Mixed method for solving the general convex programming problem

DOI10.1007/BF02667002MaRDI QIDQ1288666zbMATH OpenFDO

Authors B. N. Pshenichnyĭ, É. I. Nenakhov, V. N. Kuz'menko

Publication date 10 October 1999

Published in Cybernetics and Systems Analysis (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1007/BF02667002

zbMATH Keywords

convex programming cutting plane exact penalty linearization

Mathematics Subject Classification ID

Convex programming (90C25)

Review text

A method for solving the following convex optimization problem is suggested: \[ \text{minimize }f(x) \text{ subject to }x\in\{x\in \mathbb{R}^n\mid f_j(x)\leq 0,\;j=1,\dots, m,\;x\in M\}, \] where \(f,f_j\) are continuous convex functions and \(M\) is a convex polyhedron. The proposed method combines ideas of the following three methods known from the literature: The linearization method, the cutting plane method and the exact penalty method. The algorithm has two main features: (1) It converges under very general assumptions; (2) It provides an estimate for the Kuhn-Tucker multiplier. Since the theoretical version of the algorithm, the convergence of which the authors proved, may involve accumulating a large number of constraints, the authors suggested in the paper also a modified version of the algorithm, which overcomes this disadvantage. Although the convergence of this modified version of the algorithm was not proved, the authors report about its good efficiency and reliability during tests on numerical examples.

Recommendations

Cites work

Cited in

(10)

This page was built for publication: Mixed method for solving the general convex programming problem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1288666)