CosAge: Federated Learning with Gradient Summaries for Centralized Client Selection

Abstract

Heterogeneous data and partial participation hinder the effectiveness of federated learning (FL). To compare client selection policies under a common yardstick, we adopt the Federated Learning with Gradient Summaries for Centralized Client Selection (FL-GSCCS) model, where each client transmits a lightweight gradient summary for selection and only chosen clients perform full local training with sparsified updates. Within this framework, we propose C OS A GE , a hybrid centralized policy that combines Age of Information (AoI) with gradient dissimilarity computed from a proxy update via the cos 4 metric. Simulation results show that C OS A GE consistently outperforms AoI-only and dissimilarity-only baselines in non-IID settings, and approaches the performance of clustering-based upper bounds without requiring client-to-client coordination or server access to client statistics.