PARALLEL METHODS FOR GLOBAL OPTIMIZATION PROBLEM SOLVING
N. I. Lobachevsky State University, Software Department; President
V. P. Gergel
N. I. Lobachevsky State University, Software Department; Dean
K. A. Barkalov
N. I. Lobachevsky State University, Software Department; Senior Teacher
Abstract. Parallel algorithm for solving multiextremal optimization problems with nonconvex constraints is considered. It is based on the reduction of multidimensional problems to the set of joint one-dimensional ones. New scheme of construction of the set of Peano-type space-filling curves is proposed. Suggested scheme preserves a part of information about closeness of points in multidimensional space. The results of numerical experiments showing acceleration of algorithm convergence with use of the new scheme of multiply curves construction are presented.
Keywords: multiextremal optimization, nonconvex constraints, Peano curves, parallel algorithms.