On bounds for scaled projections and pseudoinverses (Q913886): Difference between revisions

The paper builds up on results of \textit{G. W. Stewart} [Linear Algebra Appl. 112, 189-193 (1989; Zbl 0658.15003)], how to determine for a matrix X of full column rank and for positive definite diagonal matrices D a value p such that \(\sup_{D}\| X^+_ D\| \leq p^{-1}\| X^+\|\) for the weighted pseudoinverse \(X^+_ D:=(X^ TDX)^{- 1}X^ TD.\) Let the columns of U be an orthogonal basis of R(X) and let \(U_ I\) denote any submatrix formed from a nonempty set of rows of U. Then the author shows that \(p\geq m:=\min_{U_ I}\{smallest\) nonzero singular value of \(U_ I\}\). Together with Stewart's result \(p\leq m\) this proves \(p=m\).