A Two-Timescale Framework for Bilevel Optimization: Complexity Analysis and Application to Actor-Critic

DOI10.1137/20M1387341zbMATH Open1516.90062arXiv2007.05170MaRDI QIDQ6344797FDOQ6344797

Authors: Mingyi Hong, Hoi-To Wai, Zhaoran Wang, Zhuoran Yang 

Publication date: 10 July 2020

Abstract: This paper analyzes a two-timescale stochastic algorithm framework for bilevel optimization. Bilevel optimization is a class of problems which exhibit a two-level structure, and its goal is to minimize an outer objective function with variables which are constrained to be the optimal solution to an (inner) optimization problem. We consider the case when the inner problem is unconstrained and strongly convex, while the outer problem is constrained and has a smooth objective function. We propose a two-timescale stochastic approximation (TTSA) algorithm for tackling such a bilevel problem. In the algorithm, a stochastic gradient update with a larger step size is used for the inner problem, while a projected stochastic gradient update with a smaller step size is used for the outer problem. We analyze the convergence rates for the TTSA algorithm under various settings: when the outer problem is strongly convex (resp.~weakly convex), the TTSA algorithm finds an $mathcalO(K^{-2/3})$-optimal (resp.~$mathcalO(K^{-2/5})$-stationary) solution, where $K$ is the total iteration number. As an application, we show that a two-timescale natural actor-critic proximal policy optimization algorithm can be viewed as a special case of our TTSA framework. Importantly, the natural actor-critic algorithm is shown to converge at a rate of $mathcalO(K^{-1/4})$ in terms of the gap in expected discounted reward compared to a global optimal policy.

Mathematics Subject Classification ID

Nonconvex programming, global optimization (90C26) Nonlinear programming (90C30) Stochastic programming (90C15)

This page was built for publication: A Two-Timescale Framework for Bilevel Optimization: Complexity Analysis and Application to Actor-Critic

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6344797)