Using Building Block Functions to Investigate a Building Block Hypothesis for Genetic Programming

This paper presents building block functions, i.e., functions in which explicit schemas of high fitness are defined (BB functions, for short) which are useful in investigating the character of Genetic Programming (GP) search. One conjecture we believe to be answerable through experimentation with these functions is whether GP power that, when primary partial solutions are discovered, their numbers increase and GP crossover is able to combine them into increasingly larger sub-solutions and eventually find the solution? The functions should also provide insight into more detailed aspects of the roles of GP crossover and GP genotype growth.