Loading...

An improved real-coded bayesian optimization algorithm for continuous global optimization

Moradabadi, B ; Sharif University of Technology | 2013

965 Viewed
  1. Type of Document: Article
  2. Publisher: 2013
  3. Abstract:
  4. Bayesian optimization algorithm (BOA) utilizes a Bayesian network to estimate the probability distribution of candidate solutions and creates the next generation by sampling the constructed Bayesian network. This paper proposes an improved real-coded BOA (IrBOA) for continuous global optimization. In order to create a set of Bayesian networks, the candidate solutions are partitioned by an adaptive clustering method. Each Bayesian network has its own structure and parameters, and the next generation is produced from this set of networks. The adaptive clustering method automatically determines the correct number of clusters so that the probabilistic building-block crossover (PBBC) is effectively preserved. This leads to a better search when the diversity of population is high at the beginning of search. Moreover, it tunes the solutions by automatically decreasing the number of clusters as the diversity of population decreases during the search process. The experimental results demonstrate that the proposed algorithm achieves better performance on well-known benchmark functions in the continuous global optimization
  5. Keywords:
  6. Adaptive clustering method ; Continuous global optimization ; Adaptive clustering ; Bayesian optimization algorithms ; Benchmark functions ; Better performance ; Diversity of populations ; Number of clusters ; probabilistic building-block crossover ; Search process ; Benchmarking ; Cluster analysis ; Evolutionary algorithms ; Global optimization ; Next generation networks ; Probability distributions ; Bayesian networks
  7. Source: International Journal of Innovative Computing, Information and Control ; Volume 9, Issue 6 , 2013 , Pages 2505-2519 ; 13494198 (ISSN)
  8. URL: http://www.ijicic.org/ijicic-12-03005.pdf