Towards solving expensive optimization problems with heterogeneous constraint costs
Kamrul Hasan Rahi, Hemant Kumar Singh, Tapabrata Ray · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2024
Expensive constrained optimization problems (ECOPs) are prevalent in practical applications. Surrogate-assisted constrained evolutionary algorithms (SACEAs) are commonly used to solve ECOPs. Existing SACEAs may assess candidate solutions through so-called full evaluations, where all objectives and constraints are unconditionally evaluated, or partial evaluations, where only a subset of them is evaluated. A challenging aspect of the latter category is dealing with cases where individual objective and/or constraint evaluations have different evaluation costs (referred to as heterogeneous costs). While some recent research has been conducted on heterogeneous objectives, scarce attention has been paid to problems involving heterogeneous constraints. Towards addressing this gap, this paper investigates the impact of heterogeneous constraint evaluation costs on algorithm performance. An existing approach (SParEA) that uses partial constraint evaluations is enhanced through a simple update that accounts for the individual constraint costs. The updated algorithm is compared to its base version and its full evaluation version. Numerical experiments are conducted on four systematically constructed types of problems involving different constraint evaluation scenarios to quantitatively demonstrate the benefits of the proposed approach.