DASA: Delay-Adaptive Multi-Agent Stochastic Approximation

We consider a setting in which N agents aim to speedup a common Stochastic Approximation (SA) problem by acting in parallel and communicating with a central server. We assume that the up-link transmissions to the server are subject to asynchronous and potentially unbounded time-varying delays. To mitigate the effect of delays and stragglers while reaping the benefits of distributed computation, we propose DASA, a Delay-Adaptive algorithm for multi-agent Stochastic Approximation. We provide a finite-time analysis of DASA assuming that the agents' stochastic observation processes are independent Markov chains. Significantly advancing existing results, DASA is the first algorithm whose convergence rate depends only on the mixing time τm i x and on the average delay τa v g while jointly achieving an N-fold convergence speedup under Markovian sampling. Our work is relevant for various SA applications, including multi-agent and distributed temporal difference (TD) learning, Q-learning and stochastic optimization with correlated data.