Nonnegative Moore-Penrose inverse of Gram matrices in an indefinite inner product space (Q1024248): Difference between revisions

An indefinite inner product is defined on \(\mathbb R^n\) by \([x,y]=\langle x,Ny\rangle\) where \(\langle\cdot,\cdot\rangle\) denotes the Euclidian inner product and the weight \(N\in \mathbb R^{n\times n}\) is hermitian with \(N=N^{-1}\). A space with an indefinite inner product is said to be an indefinite inner product space. For \(A\in \mathbb R^{m\times n}\) and \(B\in \mathbb R^{n\times p}\) the indefinite matrix product of \(A\) and \(B\) is defined by \(A\circ B=ANB\). The adjoint \(A^{[\star]}\) of the matrix \(A\) is defined by \(A^{[\star]}=NA^TM\) where \(M\) and \(N\) are weights of order \(m\) and \(n\) respectively. The Gram matrix of \(A\) is defined by \(A^{[\star]}\circ A\). Necessary and sufficient conditions are provided for the nonnegativity of the Moore-Penrose inverse of a Gram matrix defined in an indefinite inner product space. These conditions include the acuteness of a pair of closed convex cones.