperforms the FedDeepSC baseline across the tested SNR range, achieving up to a 58.2% relative improvement in bilingual evaluation understudy (BLEU)-1 at 18 dB. Ablation studies confirm the independent contribution of each component.