GPU-accelerated differential evolutionary Markov Chain Monte Carlo method for multi-objective optimization over continuous space
Department of Industrial Engineering, Lamar University, 211 Redbird Ln, P.O. Box 10032, Beaumont, TX 77710
In Proceeding of the 2nd workshop on Bio-inspired algorithms for distributed systems (2010), pp. 1-8
@conference{zhu2010gpu,
title={GPU-accelerated differential evolutionary Markov Chain Monte Carlo method for multi-objective optimization over continuous space},
author={Zhu, W. and Li, Y.},
booktitle={Proceeding of the 2nd workshop on Bio-inspired algorithms for distributed systems},
pages={1–8},
year={2010},
organization={ACM}
}
In this paper, the attractive features of evolutionary algorithm and Markov Chain Monte Carlo are combined into a new Differential Evolutionary Markov Chain Monte Carlo (DE-MCMC) method for multi-objective optimization problems with continuous variables. The DE-MCMC evolves a population of Markov chains through differential evolution (DE) toward a diversified set of solutions at the Pareto optimal front in the multi-objective function space. The computational results show the effectiveness of the DE-MCMC algorithm in a variety of standardized test functions as well as a protein loop structure sampling application. Moreover, the DE-MCMC algorithm can efficiently take advantage of the massive-parallel, many-core architecture, where its implementation on GPU can achieve speedup of 14~35.
December 17, 2010 by hgpu
