The reduction of computational cost of evolutionary multiobjective Pareto-optimization algorithms is necessary when time-consuming objective functions evaluation is required. To this end neural network interpolation techniques are used for objective response surface building; evolutionary multiobjective Pareto-optimization is then performed on interpolated functions. Both analytical multiobjective test problems and numerical electromagnetic design problems are considered; the study of Pareto optimal front interpolation accuracy versus neural network training cost is performed.
No takes yet. Share an insight, caveat, or question.
Bramanti et al. (2001) studied this question.