The field of federated learning and graph neural networks is moving towards addressing the challenges of non-IID data, class imbalance, and personalized learning. Researchers are proposing novel frameworks and methods to enhance the performance of federated learning models, such as using generative models to generate synthetic samples for rare classes, employing stochastic feature manifold completion to enrich the training space, and introducing adaptive layer-wise feature alignment methods. Additionally, there is a growing interest in developing personalized federated learning approaches that can deliver tailored models to individual clients. Noteworthy papers in this area include: Fed MobiLLM, which proposes a server-assisted federated side-tuning paradigm to facilitate efficient federated LLM fine-tuning across mobile devices. GraphFedMIG, which tackles class imbalance in federated graph learning via mutual information-guided generation. APFL, which proposes an analytic personalized federated learning approach via dual-stream least squares to address the non-IID issue in PFL.
Federated Learning and Graph Neural Networks
Sources
GFlowNets for Learning Better Drug-Drug Interaction Representations
Fed MobiLLM: Efficient Federated LLM Fine-Tuning over Heterogeneous Mobile Devices via Server Assisted Side-Tuning
Multi-level Collaborative Distillation Meets Global Workspace Model: A Unified Framework for OCIL
FedMP: Tackling Medical Feature Heterogeneity in Federated Learning from a Manifold Perspective
Large-Small Model Collaborative Framework for Federated Continual Learning
Improving Learning of New Diseases through Knowledge-Enhanced Initialization for Federated Adapter Tuning
Flexible Personalized Split Federated Learning for On-Device Fine-Tuning of Foundation Models
GraphFedMIG: Tackling Class Imbalance in Federated Graph Learning via Mutual Information-Guided Generation
APFL: Analytic Personalized Federated Learning via Dual-Stream Least Squares