FIARSE: Model-Heterogeneous Federated Learning via Importance-Aware Submodel Extraction

In federated learning (FL), accommodating clients' varied computational capacities poses a challenge, often limiting the participation of those with constrained