Differentially Private Gradient-Tracking-Based Distributed Stochastic Optimization over Directed Graphs

TL;DR

This paper introduces a differentially private distributed stochastic optimization algorithm for directed graphs, achieving convergence guarantees and improved privacy-utility trade-offs, validated on MNIST and CIFAR-10 datasets.

Contribution

It proposes a novel gradient-tracking-based method with privacy-preserving noise addition and two schemes for step-sizes and sampling, ensuring finite privacy budget and convergence.

Findings

Achieves almost sure and mean square convergence for nonconvex objectives.

Scheme (S1) attains polynomial convergence rate under Polyak-Lojasiewicz condition.

Numerical experiments demonstrate superior performance over existing methods.

Abstract

This paper proposes a differentially private gradient-tracking-based distributed stochastic optimization algorithm over directed graphs. In particular, privacy noises are incorporated into each agent's state and tracking variable to mitigate information leakage, after which the perturbed states and tracking variables are transmitted to neighbors. We design two novel schemes for the step-sizes and the sampling number within the algorithm. The sampling parameter-controlled subsampling method employed by both schemes enhances the differential privacy level, and ensures a finite cumulative privacy budget even over infinite iterations. The algorithm achieves both almost sure and mean square convergence for nonconvex objectives. Furthermore, when nonconvex objectives satisfy the Polyak-Lojasiewicz condition, Scheme (S1) achieves a polynomial mean square convergence rate, and Scheme (S2) achieves an exponential mean square convergence rate. The trade-off between privacy and convergence is presented. The effectiveness of the algorithm and its superior performance compared to existing works are illustrated through numerical examples of distributed training on the benchmark datasets "MNIST" and "CIFAR-10".