scientific article; zbMATH DE number 1421256
The article under review studies global optimization by both analysis and (bipartite) graph theory. The basic idea of this valuable contribution consists in following stationary points along trajectories which are reasonable for the problem. The authors are concerned with a minimization problem \(({\mathcal P})\) subject to finitely many equalities and inequalities. Throughout, the linear independence constraint qualification is assumed for the feasible set \(M\). First, the concept of critical, Karush-Kuhn-Tucker and, in particular, decomposition points of \(({\mathcal P})\) is given. By definition, for the latter points the Quadratic Index (QI) equals 1 (i.e., dimension of second-order descent \(=1\)). From Morse theory we learn: \#(local minima) -- \#(decomposition points) = 1, if \(M\) is connected. The authors apply Morse theory on the approximated feasible set \(M^\varepsilon\) defined by an additional inequality: sum of logarithms of the given inequalities \(\geq\varepsilon\) (a global smoothing via barrier functions, lateron used in a proof). There is a one-to-one correspondence between the critical points for the problems over \(M\) and \(M^\varepsilon\). Then, projected negative/positive gradients, Riemannian metrics \(R\) and, herewith, descent and ascent flows for \(({\mathcal P})\) are defined. Under a further overall smoothness and nondegeneracy assumption on \(\pm\) objective function, in case of equality constraints only, the authors prove: For generic Riemannian metrics both the 0-1-0 graph (nodes: local minima, critical points of QI = 1, edges: corresponding stable and unstable manifolds intersect) and the min-max digraph are connected. If inequalities are allowed, too, for connected \(M\) there exists a smooth \(R\) such that the resulting min-max digraph, defined via ascent/descent flows connecting some neighbours of our stationary points now, is strongly connected. The presentation is clear, well written, illustrated and examplified. This article may serve as theoretical ``discrete guideline for algorithmical treatments in nonlinear optimization. For the entire collection see [Zbl 0921.00022].
- Smooth Optimization Methods for Minimax Problems
- A smoothing scheme for optimization problems with max-min constraints
- scientific article; zbMATH DE number 3911152
- scientific article; zbMATH DE number 4164577
- Min-Max-Min Optimization with Smooth and Strongly Convex Objectives
- Minimization methods for smooth nonconvex functions
- Nonsmooth optimization: Theory and algorithms
- Nonsmooth global optimization
- scientific article; zbMATH DE number 3892946
- Smoothing approximations to nonsmooth optimization problems
- scientific article; zbMATH DE number 1985655 (Why is no real title available?)
- scientific article; zbMATH DE number 2068062 (Why is no real title available?)
- On absorbing cycles in min-max digraphs
- Smoothing by mollifiers. II: Nonlinear optimization
- Smoothing by mollifiers. I: Semi-infinite optimization
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4946288)
