On the numerical treatment of linearly constrained semi-infinite optimization problems

DOI10.1016/S0377-2217(99)00042-9MaRDI QIDQ1969893zbMATH OpenFDO

Authors E. Vercher, Teresa León, S. San Matías

Publication date 19 November 2002

Published in European Journal of Operational Research (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/S0377-2217(99)00042-9

zbMATH Keywords

descent rules feasible-direction implementation linear semi-infinite optimization pivoting rules steplength computation

Mathematics Subject Classification ID

Numerical mathematical programming methods (65K05) Semi-infinite programming (90C34)

Review text

This article is a valuable contribution in the practical treatment of linear semi-infinite optimization. Semi-infinite problems may imply a possibly infinite number of inequalities; they are motivated, e.g., by Chebyshev approximation, monotonic linear boundary value problems, experimental and engineering design. The treated problems are linear in the state \(x\in \mathbb{R}^n\) and continuously differentiable (of some higher order) in the inequality parameters \(s_i\) from \(q\) closed intervals. The \(q\) slack variables are denoted by \(z_i\) (subsequently we use vector notation). Simplex-type and feasible-direction methods are natural approaches. Among further preparations, the authors assume that the set of active constraints is finite for every feasible \((x,z)\). They recall algebraical characterizations for being an extreme point, a non-degenerate one, or an optimal non-degenerate one, respectively. Moreover, under Slater condition, optimality implies a Karush-Kuhn-Tucker type condition. Then, the authors give theoretical and practical algorithmic statements in the following line: pivoting rules for non-degenerate extreme points, purification procedure, descent rules for feasible solutions (by finite linear optimization, incorporated in a new characterization theorem on optimality), steplength computation, and implementation issues. For these items, a further rank assumption and a finiteness assumption on the number of minima for the lower stage problems on the slacks \(z_i(x, \cdot)\) are made. Hereon, Algorithms 1, 2 base. While Algorithm 1 combines a simplex-type strategy with a search direction scheme, Algorithm 2 is a feasible direction method (descent rules) using extreme points sometimes. Both algorithms were numerically tested and compared with ones from other literature/libraries. Algorithm 1 is competitive for problems with \(q=1\); Algorithm 2 betters its performance for minimax-type problems. The article is clearly written, with rich examples and computation experience.

Recommendations

Cites work

Cited in

(23)

This page was built for publication: On the numerical treatment of linearly constrained semi-infinite optimization problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1969893)