Deterministic sampling designs with local methods for optimization problems

The problem of numerically minimizing a functional cost involves the following issues: (i) the definition of a sampling of the domain where the functional is evaluated; (ii) the choice of a class of models to approximate the solution. This work presents a comparison of performances among several deterministic sampling designs (low-discrepancy sequences, orthogonal arrays and Latin hypercubes) in case local models are employed, in different functional optimization contexts.