MILP models for the selection of a small set of well-distributed points

Abstract

Motivated by the problem of fitting a surrogate model to a set of feasible points in the context of constrained derivative-free optimization, we consider the problem of selecting a small set of points with good space-filling and orthogonality properties from a larger set of feasible points. We propose four mixed-integer linear programming models for this task and we show that the corresponding optimization problems are NP-hard. Numerical experiments show that our models consistently yield well-distributed points that, on average, help reducing the variance of model fitting errors.