Seminars: Dept Seminar

Sliced Space-filling Designs

Abstract:

Space-filling designs are becoming increasingly popular in computer experiments, numerical integration, stochastic optimization and machine learning. The standard framework for space-filling designs assumes that all the factors are quantitative.  We propose an approach to constructing a new type of design, called sliced space-filling design, to accommodate both qualitative and quantitative factors. It starts with constructing a Latin hypercube design based on a special orthogonal array for the quantitative factors and then partitions the design into groups corresponding to different level combinations of the qualitative factors. The points in each group have good space-filling properties. Sampling properties of the constructed designs will be discussed. Sliced space-filling designs are useful for computer experiments with qualitative and quantitative factors, ensembles of multiple computer models and cross-validation.