scientific article; zbMATH DE number 2128113

From MaRDI portal
Publication:3155152





An Hadamard matrix is a square matrix \(H\) of order \(n\) with \(n= 1,2\), or \(4t\) with all entries \(1\) or \(-1\) and which satisfies \(HH^T= nI_n\). These matrices are used extensively in coding and communications [\textit{J. Seberry} and \textit{M. Yamada}, Contemporary desing theory, Collect. Surv., 431--560 (1992; Zbl 0776.05028)]. In the present paper the authors survey a large number of algorithms for the construction of Hadamard matrices, and they begin Section 1 with results from the use of Williamson matricies [(*) \textit{J. Horton}, \textit{C. Koukouvinos} and \textit{J. Seberry}, Bull. Inst. Comb. Appl. 35, 75--88 (2002; Zbl 0990.05018)] of order \(n\) to get Hadamard matrices of order \(4n\). Then they describe over several subsections two recent intricate algorithms of this type of their own. They used a server which broke the work into smaller portions and at various times as many as 37 other computers. The first method of theirs used \(n\)th roots of unity of \(n\) and the second dealt with \(n = pq\) [\textit{C. Koukouvinos} and \textit{S. Kounias}, Discrete Math. 68, 45--57 (1988; Zbl 0667.05010)] and used circulant matrices. Results for orders \(n= 100\) to \(n= 180\) are included in [(*)]. Section 1 also contains methods of obtaining Hadamard matrices from generalized Legendre pairs using the discrete Fourier transform, from generalized Legendre pairs using supplementary difference sets, and from two circulant matrices. Section 2 deals with inequivalent Hadamard matrices which have been normalized. Since one can multiply a row or column of a Hadamard matrix without changing the the equation \(HH^T= nI_n\), one can normalize a Hadamard matrix so that its first row and column are all ones. Equivalency is a difficult computational problem even for normalized matrices. The results so far are: Hadamard matrices of orders less than 16 are unique up to equivalence (any sequence of row negations, row permutations, column negations, column permutations); there are five equivalence classes for \(n= 16\); three equivalence classes of order 20; sixty equivalence classes of order 24; there are 487 classes of order 28. The classification of orders \(n\) larger than 28 is still open (exhaustive search is an NP hard problem). Section 3 deals with algorithms for constructing orthogonal designs and is partly based on [\textit{A. V. Geramita} and \textit{J. Seberry}, Orthogonal designs. Quadratic forms and Hadamard matrices (Lecture Notes in Pure and Applied Mathematics 45, New York, Basel: Marcel Dekker) (1979; Zbl 0411.05023)] but contains insightful summaries of recent work. Section 4 deals with short amicable sets and Kharaghani type orthogonal designs and is quite recent, based on results in [\textit{S. Georgiou}, \textit{C. Koukouvinos} and \textit{J. Seberry}, Int. J. Appl. Math. 9, No. 2, 161--187 (2002; Zbl 1024.05013)]. The authors have masterfully summarized a huge body of work for the interested reader (38 of the 90 citations are due to one or more of the co-authors). For the entire collection see [Zbl 1050.05003].




Cited in
(31)


Describes a project that uses

Uses Software






This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3155152)