Evolved Linker Gene Expression Programming: A New Technique for Symbolic Regression

Abstract

This paper utilises Evolved Linker Gene Expression Programming (EL-GEP), a new variant of Gene Expression Programming (GEP), to solve symbolic regression and sequence induction problems. The new technique was first proposed in [6] to evolve modularity in robotic behaviours. The technique extends the GEP algorithm by incorporating a new alphabetic set (linking set) from which genome linking functions are selected. Further, the EL-GEP algorithm allows the genetic operators to modify the linking functions during the evolution process, thus changing the length of the chromosome during a run. In the current work, EL-GEP has been utilised to solve both symbolic regression and sequence induction problems. The achieved results are compared with those derived from GEP. The results show that EL-GEP is a suitable method for solving optimisation problems.