An Ensemble Surrogate-Based Framework for Expensive Multiobjective Evolutionary Optimization

Surrogate-assisted evolutionary algorithms (SAEAs) have become very popular for tackling computationally expensive multiobjective optimization problems (EMOPs), as the surrogate models in SAEAs can approximate EMOPs well, thereby reducing time cost of process. However, with increased number decision variables EMOPs, prediction accuracy will deteriorate, which inevitably worsens performance SAEAs. To deal this issue, article suggests an ensemble surrogate-based framework EMOPs. In framework, a global model is trained under entire search space to explore area, while submodels are different subspaces exploit subarea, so enhance and reliability. Moreover, new infill sampling criterion designed based on set reference vectors select promising samples training models. validate generality effectiveness our three state-of-the-art [nondominated sorting genetic algorithm III (NSGA-III), decomposition differential evolution (MOEA/D-DE) vector-guided (RVEA)] embedded, significantly improve their solving most test adopted article. When compared some competitive up 30 variables, experimental results also advantages approach cases.