A Vector Division Algorithm for Constructing Orthogonal Latin Hypercube Designs

Latin hypercube designs are commonly used in many applications such as, computer and physical experiments, data selection by sampling techniques and recently in machine learning. The aim of this work is to develop an algorithm concerning their construction. The proposed algorithm is based on a vector division procedure and employs only permutations and negations of vectors. Therefore, it produces the orthogonal Latin hypercube designs cost effectively without performing any floating point computations. The algorithm is compared with other existing methods and it is proved that the developed method produces Latin hypercube designs that are isomorphic with the ones produced by other methods.