ASAP.SGD: Instance-based Adaptiveness to Staleness in Asynchronous SGD

Concurrent algorithmic implementations of Stochastic Gradient Descent (SGD) give rise to critical questions for compute-intensive Machine Learning (ML). Asynchrony implies speedup in some contexts, and challenges in others, as stale updates may lead to slower, or non-converging executions. While previous works showed asynchrony-adaptiveness can improve stability and speedup by reducing the step size for stale updates according to static rules, there is no one-size-fits-all adaptation rule, since the optimal strategy depends on several factors. We introduce (i) ASAP.SGD, an analytical framework capturing necessary and desired properties of staleness-adaptive step size functions and (ii) TAIL-T, a method for utilizing key properties of the execution instance, generating a tailored strategy that not only dampens the impact of stale updates, but also leverages fresh ones. We recover convergence bounds for adaptiveness functions satisfying the ASAP.SGD conditions, for general, convex and non-convex problems, and establish novel bounds for ones satisfying the Polyak-Lojasiewicz property. We evaluate TAIL-T with representative AsyncSGD concurrent algorithms, for Deep Learning problems, showing TAIL-T is a vital complement to AsyncSGD, with (i) persistent speedup in wall-clock convergence time in the parallelism spectrum, (ii) considerably lower risk of non-convergence, as well as (iii) precision levels for which original SGD implementations fail.