PANMIN: Sequential and parallel global optimization procedures with a variety of
options for the local search strategy
Description
PANMIN is a suite of programs for global optimization that take advantage of the
Merlin/MCL environment.
Specificaly PANMIN contains implementations of two algorithms that belong to the
stochastic class.
The first one is a modification of the "Controlled Random Search" (CRS) originally
introduced by Price (W. L. Price in Toward Global Optimization 2, L.C.W. Dixon and
G.P.Szego eds.), and aims to locate only one global minimum.
The second one is the "Healed Topographical Multilevel Single Linkage" (HTMLSL)
that is based on the Multi Level Single Linkage method of Kan and Timmer
(Math. Prog. 39, 1987, 27, Math. Prog. 39, 1987, 57, Ph.D. Thesis, Erasmus University Rotterdam),
with topographical modifications inspired from the articles by Ali and Storey
(J. Glob. Opt. 5, 1994, 349) and by Torn and Viitanen (J. Glob. Opt. 5, 1994, 267).
The above two codes are sequential and do not take advantage of environments with multiple
processors.
Note that CRS can hardly benefit from such a computational environment due to the
nature of its algorithmic structure.
HTMLSL lends itself to parallel processing and PANMIN contains a parallel implementation
of the algorithm that uses the Message Passing Interface (MPI) to take advantage of
either a shared memory architecture or a distributed environment with interconnected
computers.
PANMIN is described in the following article:
-
PANMIN: Sequential and parallel global optimization procedures with a variety of
options for the local search strategy
F.V. Theos, I.E. Lagaris, D.G. Papageorgiou
Comput. Phys. Commun. 159 (2004) 63-69.
Associated software in the CPC Program Library:
ADSU_v1_0
DOI: 10.1016/j.cpc.2003.11.001
Downloads
|