Convergence detection for optimization algorithms: Approximate-KKT stopping criterion when Lagrange multipliers are not available

Autor(es)

Haeser, Gabriel ^[1] ; de Melo, Vinicius V. ^[2]

Afiliação do(s) autor(es)

^[1] Univ Sao Paulo, Inst Math & Stat, Sao Paulo, SP - Brazil

^[2] Univ Fed Sao Paulo, Inst Sci & Technol, Sao Jose Dos Campos, SP - Brazil

• Tipo de documento: Artigo Científico
• Fonte: OPERATIONS RESEARCH LETTERS; v. 43, n. 5, p. 484-488, SEP 2015.
• Citações Web of Science: 3

Resumo

In this paper we investigate how to efficiently apply Approximate-Karush-Kuhn-Tucker proximity measures as stopping criteria for optimization algorithms that do not generate approximations to Lagrange multipliers. We prove that the KKT error measurement tends to zero when approaching a solution and we develop a simple model to compute the KKT error measure requiting only the solution of a non-negative linear least squares problem. Our numerical experiments on a Genetic Algorithm show the efficiency of the strategy. (C) 2015 Elsevier B.V. All rights reserved. (AU)

• Processo FAPESP: 10/19720-5 - Condições de otimalidade e restauração inexata
• Beneficiário: Gabriel Haeser
• Modalidade de apoio: Auxílio à Pesquisa - Jovens Pesquisadores