Comparing some graph crossover in genetic network programming
Hideki Katagiri, Kotaro Hirasawa, Jinglu Hu, Junichi Murata · 2003
In this paper, we studied the crossover for graph-based programs. The graph crossover is unsatisfactory on several counts to divide and combine graphs unlike with string or tree crossover, because a relatively large number of random factors should be inevitable to operate the crossover. In this paper, we compared the performance of several crossover operators for the genetic network programming experimentally. The experimental results show the advantages and drawbacks of each method.