Synthetic objective function to improve the performance of DE – Initial study
Adam Viktorin, Michal Pluháček, Roman Šenkeřík · AIP conference proceedings · 2017
In this initial study, the idea of synthesizing objective function during the evolution process is tested for the improvement of optimization performance of Differential Evolution (DE) algorithm. Since many of the real world problems require computationally expensive simulations there is a demand for specialized optimization algorithms to solve them in as few objective function evaluations as possible. This paper proposes a new approach which combines DE with Analytical Programming (AP), a symbolic regression tool used for the synthesis of objective function in order to adapt the control parameter settings during evolution.