FedPCD: Priority-Constrained Federated Aggregation
FedPCD protects weak federated clients with minimal corrections while preserving global accuracy.
FedPCD is a priority-constrained federated optimization method that keeps the conventional federated aggregate as the primary objective and applies only the minimum correction needed to satisfy client performance constraints. The authors argue this avoids changing the global optimization direction, unlike client re-weighting or symmetric multi-objective fairness methods. Across heterogeneous federated-learning benchmarks, they report the highest global accuracy in every configuration and leading worst-10% client accuracy in nearly all, with faster empirical convergence.
- FedPCD keeps the global aggregate primary and applies only minimal client corrections.
- It reports the highest global accuracy across the tested heterogeneous FL settings.
- Worst-10% client accuracy leads in nearly all reported configurations.
- Empirical convergence needs fewer communication rounds than fairness-aware baselines.
From the original article
1 min read at arxiv.orgFederated learning (FL) seeks to optimize a shared global model from decentralized client data. While effective for average performance, this process can produce unfair performance across clients, particularly under statistically heterogeneous data distributions.
By Yousef Irshaid · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.