A Performance Analysis of Mono and Multi-objective Evolutionary Algorithms Assisted by Meta-modeling

Evolutionary Algorithms can be inefficient in optimizing problems in which fitness evaluation of candidate solutions is computationally expensive. In this paper, single and multi-objective evolutionary methods assisted by meta-models are proposed and analyzed. Meta-models are used to identify promising regions of search space in order to save evaluations of objective-functions. The meta-models are produced using regularized Radial Basis Functions networks. The study in this work shows that the method assisted by meta-modeling accelerates the convergence of the evolutionary process in mono and multi-objectives optimizations.