A New Iterative Method to Find Polar Decomposition

Abstract

In this paper, we present a novel and efficient iterative approach for computing the polar decomposition of rectangular (or square) complex (or real) matrices. The method proposed herein entails four matrix multiplications in each iteration, effectively circumventing the need for matrix inversions. We substantiate that this method exhibits fourth-order convergence. To illustrate its efficacy relative to alternative techniques, we conduct numerical experiments using randomly generated matrices of dimensions

n\times n

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$n\times n$$\end{document}

, where n assumes values of 80, 90, 100, 120, 150, 180, and 200. Through two illustrative examples, we provide numerical results. We gauge the performance of different methods by calculating essential metrics based on ten matrices for each dimension. These metrics include the average iteration count, the average total matrix multiplication count, the average precision, and the average execution time. Through meticulous comparison, our newly devised method emerges as a proficient and rapid solution, boasting a reduced computational overhead.

Keywords

Polar decomposition / Iterative method / Fourth-order convergent / Matrix multiplications / 15A23 / 65F99

References

[1] Autonne L. Sur les groupes lineaires, reels et orthogonaux. Bull. Sot. Math. France, 1902, 30: 121-134

[2] Cordero A, Torregrosa JR. A sixth-order iterative method for approximating the polar decomposition of an arbitrary matrix. J. Comput. Appl. Math., 2017, 318: 591-598

[3] Esmaeili H. A class of iterative methods for computing polar decomposition. Int. J. Comput. Math., 2011, 88: 207-220

[4] Gander W. Algorithms for polar decomposition. SIAM J. Sci. Stat. Comput., 1990, 11: 1102-1115

[5] Golub G, Van Loan C. Matrix Computations, 20134Baltimore, The Johns Hopkins University Press

[6] Higham, N.J.: Functions of Matrices: Theory and Computation. Society for Industrial and Applied Mathematics, Philadelphia (2008)

[7] Higham NJ, Mackey DS, Mackey N, Tisseur F. Computing the polar decomposition and the matrix sign decomposition in matrix groups. SIAM J. Matrix Anal. Appl., 2004, 25: 1178-1192

[8] Khaksar F, Soleymani F. On a fourth-order matrix method for computing polar decomposition. Comput. Appl. Math., 2015, 34: 389-399

[9] Kiyoumarsi F. Some new high-order computational methods for polar decomposition of complex matrices. Iran. J. Sci. Technol. Trans. A Sci., 2018, 42(4): 2293-2299

[10] Kovarik Z. Some iterative methods for improving orthogonality. SIAM J. Numer. Anal., 1970, 7: 386-389

[11] Petcu D, Popa C. A new version of Kovarik’s approximate orthogonalization algorithm without matrix inversion. Int. J. Comput. Math., 2005, 82: 1235-1246

[12] Soleymani F, Khaksar Haghani F, Shateyi S. Several numerical methods for computing unitary polar factor of a matrix. Adv. Differ. Equ., 2016, 2016(4): 1-11

[13] Soleymani F, Stanimirovic PS, Stojanovic I. A novel iterative method for polar decomposition and matrix sign function. Discrete Dyn. Nat. Soc., 2015, 2015: 1-11

RIGHTS & PERMISSIONS

Shanghai University