Publication Details

A PROBLEM KNOWLEDGE BASED BAYESIAN OPTIMIZATION ALGORITHM APPLIED IN MULTIPROCESSOR SCHEDULING

SCHWARZ Josef and JAROŠ Jiří. A PROBLEM KNOWLEDGE BASED BAYESIAN OPTIMIZATION ALGORITHM APPLIED IN MULTIPROCESSOR SCHEDULING. In: Mendel Conference on Soft Computing. Brno: Faculty of Mechanical Engineering BUT, 2004, pp. 83-88. ISBN 80-214-2676-4.
Czech title
Znalostně orientovaný Bayesovský optimalizační algoritmus
Type
conference paper
Language
english
Authors
Schwarz Josef, doc. Ing., CSc. (DCSY FIT BUT)
Jaroš Jiří, doc. Ing., Ph.D. (DCSY FIT BUT)
Keywords

optimization problems, multiprocessor scheduling problem, evolutionary algorithms, Bayesian optimization algorithm, problem knowledge.

Abstract

This paper deals with the multiprocessor scheduling problem, which  belongs to the class of frequently solved decomposition tasks. The goals is to experimentally compare the performance of the recently proposed Mixed Bayesian Optimization Algorithm (MBOA) based on probabilistic model with  the newly derived  knowledge  based MBOA version (KMBOA) This algorithm includes  utilization of prior knowledge about the structure of a task graph to speed-up the  convergence  and the  solution quality. The performance of standard  genetic algorithm was also tested on the same benchmarks.

Published
2004
Pages
83-88
Proceedings
Mendel Conference on Soft Computing
Conference
Tenth International Mendel Conference on Soft Computing, FME, VUT BRNO, CZ
ISBN
80-214-2676-4
Publisher
Faculty of Mechanical Engineering BUT
Place
Brno, CZ
BibTeX 
@INPROCEEDINGS{FITPUB7519,
   author = "Josef Schwarz and Ji\v{r}\'{i} Jaro\v{s}",
   title = "A PROBLEM KNOWLEDGE BASED BAYESIAN OPTIMIZATION ALGORITHM APPLIED IN MULTIPROCESSOR SCHEDULING",
   pages = "83--88",
   booktitle = "Mendel Conference on Soft Computing",
   year = 2004,
   location = "Brno, CZ",
   publisher = "Faculty of Mechanical Engineering BUT",
   ISBN = "80-214-2676-4",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/7519"
}
Back to top