Stochastic alternating direction method of multipliers for Byzantine-robust distributed learning

This paper aims to solve a distributed learning problem under Byzantine attacks. In the underlying master-worker architecture, there exist number of unknown but malicious workers that can send arbitrary messages master deviate process, called workers. literature, total variation (TV) norm-penalized approximation formulation has been investigated alleviate effect To be specific, TV norm penalty not only forces local variables at regular close, is robust outliers sent by as well. For handling separable formulation, we propose Byzantine-robust stochastic alternating direction method multipliers (ADMM). Theoretically, prove proposed converges bounded neighborhood optimal solution rate O(1/k) mild assumptions, where k iterations and size determined Numerical experiments on MNIST COVERTYPE datasets further demonstrate effectiveness various