client-heterogeneity

The paper introduces FedeRage, a risk-averse federated learning method that uses conditional value-at-risk to address unknown client participation and heterogeneity, demonstrating enhanced accuracy, fairness, and convergence in experiments.