ShapCCS: Shapley-Driven Client Coreset Selection in Federated Learning

Computation overhead has emerged as a critical bottleneck in Federated Learning (FL). Coreset selection tackles this challenge by constructing an informative subset to represent the full dataset. However, existing approaches optimize coreset construction solely at the data level and enforce a uniform retention ratio across all clients, ignoring client heterogeneity and introducing detrimental fragmented clients. In this paper, we first introduce a *gradient projection Shapley value* (GPSV) to evaluate client contributions. GPSV captures both the directional and magnitude information of client updates and enables exact Shapley value calculation with $\mathcal{O}(1)$ per-coalition evaluation. Building on GPSV, we then propose ShapCCS, the first client-level coreset selection strategy for FL. ShapCCS prioritizes clients with high GPSV scores while excluding fragmented clients with negligible or even negative GPSV. As a client-level coreset selection strategy, ShapCCS can be integrated with a data-level selection approach, and additionally reduces communication costs, an advantage unattainable by data-level methods alone. Extensive experiments demonstrate the superiority of ShapCCS on model performance and robustness to noise. The code is available at https://anonymous.4open.science/r/ShapCCS-5CBB.

This paper studies how to make federated learning faster and more efficient. We propose a new method called ShapCCS that identifies which participants provide the most useful information for training. Instead of treating all participants equally, our method prioritizes the most informative ones and removes participants that contribute little or may even hurt performance. To measure contribution, we introduce a new scoring method that evaluates how helpful each participant’s training updates are to the overall learning process. By selecting only the most valuable participants and data, ShapCCS reduces both computation and communication costs while maintaining model accuracy. Experiments on several benchmark datasets show that our approach consistently outperforms existing selection methods and remains robust even when some participants contain noisy data.