Title: Differentially Private Distributed Nonconvex Stochastic Optimization with Quantized Communications

Abstract: This paper proposes a new distributed nonconvex stochastic optimization algorithm that can achieve privacy protection, communication efficiency and convergence simultaneously. Specifically, each node adds time-varying privacy noises to its local state to avoid information leakage, and then quantizes its noise-perturbed state before transmitting to improve communication efficiency. By employing the subsampling method controlled through the sample-size parameter, the proposed algorithm reduces the impact of privacy noises, and enhances the differential privacy level. When the global cost function satisfies the Polyak-Lojasiewicz condition, the mean and high-probability convergence rate and the oracle complexity of the proposed algorithm are given. Importantly, the proposed algorithm achieves both the mean convergence and a finite cumulative differential privacy budget over infinite iterations as the sample-size goes to infinity. A numerical example of the distributed training on the "MNIST" dataset is given to show the effectiveness of the algorithm.