FedMDO: Privacy-Preserving Federated Learning via Mixup Differential Objective.

Accuracy Degrading: Toward Participation-Fair Federated Learning.

Reschedule Gradients: Temporal Non-IID Resilient Federated Learning.