Regret Minimization in Stochastic Non-Convex Learning via a Proximal-Gradient Approach

Nadav Hallak · Panayotis Mertikopoulos · Volkan Cevher

Keywords: [ Non-convex Optimization ]

[ Abstract ]
[ Paper ]
[ Visit Poster at Spot A3 in Virtual World ]
Wed 21 Jul 9 a.m. PDT — 11 a.m. PDT
Spotlight presentation: Optimization (Nonconvex)
Wed 21 Jul 5 a.m. PDT — 6 a.m. PDT


This paper develops a methodology for regret minimization with stochastic first-order oracle feedback in online, constrained, non-smooth, non-convex problems. In this setting, the minimization of external regret is beyond reach for first-order methods, and there are no gradient-based algorithmic frameworks capable of providing a solution. On that account, we propose a conceptual approach that leverages non-convex optimality measures, leading to a suitable generalization of the learner's local regret. We focus on a local regret measure defined via a proximal-gradient mapping, that also encompasses the original notion proposed by Hazan et al. (2017). To achieve no local regret in this setting, we develop a proximal-gradient method based on stochastic first-order feedback, and a simpler method for when access to a perfect first-order oracle is possible. Both methods are order-optimal (in the min-max sense), and we also establish a bound on the number of proximal-gradient queries these methods require. As an important application of our results, we also obtain a link between online and offline non-convex stochastic optimization manifested as a new proximal-gradient scheme with complexity guarantees matching those obtained via variance reduction techniques.

Chat is not available.