Online Multi-Agent Decentralized Byzantine-robust Gradient Estimation

TL;DR

This paper introduces a decentralized method for robust gradient estimation in multi-agent systems that is resilient to Byzantine failures, utilizing perturbation, secure estimation, and stochastic approximation techniques.

Contribution

It presents a novel iterative algorithm combining perturbation, secure state estimation, and stochastic approximation for Byzantine-robust gradient estimation in distributed settings.

Findings

Algorithm demonstrates resilience to Byzantine failures in simulations

Numerical experiments validate the effectiveness of the proposed method

The approach outperforms existing decentralized gradient estimation techniques

Abstract

In this paper, we propose an iterative scheme for distributed Byzantineresilient estimation of a gradient associated with a black-box model. Our algorithm is based on simultaneous perturbation, secure state estimation and two-timescale stochastic approximations. We also show the performance of our algorithm through numerical experiments.