Phenotypic Forking GA with Moving Windows
Shigeyoshi Tsutsui, Ashish Kumar Ghosh, Masato Takiguchi · 1999
The phenotypic forking GA (p-fGA) which divides the whole search space into sub-spaces using the information of the convergence status of the population and the solutions obtained so far had already been developed. In that work, a neighborhood hypercube was defined around the best individual at the time of forking in the phenotypic feature space with the best solution as the center of that hypercube; and that sub-space was searched in exploitation mode. The rest of the search space was explored. In this investigation, we extend this concept by moving the center of the neighborhood hypercubes with generations. Here, the position of the center of the hypercube is updated every generation such that it becomes the current best solution; thereby dynamically modifying the sub-spaces for exploration and exploitation. This enhances the scope of tracing the optimum solution and gives more flexibility for choosing the size of the hypercube. Empirical results on complex function optimization prob...