Solving an inverse eigenvalue problem with triple constraints on eigenvalues, singular values, and diagonal elements. (19th June 2017)

• Record Type: Journal Article
• Authors: Wu, Sheng-Jhih; Chu, Moody T
• Is Part Of: Inverse problems. Volume 33:Number 8(2017:Aug.)
• Publication Date: 2017-06-19
• Subjects: inverse eigenvalue problem -- majorization relationships -- projected gradient -- projected Hessian -- analytic gradient dynamics; Inverse problems (Differential equations) -- Periodicals; 515.357
• DOI: 10.1088/1361-6420/aa76c4
• Languages: English
• ISSNs: 0266-5611

Abstract

Abstract: An inverse eigenvalue problem usually entails two constraints, one conditioned upon the spectrum and the other on the structure. This paper investigates the problem where triple constraints of eigenvalues, singular values, and diagonal entries are imposed simultaneously. An approach combining an eclectic mix of skills from differential geometry, optimization theory, and analytic gradient flow is employed to prove the solvability of such a problem. The result generalizes the classical Mirsky, Sing–Thompson, and Weyl-Horn theorems concerning the respective majorization relationships between any two of the arrays of main diagonal entries, eigenvalues, and singular values. The existence theory fills a gap in the classical matrix theory. The problem might find applications in wireless communication and quantum information science. The technique employed can be implemented as a first-step numerical method for constructing the matrix. With slight modification, the approach might be used to explore similar types of inverse problems where the prescribed entries are at general locations.