Cross-Training with Prototypical Distillation for improving the generalization of Federated Learning | IEEE Conference Publication | IEEE Xplore