Watson, Richard A., and Jordan B. Pollack. “Hierarchically consistent test problems for genetic algorithms: Summary and additional results.” In Procs. of 1999 Congress on Evolutionary Computation, CEC-99 . 1999.
URL1
This paper gives additional data for experiments presented in previous work on hierarchically consistent test problems. The experiments utilize Hierarchical-if-and-only-if (H-IFF), the basic example of a hierarchically consistent building-block problem, and several variants of H-IFF designed to enable the difficulty of the function to be `tuned’. We review the H-IFF function and its variants, and give data showing the performance of the regular GA and a fitness-sharing GA for various parameters affecting difficulty. 1 Introduction The Building-Block Hypothesis [Holland 1975, Goldberg 1989] suggests that the GA will perform well when it is able to identify above-average-fitness low-order schemata and recombine them to produce higher-order schemata of higher fitness. We suppose that the recombinative process continues recursively, combining schemata of successively higher orders as search progresses.