The field of federated learning is moving towards addressing the challenges of data heterogeneity, concept drift, and fairness. Researchers are proposing innovative solutions such as federated incomplete multi-view clustering, knowledge distillation, and dynamic client clustering to improve the performance and robustness of federated learning models. Additionally, there is a growing focus on fairness and fairness evaluation in federated learning, with the development of libraries and benchmarks to support more robust and reproducible research. Notable papers in this area include Federated Incomplete Multi-view Clustering with Globally Fused Graph Guidance, which proposes a novel method for federated multi-view clustering, and FeDa4Fair, which introduces a library for generating tabular datasets to evaluate fair FL methods under heterogeneous client bias. Other noteworthy papers include FedDAA, which proposes a dynamic clustered FL framework to adapt to multi-source concept drift, and HybridQ, which introduces a hybrid classical-quantum generative adversarial network for skin disease image generation.
Advancements in Federated Learning and Fairness
Sources
Client Clustering Meets Knowledge Sharing: Enhancing Privacy and Robustness in Personalized Peer-to-Peer Learning
Tackling Data Heterogeneity in Federated Learning through Knowledge Distillation with Inequitable Aggregation