Two Strategies of Adaptive Cluster Covering with Descent and Their Comparison to Other Algorithms

  • Authors:
  • D. P. Solomatine

  • Affiliations:
  • International Institute for Infrastructural, Hydraulic and Environmental Engineering, P.O. Box 3015, 2601 DA Delft, The Netherlands (e-mail:sol@ihe.nl)

  • Venue:
  • Journal of Global Optimization
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

Two strategies of randomized search, namely adaptive cluster covering (ACCO), and adaptive cluster covering with descent (ACD), are introduced and positioned in the group of the global optimization techniques. Several algorithms based on these new strategies are compared with other techniques of global randomized search in terms of effectiveness, efficiency and reliability. The other techniques include two versions of multistart, two versions of the controlled random search (CRS2 and CRS4) and the canonical genetic algorithm. Thirteen minimization problems including two parameter identification problems (for a flexible membrane mirror model and a hydrologic model) are solved. The algorithm ACCO, and a version of CRS4 algorithm (Ali and Storey 1994) show the highest efficiency, effectiveness and reliability. The second new algorithm, ACD, is in some runs very efficient and effective, but its reliability needs further improvement.