An application of the Kiefer-Wolfowitz equivalence theorem to a problem in Hadamard transform optics (Q1094054): Difference between revisions

Let \(\Omega =\{x=(x_ 1,...,x_ n)^ T:\) \(0\leq x_ i\leq 1\}\) be the unit cube in \(R^ n\). For any probability measure \(\xi\) on \(\Omega\), let \(M(\xi)=\int_{\Omega}xx^ T\xi (dx)\). \textit{M. Harwit} and \textit{N. J. A. Sloane}, Masks for Hadamard transform optics, and weighing designs. Appl. Optics 15, 107-114 (1976), conjectured that if \(X^*\) is the incidence matrix of a balanced incomplete block design (BIBD) with n treatments and n blocks of size \((n+1)/2\), then \(X^*\) minimizes \(tr(X^ TX)^{-1}\) over the \(n\times n\) matrices with entries \(0\leq x_{ij}\leq 1\). This arises from a problem in spectroscopy. In order to solve the conjecture, we consider the more general problem of maximizing \(j_ a(M(\xi))\) over the probability measures on \(\Omega\) for -\(\infty \leq a\leq 1\), where \(j_ 0(M(\xi))=(\det M(\xi))^{1/n}\), \(j_{-\infty}(M(\xi))=the\) minimum eigenvalue of M(\(\xi)\) and \(j_ a(M(\xi))=\{n^{-1} tr[M(\xi)]^ a\}^{1/a}\) for other a's. A complete solution is obtained by using the equivalence theorem in optimal design theory. Let \(\xi_ k\) be the uniform measure on the vertices of \(\Omega\) with k coordinates equal to 1. Then depending on the value of a, optimality is attained by \(\xi_ k\) or a mixture of \(\xi_ k\) and \(\xi_{k+1}\) with \(k\geq [(n+1)/2]\). Optimal \(\xi\) 's with a smaller support can be found by using BIBDs. It follows that if n is odd and \(X^*\) is the block-treatment incidence matrix of a BIBD with n treatments and N blocks of size \((n+1)/2\), then \(X^*\) minimizes \(tr(X^ TX)^ a\) for all \(a<0\) and maximizes \(\det (X^ TX)\) and \(tr(X^ TX)^ a\) for all \(0<a\leq 1-\ln (n/2+1)/\ln (n+1)\) over the \(N\times n\) matrices with entries \(0\leq x_{ij}\leq 1\). Similar results are derived for the even case and the incidence matrices of BIBDs of larger block sizes.