Fine-tuning Global Model via Data-Free Knowledge Distillation for Non-IID Federated Learning | IEEE Conference Publication | IEEE Xplore