Fitness Inference System Considering Similarity of Chomosomes.
Hiroharu Kawanaka, Tomohiro Yoshikawa, Reiko Mitsuhashi, Yoshifumi Banno, Tsuyoshi Shinogi, Shinji Tsuruoka · IEEJ Transactions on Electronics Information and Systems · 2003
Recently, Evolutionary Computations (ECs) are widely studied. Generally, it takes a lot of time for the calculation of ECs to acquire expected solutions because they need repeated calculation for searching solutions. In this paper, we propose a new fitness inference method with similarity of chromosomes to reduce the total number of evaluation in GA. This method uses an idea of territory that represents the threshold of similarity of chromosomes, and it infers most of the fitness values of chromosomes without actual evaluation process using other evaluated chromosomes. The time for evaluation can be reduced drastically and the effective search can be done by the proposed method. This paper also studies the territory of the proposed method.