Refining asymptotic complexity bounds for nonconvex optimization methods, including why steepest descent is o(eps^{-2}) rather than O(eps^{-2})

Serge Gratton, Chee_Khian Sim, Philippe Toint

Research output: Working paper

82 Downloads (Pure)

Abstract

We revisit the standard ``telescoping sum'' argument ubiquitous in the final steps of analyzing evaluation complexity of algorithms for smooth nonconvex optimization, and obtain a refined formulation of the resulting bound as a function of the requested accuracy eps. While bounds obtained using the standard argument typically are of the form O(eps^{-\alpha}) for some positive alpha, the refined results are of the form o(eps^{-\alpha}). We then explore to which known algorithms our refined bounds are applicable and finally describe an example showing how close the standard and refined bounds can be.
Original languageEnglish
PublisherArxiv
Volume2408.09124
Publication statusPublished - 20 Aug 2024

Keywords

  • nonlinear optimisation
  • complexity analysis

Fingerprint

Dive into the research topics of 'Refining asymptotic complexity bounds for nonconvex optimization methods, including why steepest descent is o(eps^{-2}) rather than O(eps^{-2})'. Together they form a unique fingerprint.

Cite this