A surrogate assisted differential evolution to solve constrained optimization problems
Rafael de Paula Garcia, Beatriz Souza Leite Pires de Lima, Afonso Celso de Castro Lemonge · 2017
Many real optimization problems are defined by functions whose evaluations are very expensive in terms of time consuming. This is an important issue to the applications of evolutionary algorithms, which demand a large number of function evaluations. In this sense, surrogate models can provide good approximations maintaining the accuracy of the search. This paper presents a similarity-based surrogate coupled to a differential evolution to solve constrained optimization problems. The database management, where solutions evaluated by the exact model are stored and used to approximate other solutions, is done according to a “merit” scheme. A set of 24 constrained benchmark problems is used in the computational experiments and the results showed that the proposed scheme achieves good solutions with a reduced number of function evaluations.