An Energy-Aware Combinatorial Contextual Neural Bandit Approachfor Joint Performance Optimization in Client Selection for Federated Learning

In the evolving landscape of machine learning (ML), federated learning (FL) stands out as an innovative strategy for training models across dispersed devices without centralizing raw data. Such an approach, however, grapples with data heterogeneity challenges, violating the independent and identically distributed (IID) assumption and undermining the global model accuracy. To address this, we present federated adversary-resilient neural selector (FANS), a sophisticated context-aware client selection algorithm, leveraging a combinatorial contextual neural bandit framework. This algorithm accentuates the enhanced extraction of contextual information by evaluating each local client with a universally standardized dataset, subsequently yielding a more insightful contextual representation tailored for federated settings. Additionally, we further address another crucial aspect of client selection — energy consumption. Considering this key factor along with the global accuracy jointly, greatly increase the adaptability of FL in real-world applications. We then introduce the Selection Robustness Score (SRS), a novel metric designed to quantify the efficacy of client selection under both adversarial and energy-constrained conditions. Using this metric, we demonstrate FANS’s effectiveness in enhancing the FL process. Empirical evaluations across diverse settings reveal our method’s superiority over current state-of-the-art solutions, with significant improvement in the SRS and energy conservation.