"Annals. Computer Science Series" Journal Romania, 300559 Timişoara, 6 Lascăr Catargiu str. Phone: 004 0256 220 687  Fax: 004 0256 220 690 Email: conference.fcia [@] tibiscus [.] ro  conference.fcia [@] gmail [.] com 



» An Algorithm for Constructing Orthogonal Array Latin Hypercube Designs using Hadamard Matrices of Orders 4λ and 8λ » Kazeem A. Osuolale, Babatunde L. Adeleke and Waheed B. Yahya ABSTRACT: Orthogonal array Latin hypercube designs have become popular in practice among strategies used for developing computer experiments. Hadamard matrices have been used to construct orthogonal arrays based on the connection between Hadamard matrices and orthogonal arrays (OAs). A Hadamard matrix is a square matrix of +1 and 1 whose rows are orthogonal. This study aimed at proposing an enhanced algorithm that employed the maximin criterion in the kNearest Neighbour with Euclidean distance for constructing optimal spacefilling design called Orthogonal Array Latin Hypercube Design (OALHD). Orthogonal arrays (OAs) were constructed from Hadamard matrices of orders 4λ and 8λ which are subsequently used to construct the desired OALHD. The Orthogonal array (n, k) Latin hypercube designs were constructed at parameter values of OA (n, k, s, t, λ) = (8, 7, 2, 2, 2) and (16, 8, 2, 3, 2). The OA (8, 7) LHD and OA (16,8) LHD constructed have better spacefilling properties and they achieve uniformity in each dimension. MATLAB 2015 computer package was used for the development of the algorithm that constructs the OALHDs. KEYWORDS:Computer experiments, Hadamard matrices, Latin hypercube designs, Orthogonal array, Spacefilling designs. » Received: 20 November 2016» Accepted: 21 July 2017 
9  18

» www.fcia.tibiscus.ro  » www.ccsitia.tibiscus.ro  » www.geogebra.tibiscus.ro  » www.tibiscus.ro   © 20032017, Last update: 26.07.2017 / 11:15 