Interactive genetic algorithms with variational population size

  • Authors:

  • Jie Ren;Dun-Wei Gong;Xiao-Yan Sun;Jie Yuan;Ming Li

  • Affiliations:

  • School of Information and Electrical Engineering, China University of Mining and Technology, Xuzhou, China;School of Information and Electrical Engineering, China University of Mining and Technology, Xuzhou, China;School of Information and Electrical Engineering, China University of Mining and Technology, Xuzhou, China;School of Information and Electrical Engineering, China University of Mining and Technology, Xuzhou, China;School of Information and Electrical Engineering, China University of Mining and Technology, Xuzhou, China

  • Venue:
  • ICIC'09 Proceedings of the Intelligent computing 5th international conference on Emerging intelligent computing technology and applications
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Traditional interactive genetic algorithms often have small population size because of a limited human-computer interface and user fatigue, which restricts these algorithms' performances to some degree. In order to effectively improve these algorithms' performances and alleviate user fatigue, we propose an interactive genetic algorithm with variational population size in this paper. In the algorithm, the whole evolutionary process is divided into two phases, i.e. fluctuant phase and stable phase of the user's cognition. In fluctuant phase, a large population is adopted and divided into several coarse clusters according to the similarity of individuals. The user only evaluates these clusters' centers, and the other individuals' fitness is estimated based on the acquired information. In stable phase, the similarity threshold changes along with the evolution, leading to refined clustering of the population. In addition, elitist individuals are reserved to extract building blocks. The offspring is generated based on these building blocks, leading to a reduced population. The proposed algorithm is applied to a fashion evolutionary design system, and the results validate its efficiency.