Canonical dual \(K\)-g-Bessel sequences and \(K\)-g-frame sequences (Q1743225)

A sequence of operators \(\Lambda_j\in B(\mathcal H,\mathcal K_j)\) is a \(K\)-g-frame for \(\mathcal H\) with respect to the closed subspaces \(\mathcal K_j\subseteq\mathcal K\), if there are constants \(0 < C\leq D<\infty\) (the frame bounds) such that \[ C\| K^\ast f\|^2\leq\sum_j\| \Lambda_j f\|^2\leq D\| f \|^2, \forall f\in\mathcal H, \] where \(\mathcal H\), \(\mathcal K\) are Hilbert spaces and \(K\in B(\mathcal H)\) (see also definition 1.2 of the author). This was initially defined in [\textit{X. C. Xiao} et al. Rocky Mountain J. Math. 45, 675--693 (2015; Zbl 1322.42039)] and generalizes the notions of g-frames and \(K\)-frames simultaneously. As the author comments, the corresponding frame operator is not always invertible, and this presents an obstruction towards the formation of a duality theory of said frames. The author overcomes this obstacle by considering canonical duals of \(K\)-g-Bessel sequences (they are defined as the \(K\)-g-frames, without the lower bound condition), extending the ideas of [\textit{X. X. Guo}, J. Math. Anal. Appl. 444, 598--609 (2016; Zbl 1417.42034)]. Since dual \(K\)-g-Bessel sequences of a given \(K\)-g-frame exist in abundance, the canonical dual is the one whose synthesis operator has the smallest norm, and is unique. This could be seen as an infinite dimensional analogue of the pseudoinverse (or Moore-Penrose inverse) of a linear map.