Comparison of three novelty approaches to constants (Ks) handling in analytic programming powered by SHADE

Abstract

This research deals with the comparison of three novelty approaches for constant estimation in analytic programming (AP) powered by Success-history based Differential evolution (SHADE). AP is a tool for symbolic regression tasks which enables to synthesise an analytical solution based on the required behaviour of the system. This paper offers another strategy to already known and used by the AP from the very beginning and approaches published recently in 2016. This paper compares these procedures and the discussion also includes nonlinear fitting and metaevolutionary approach. As the main evolutionary algorithm, a differential algorithm in the version SHADE for the main process of AP is used. The proposed comparison is performed out on quintic, sextic, Sine 3 and Sine 4 benchmark problems.