Ensemble of surrogates in black-box-type engineering optimization: Recent advances and applications

Abstract

Due to its high efficiency, surrogate models have been extensively used in black-box-type engineering optimization problems. However, due to the nature of black-box functions, it is difficult to decide which surrogate model is the best for a given problem. In addition, it is difficult to accurately estimate the prediction accuracy of a surrogate on an expensive black-box function in advance. Because we cannot use many real expensive samples to test the prediction accuracy of different surrogates due to the limited computational resources. Ensembles of surrogates fully exploit the advantages of each surrogate model in predicting a given implicit function and have been shown to be effective in improving the robustness of surrogates. The use of ensembles of surrogates to the engineering design optimization has recently attracted huge attention. This work reviews the state-of-the-art ensembles of surrogates and their applications in black-box-type engineering optimization problems. We first familiarize readers with ensembles of surrogates, and give a description of different types of ensembles of surrogates. Further, the recent advances in using ensembles of surrogates for black-box-type engineering optimization are reviewed, and the applications in various optimization needs are highlighted. Finally, we identify the research gaps and most important trends. This review aims to serve as a comprehensive guide that enhances the accessibility of ensembles of surrogates, which can act as an insurance policy in predicting expensive black-box functions and are increasingly used in the field of black-box-type engineering optimization.