Tag
FedImp is a novel federated learning algorithm that uses impurity-based weighting to enhance convergence speed in non-IID data settings, showing significant reductions in communication rounds compared to baselines.