Decentralized Gradient-Free Methods for Stochastic Non-smooth Non-convex Optimization

We develop DGFM and DGFM+ for decentralized optimization of nonsmooth, nonconvex objectives using function evaluations. The methods combine randomized smoothing and gradient tracking, with variance reduction in DGFM+ improving the oracle complexity for reaching approximate Goldstein stationarity.

Published in Proceedings of the AAAI Conference on Artificial Intelligence 38 (16), 17477–17486 (2024).