Agent-Level Differentially Private Federated Learning via Compressed Model Perturbation | IEEE Conference Publication | IEEE Xplore