An integer programming model for obtaining cyclic quasi-difference matrices

Abstract

Orthogonal arrays are of great importance in mathematical sciences. This paper analyses a certain practical advantage of quasi-difference matrices over difference matrices to obtain orthogonal arrays with given parameters. We also study the existence of quasi-difference matrices over cyclic groups originating orthogonal arrays with $t=2$ and $\lambda =1$, proving their existence for some parameters sets. Moreover, we present an Integer Programming model to find such quasi-difference matrices and also a Bimodal Local Search algorithm to obtain them. We provide a conjecture related to the distributions of differences along rows and columns of arbitrary square matrices with entries in a cyclic group in positions outside the main diagonal which shows an intriguing symmetry, and we prove it when the matrix is a quasi-difference matrix.