SplitGP: Achieving Both Generalization and Personalization in Federated Learning | IEEE Conference Publication | IEEE Xplore