RUS  ENG
Full version
JOURNALS // Diskretnyi Analiz i Issledovanie Operatsii // Archive

Diskretn. Anal. Issled. Oper., 2015 Volume 22, Issue 1, Pages 64–85 (Mi da807)

Numerical search for global solutions in problems of non-symmetric bilinear separability

A. V. Orlov

Institute for System Dynamics and Control Theory SB RAS, 134 Lermontov St., 664033 Irkutsk, Russia

Abstract: The paper is devoted to the bilinear separability problem of two sets (the non-symmetrical case). The optimization approach to the problem is applied. This approach is based on the reduction of the bilinear separability problem to an equivalent nonconvex bilinear optimization problem with disjoint constraints. The latter problem is solved by Global Search Theory developed by A. S. Strekalovsky. According to that theory, the local and global search methods for the problem under scrutiny were elaborated. Computational testing of the developed methods has shown the competitive efficiency of the approach on a rather large number of test problems of bilinear separability. Ill. 5, tab. 3, bibliogr. 29.

Keywords: classification problem, bilinear separability, optimization approach, local search, global search, test problem generation, numerical experiment.

UDC: 519.853.4

Received: 25.03.2014
Revised: 26.08.2014

DOI: 10.17377/daio.2015.22.450



Bibliographic databases:


© Steklov Math. Inst. of RAS, 2024