Parameterised Indexed FOR-Loops in Genetic Programming and Regular Binary Pattern Strings

Abstract

We present two methods to represent and use parameterised indexed FOR-loops in genetic programming. They are tested on learning the repetitive unit of regular binary pattern strings to reproduce these patterns to user specified arbitrary lengths. Particularly, we investigate the effectiveness of low-level and high-level functions inside these loops for the accuracy and the semantic efficiency of solutions. We used 5 test cases at increasing difficulty levels and our results show the high-level approach producing solutions in at least 19% of the runs when the low-level approach struggled to produce any in most cases.