A self-adaptive approach to representation shifts in cultural algorithms

Describes how a formal model of self-adaptation (Angeline, 1995) can be expressed in terms of “cultural algorithms”. A particular form of self-adaptation concerns the shifting of the representational bias used to described the set of learned beliefs within the cultural algorithms. A version of a cultural algorithm with the ability to shift its representational bias was used to solve the “royal road” problem suggested by Mitchell, Forrest and Holland (1991). The results of the presented experiments indicate that representational self-adaptations such as this can produce significant performance improvements over systems without such capabilities for problems whose performance function is inherently hierarchical, as is the case for the royal road function