Publication Details
ACCELERATED BAYESIAN OPTIMIZATION ALGORITHMS FOR ADVANCED HYPERGRAPH PARTITIONING, accepted paper
Očenášek Jiří, Ing.
Optimization problems, decomposition and allocation problems, graphicalprobabilistic model, Bayesian network, Bayesian-Dirichlet metric,Bayesian optimization algorithm, problem knowledge, parallelization,hypergraph partitioning.
The paper summarizes our recent work on the design, analysis andapplications of the Bayesian optimization algorithm (BOA) and itsadvanced accelerated variants for solving complex - sometimesNP-complete - combinatorial optimization problems from circuit design.We review the methods for accelerating BOA for hypergraph-partitioningproblem. The first method accelerates the convergence of sequential BOAby utilizing specific knowledge about the optimized problem and thesecond method is based on the parallel construction of a probabilisticmodel. In the experimental part we analyze the advantages ofacceleration techniques and prove that BOA is able to solve hypergraphpartitioning problems reliably, effectively, and without the need forspecifying control parameters and encoding schemes as inrecombination-based genetic algorithms.
@inproceedings{BUT13984,
author="Josef {Schwarz} and Jiří {Očenášek}",
title="ACCELERATED BAYESIAN OPTIMIZATION ALGORITHMS FOR ADVANCED HYPERGRAPH PARTITIONING, accepted paper",
booktitle="Procceedings of MENDEL 2003",
year="2003",
pages="133--141",
publisher="Faculty of Mechanical Engineering BUT",
address="Brno",
isbn="80-214-2411-7"
}