When is a matrix unitary or Hermitian plus low rank?

Hermitian and unitary matrices are two representatives of the class of normal matrices whose full eigenvalue decomposition can be stably computed in quadratic computing com plexity. Recently, fast and reliable eigensolvers dealing with low rank perturbations of unitary and Hermitian matrices were proposed. These structured eigenvalue problems appear naturally when computing roots, via confederate linearizations, of polynomials expressed in, e.g., the monomial or Chebyshev basis. Often, however, it is not known beforehand whether or not a matrix can be written as the sum of an Hermitian or unitary matrix plus a low rank perturbation. We propose necessary and sufficient conditions characterizing the class of Hermitian or unitary plus low rank matrices. The number of singular values deviating from 1 determines the rank of a perturbation to bring a matrix to unitary form. A similar condition holds for Hermitian matrices; the eigenvalues of the skew-Hermitian part differing from 0 dictate the rank of the perturbation. We prove that these relations are linked via the Cayley transform. Based on these conditions we are able to identify the closest Hermitian and unitary plus low rank matrix in Frobenius and spectral norm and a practical Lanczos iteration to detect the low rank perturbation is presented. Numerical tests prove that this straightforward algorithm is robust with respect to noise.

[1]  Charles R. Johnson,et al.  Matrix analysis , 1985, Statistical Inference for Engineers and Data Scientists.

[2]  Marko Huhtanen,et al.  A Stratification of the Set of Normal Matrices , 2001, SIAM J. Matrix Anal. Appl..

[3]  David W. Lewis,et al.  Matrix theory , 1991 .

[4]  B. Parlett The Symmetric Eigenvalue Problem , 1981 .

[5]  Raf Vandebril,et al.  Fast and backward stable computation of the eigenvalues of matrix polynomials , 2016 .

[6]  Israel Gohberg,et al.  Efficient eigenvalue computation for quasiseparable Hermitian matrices under low rank perturbations , 2008, Numerical Algorithms.

[7]  Raf Vandebril,et al.  Fast and stable unitary QR algorithm , 2015 .

[8]  Paul Van Dooren,et al.  Implicit double shift QR-algorithm for companion matrices , 2010, Numerische Mathematik.

[9]  Yuli Eidelman,et al.  Implicit QR for rank-structured matrix pencils , 2014 .

[10]  L. Reichel,et al.  A divide and conquer method for unitary and orthogonal eigenproblems , 1990 .

[11]  Lothar Reichel,et al.  The Arnoldi Process and GMRES for Nearly Symmetric Matrices , 2008, SIAM J. Matrix Anal. Appl..

[12]  S. Chandrasekaran,et al.  Fast and stable eigendecomposition of symmetric banded plus semi-separable matrices , 2000 .

[13]  Angelika Bunse-Gerstner,et al.  Schur parameter pencils for the solution of the unitary eigenproblem , 1991 .

[14]  Gianna M. Del Corso,et al.  Fast QR iterations for unitary plus low rank matrices , 2020, Numerische Mathematik.

[15]  Raf Vandebril,et al.  Efficient Ehrlich–Aberth iteration for finding intersections of interpolating polynomials and rational functions , 2017 .

[16]  Raf Vandebril,et al.  Fast and backward stable computation of eigenvalues and eigenvectors of matrix polynomials , 2016, Math. Comput..

[17]  Dario Bini,et al.  A Fast Implicit QR Eigenvalue Algorithm for Companion Matrices , 2010 .

[18]  R. C. Thompson Principal submatrices IX: Interlacing inequalities for singular values of submatrices , 1972 .

[19]  Nathan Halko,et al.  Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions , 2009, SIAM Rev..

[20]  Thomas Huckle The Arnoldi Method for Normal Matrices , 1994 .

[21]  L. Elsner,et al.  Normal matrices: an update , 1998 .

[22]  S. Barnett Polynomials and linear control systems , 1983 .

[23]  Luca Gemignani,et al.  A unitary Hessenberg QR-based algorithm via semiseparable matrices , 2005 .

[24]  John Chollet,et al.  On principal submatrices , 1982 .

[25]  R. Vandebril,et al.  Matrix Computations and Semiseparable Matrices: Linear Systems , 2010 .

[26]  Inderjit S. Dhillon,et al.  The design and implementation of the MRRR algorithm , 2006, TOMS.

[27]  Luca Gemignani,et al.  Fast Hessenberg Reduction of Some Rank Structured Matrices , 2017, SIAM J. Matrix Anal. Appl..

[28]  Fei Xue,et al.  Short-Term Recurrence Krylov Subspace Methods for Nearly Hermitian Matrices , 2012, SIAM J. Matrix Anal. Appl..

[29]  Leandro Moral,et al.  Five-diagonal matrices and zeros of orthogonal polynomials on the unit circle , 2002 .

[30]  Gianna M. Del Corso,et al.  A CMV-Based Eigensolver for Companion Matrices , 2015, SIAM J. Matrix Anal. Appl..

[31]  Raf Vandebril,et al.  On an Economic Arnoldi Method for BML-Matrices , 2018, SIAM J. Matrix Anal. Appl..

[32]  David S. Watkins,et al.  The matrix eigenvalue problem - GR and Krylov subspace methods , 2007 .

[33]  Yousef Saad,et al.  Iterative methods for sparse linear systems , 2003 .

[34]  Federico Poloni,et al.  Factoring Block Fiedler Companion Matrices , 2019, Structured Matrices in Numerical Linear Algebra.

[35]  James Demmel,et al.  Performance and Accuracy of LAPACK's Symmetric Tridiagonal Eigensolvers , 2008, SIAM J. Sci. Comput..

[36]  Dario A. Bini,et al.  Quasiseparable Hessenberg reduction of real diagonal plus low rank matrices and applications , 2015, 1501.07812.

[37]  R. Vandebril,et al.  Matrix Computations and Semiseparable Matrices , 2007 .

[38]  Froilán M. Dopico,et al.  Condition numbers for inversion of Fiedler companion matrices , 2013 .

[39]  William B. Gragg,et al.  The QR algorithm for unitary Hessenberg matrices , 1986 .

[40]  I. Gohberg,et al.  Fast QR Eigenvalue Algorithms for Hessenberg Matrices Which Are Rank-One Perturbations of Unitary Matrices , 2007, SIAM J. Matrix Anal. Appl..

[41]  Jianlin Xia,et al.  A Fast QR Algorithm for Companion Matrices , 2007 .

[42]  Teri Barth,et al.  Multiple Recursion Conjugate Gradient Algorithms Part I: Sufficient Conditions , 2000, SIAM J. Matrix Anal. Appl..

[43]  Jörg Liesen When is the Adjoint of a Matrix a Low Degree Rational Function in the Matrix? , 2007, SIAM J. Matrix Anal. Appl..

[44]  Zdenek Strakos,et al.  On Optimal Short Recurrences for Generating Orthogonal Krylov Subspace Bases , 2008, SIAM Rev..

[45]  Dario A. Bini,et al.  On a Class of Matrix Pencils and $\ell$-ifications Equivalent to a Given Matrix Polynomial , 2014, 1406.1025.

[46]  Raf Vandebril,et al.  An Implicit Multishift $QR$-Algorithm for Hermitian Plus Low Rank Matrices , 2010 .

[47]  T. Chan Rank revealing QR factorizations , 1987 .

[48]  J. Neumann,et al.  Allgemeine Eigenwerttheorie Hermitescher Funktionaloperatoren , 1930 .

[49]  Raf Vandebril,et al.  Computing the eigenvalues of symmetric tridiagonal matrices via a Cayley transformation , 2017 .

[50]  David S. Watkins,et al.  Chasing algorithms for the eigenvalues problem , 1991 .

[51]  I. Gohberg,et al.  Implicit QR with Compression , 2012 .

[52]  Paul Van Dooren,et al.  A Framework for Structured Linearizations of Matrix Polynomials in Various Bases , 2016, SIAM J. Matrix Anal. Appl..