Federated Learning with Energy-Based Structured Probabilistic Inference
2026-06-29 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors address a problem in federated learning where combining updates from different clients using fixed rules can be inefficient because clients have different kinds of data and levels of contribution. They introduce a method that uses Conditional Random Fields (CRFs) to better decide how much weight each client's update should have, taking into account both individual reliability and relationships between clients. Their approach helps the global learning model improve more effectively, especially when client data is very different. Tests show their method works better than common existing techniques in these challenging scenarios.
federated learningclient aggregationConditional Random Fieldsnon-IID datamachine learning convergenceunary potentialspairwise potentialsglobal modelheterogeneous data
Authors
Dario Fenoglio, Daniil Kirilenko, Martin Gjoreski, Marc Langheinrich
Abstract
Federated learning typically aggregates client updates using fixed or heuristic weighting rules, which can be suboptimal when clients have heterogeneous data and varying contributions to the global model. We propose a framework that refines client aggregation weights using Conditional Random Fields (CRFs). Our method defines unary potentials for individual clients and pairwise potentials for all client pairs, allowing the server to model both client-specific reliability and interactions between clients. The resulting CRF inference produces aggregation weights that enable better convergence of the global training objective. Experiments show that, under non-IID heterogeneity, our approach consistently improves performance over well-established federated learning baselines.