A Survey of Surrogate Approaches for Expensive Constrained Black-Box Optimization

Numerous practical optimization problems involve black-box functions whose values come from computationally expensive simulations. For these problems, one can use surrogates that approximate the expensive objective and constraint functions. This paper presents a survey of surrogate-based or surrogate-assisted methods for computationally expensive constrained global optimization problems. The methods can be classified by type of surrogate used (e.g., kriging or radial basis function) or by the type of infill strategy. This survey also mentions algorithms that can be parallelized and that can handle infeasible initial points and high-dimensional problems.