Generalised rank-constrained approximations of Hilbert-Schmidt operators on separable Hilbert spaces and applications

• Published in: arXiv.org 2024-08
• Main Authors: Carere, Giuseppe, Han Cheng Lie
• Format: Article
• Language: English
• Subjects: Approximation; Hilbert space; Linear operators; Matrix methods; Operators (mathematics); Singular value decomposition
• ISSN: 2331-8422

In this work we solve, for given bounded operators \(B,C\) and Hilbert--Schmidt operator \(M\) acting on potentially infinite-dimensional separable Hilbert spaces, the reduced rank approximation problem, \(\text{min}\{\Vert{M-BXC}\Vert_{HS}:\ \text{dim}\ \text{ran}(X)\leq r\}\). This extends the result of Sondermann (Statistische Hefte, 1986) and Friedland and Torokhti (SIAM J. Matrix Analysis and Applications, 2007), which studies this problem in the case of matrices \(M\), \(B\), \(C\), \(X\), and the analysis involves the Moore-Penrose inverse. In classical approximation problems that can be solved by the singular value decomposition or Moore--Penrose inverse, the solution satisfies a minimal norm property. Friedland and Torokhti also state such a minimal norm property of the solution. However, we show that this minimal norm property does not hold in general and give a modified minimality property that is satisfied in general. We also show that the solution may be discontinuous in infinite-dimensional settings. Necessary and sufficient conditions for continuity of the solutions are given and continuous approximations are constructed when such conditions are not met. Finally, we study problems from signal processing, reduced rank regression and linear operator learning under a rank constraint. We show that our theoretical results enable us to explicitly find the solutions to these problems and to fully characterise their existence, uniqueness and minimality property.