English

Contractivity of Runge-Kutta methods for convex gradient systems

Numerical Analysis 2021-04-02 v3 Numerical Analysis Optimization and Control

Abstract

We consider the application of Runge-Kutta (RK) methods to gradient systems (d/dt)x=V(x)(d/dt)x = -\nabla V(x), where, as in many optimization problems, VV is convex and V\nabla V (globally) Lipschitz-continuous with Lipschitz constant LL. Solutions of this system behave contractively, i.e. the Euclidean distance between two solutions x(t)x(t) and x~(t)\widetilde{x}(t) is a nonincreasing function of tt. It is then of interest to investigate whether a similar contraction takes place, at least for suitably small step sizes hh, for the discrete solution. Dahlquist and Jeltsch results' imply that (1) there are explicit RK schemes that behave contractively whenever LhLh is below a scheme-dependent constant and (2) Euler's rule is optimal in this regard. We prove however, by explicit construction of a convex potential using ideas from robust control theory, that there exists RK schemes that fail to behave contractively for any choice of the time-step hh.

Keywords

Cite

@article{arxiv.1909.09971,
  title  = {Contractivity of Runge-Kutta methods for convex gradient systems},
  author = {J. M. Sanz-Serna and Konstantinos C. Zygalakis},
  journal= {arXiv preprint arXiv:1909.09971},
  year   = {2021}
}

Comments

13 pages, 2 figures

R2 v1 2026-06-23T11:22:27.983Z