Privacy-Preserving Federated Primal-Dual Learning for Non-Convex Problems With Non-Smooth Regularization | IEEE Conference Publication | IEEE Xplore