Review of Efficient Surrogate Infill Sampling Criteria with Constraint Handling

This paper discusses the benefits of different infill sampling criteria used in surrogate-model-based constrained global optimization. Here surrogate models are used to approximate both the objective and constraint functions with the assumption that these are computationally expensive to compute. The construction of these surrogates (also known as meta models or response surface models) involves the selection of a limited number of designs, evaluated using the original expensive functions. Conventionally this involves two stages. First the surrogate is built using an initial sampling plan; the second stage uses infill sampling criteria to select further designs that offer model improvement. This paper provides a comparison of three different infill criteria previously used in constrained global optimization problems. Particular attention is paid to the need to balance the needs of wide ranging exploration and focussed exploitation during global optimization if good results are to be achieved.