A note on the existence of the hyperbolic singular value decomposition (Q1915616): Difference between revisions

The hyperbolic singular value decomposition was proposed as a decomposition more suitable for the numerical computation of the eigendecomposition of the difference of the outer product \(H = A_1A_1^H -A_2A_2^H\), where explicit formation of \(H\) is avoided for reasons of numerical accuracy (cf. R. Onn, A.O. Steinhardt, and A. Bojanczyk, \textit{The hyperbolic singular value decomposition and applications}, IEEE Trans. Signal Process. 39: 1575-1588 (1991)). The basic idea is first to write \(H\) in the product form \(H = A{\Phi}A^H\) with \(A = (A_1,A_2), \Phi = \text{diag}(I,-I)\) and then apply a unitary matrix \(U\) and hyperexchange matrix \(V\) to \(A\) such that \(A = U{\Sigma}V^H\), where \(\Sigma\) is diagonal. The above decomposition is not always possible for a general matrix \(A\). In the paper under review the author presents a version of the hyperbolic singular value decomposition for a matrix pair \((A,B)\) with \(A\) Hermitian. The hyperbolic singular value decomposition comes as a special case of the decomposition with \(A\) set to be the signature matrix.