Disentangling Client Contributions: Improving Federated Learning Accuracy in the Presence of Heterogeneous Data | IEEE Conference Publication | IEEE Xplore