Department of Applied Mathematics & Physics, Kyoto University

Technical Report 2007-002 (January 17, 2007)

Genetic Algorithms with Automatic Accelerated Termination
by Abdel-Rahman Hedar, Bun Theang Ong, and Masao Fukushima

pdf File

The standard versions of Evolutionary Algorithms (EAs) have two main drawbacks: unlearned termination criteria and slow convergence. Although several attempts have been made to modify the original versions of Evolutionary Algorithms (EAs), only very few of them have considered the issue of their termination criteria. In general, EAs are not learned with automatic termination criteria, and they cannot decide when or where they can terminate. On the other hand, there are several successful modifications of EAs to overcome their slow convergence. One of the most effective modifications is Memetic Algorithms. In this paper, we modify genetic algorithm (GA), as an example of EAs, with new termination criteria and acceleration elements. The proposed method is called GA with Automatic Accelerated Termination (G3AT). In the G3AT method, Gene Matrix (GM) is constructed to equip the search process with a self-check to judge how much exploration has been done. Moreover, a special mutation operation called ``Mutagenesis