Special Session 104: Recent Developments in High-Order Numerical Methods for Multiscale/Multiphysics Partial Differential Equations

Explicit Runge-Kutta methods for quadratic optimization via gradient flow equations

Tuo Liu
King Abdullah University of Science and Technology
Saudi Arabia
Co-Author(s):    Tuo Liu, David Ketcheson
Abstract:
This talk focuses on analyzing and developing accelerated optimization methods for the class of smooth and strongly convex functions. We derive a family of optimal gradient-based methods (named OERKD) for quadratic programming by building a mapping between continuous dynamics and discrete updates of optimization algorithms. Optimality of convergent rates is proved by analysis of explicit Runge-Kutta methods on the gradient flow equation with linear stability conditions. Experiments demonstrate the effectiveness of the proposed algorithm even on classical nonlinear problems. A noteworthy byproduct is proving the asymptotic equivalence between OERKD and Polyak`s heavy ball method, which subtly bridges two primary integration schemes.