Remapping Subpartitions of Hyperspace Using Iterative Genetic Search Remapping Subpartitions of Hyperspace Using Iterative Genetic Search Remapping Subpartitions of Hyperspace Using Iterative Genetic Search

Various strategies for remapping hyperspace associated with two iterative genetic algorithms are reviewed, compared and empirically tested on a particular \real-world" problem. Both algorithms reduce the hypercube at each iteration such that each new mapping samples only subpartitions of the search space. Delta coding remaps the search space at each iteration by sampling the subspace numerically adjacent to the previous solution. Canonical delta folding canonically reorders and folds hyperspace producing new mappings that sample subpartitions distributed across the search space. Empirical results suggest that more local sampling strategies generally produce better solutions and that identi cation of \good" mappings is di cult.