Surrogate-assisted cooperative coevolutionary optimization with a new grouping strategy for large-scale expensive problems

Large-scale expensive optimization problems (LSEOPs) often face significant challenges because of both the expensive objective evaluation and dimensional curse. A few surrogate-assisted evolutionary algorithms (SAEAs) have been proposed for addressing LSEOPs in the cooperative coevolutionary framework. However, good methods are still expected for finding better solutions for LSEOPs in a limited budget. Thus, in this approach, we introduced a new grouping technology to enhance the chances of grouping correlational variables into the same sub-problems. Then the search strategy is modified on the competitive swarm optimizer and two different approaches are proposed for infill sampling in two stages, respectively. We conducted experiments using the CEC’2013 test suite and compared our results to four outstanding approaches for LSEOPs. The results reflect that our proposed algorithm achieves better experimental results when dealing with expensive LSOPs.