Stationarity in nonsmooth optimization between geometrical motivation and topological relevance
Abstract
The goal of this paper is to compare alternative stationarity notions in structured nonsmooth optimization (SNO). Here, nonsmoothness is caused by complementarity, vanishing, orthogonality type, switching, or disjunctive constraints. On one side, we consider geometrically motivated notions of -, -, and -stationarity in terms of Fr\'echet, Mordukhovich, and Clarke normal cones to the feasible set, respectively. On the other side, we advocate the notion of topologically relevant T-stationarity, which adequately captures the global structure of SNO. Our main findings say that (a) -stationary points include all local minimizers; (b) -stationary points, which are not -stationary, correspond to the singular saddle points of first order; (c) T-stationary points, which are not -stationary, correspond to the regular saddle points of first order; (d) -stationary points, which are not T-stationary, are irrelevant for optimization purposes. Overall, a hierarchy of stationarity notions for SNO is established.
Keywords
Cite
@article{arxiv.2409.04222,
title = {Stationarity in nonsmooth optimization between geometrical motivation and topological relevance},
author = {Vladimir Shikhman},
journal= {arXiv preprint arXiv:2409.04222},
year = {2024}
}
Comments
Keywords: nonsmooth optimization, stationarity, normal cones, lower level sets, Morse theory, saddle points