What makes a problem GP-hard? Validating a hypothesis of structural causes

Abstract

This paper provides an empirical test of a hypothesis, which describes the effects of structural mechanisms in genetic programming. In doing so, the paper offers a test problem anticipated by this hypothesis. The problem is tunably difficult, but has this property because tuning is accomplished through changes in structure. Content is not involved in tuning. The results support a prediction of the hypothesis - that GP search space is significantly constrained as an outcome of structural mechanisms. © Springer-Verlag Berlin Heidelberg 2003.