JAM Journal of Applied Mathematics 1687-0042 1110-757X Hindawi Publishing Corporation 412094 10.1155/2014/412094 412094 Research Article The Solutions to Matrix Equation A X = B with Some Constraints http://orcid.org/0000-0002-5591-9204 Dong Chang-Zhou Zhang Yu-Ping Rafei Morteza School of Mathematics and Science Shijiazhuang University of Economics Shijiazhuang 050031 China sjzue.edu.cn 2014 742014 2014 14 11 2013 15 03 2014 7 4 2014 2014 Copyright © 2014 Chang-Zhou Dong and Yu-Ping Zhang. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Let P be a given Hermitian matrix satisfying P 2 = I . Using the eigenvalue decomposition of P , we consider the least squares solutions to the matrix equation A X = B with the constraints P X = X P and X * = X . A similar problem of this matrix equation with generalized constrained is also discussed.

1. Introduction

Throughout we denote the complex m × n matrix space by m × n . The symbols I ,   A * ,   A - 1 , and A stand for the identity matrix with the appropriate size, the conjugate transpose, the inverse, and the Frobenius norm of A m × n , respectively.

It is a very active research topic to study solutions to various matrix equations . There are many authors who have investigated the classical matrix equation (1) A X = B with different constraints such as symmetric, reflexive, Hermitian-generalized Hamiltonian, and repositive definite . By special matrix decompositions such as singular value decompositions (SVDs) and CS decompositions , Hu and his collaborators , Dai , and Don  have presented the existence conditions and detailed representations of constrained solutions for (1) with corresponding constraints, respectively. For instance, Peng and Hu  presented the eigenvectors-involved solutions to (1) with reflexive and antireflexive constraints; Wang and Yu  derived the bi(skew-)symmetric solutions and the bi(skew-)symmetric least squares solutions with the minimum norm to this matrix equation; Qiu and Wang  proposed an eigenvectors-free method to (1) with P X = X P and X * = s X constraints, where P is a Hermitian involutory matrix and s = ± 1 .

Inspired by the work mentioned above, we focus on the matrix equation (1) with P X = X P and X * = X constraints, which can be described as follows: find X such that (2) { A X - B 2 = min , P X = X P , X * = X } .

Moreover, we also discuss the least squares solutions of (1) with P X = X G P G * and X * = X constraints, where G is a given unitary matrix of order n .

In Section 2, we present the least squares solutions to the matrix equation (1) with the constraints P X = X P and X * = X . In Section 3, we derive the least squares solutions to the matrix equation (1) with the constraints P X = X G P G * and X * = X . In Section 4, we give an algorithm and a numerical example to illustrate our results.

2. Least Squares Solutions to the Matrix Equation (<xref ref-type="disp-formula" rid="EEq1.1">1</xref>) with the Constraints <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M32"> <mml:mi>P</mml:mi> <mml:mi>X</mml:mi> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mo mathvariant="bold">=</mml:mo> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mi>X</mml:mi> <mml:mi>P</mml:mi></mml:math> </inline-formula>  and <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M33"> <mml:msup> <mml:mrow> <mml:mi>X</mml:mi></mml:mrow> <mml:mrow> <mml:mi>*</mml:mi></mml:mrow> </mml:msup> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mo mathvariant="bold">=</mml:mo> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mi>X</mml:mi></mml:math> </inline-formula>

It is required to transform the constrained problem to unconstrained one. To this end, let (3) P = U diag ( I k , - I n - k ) U * be the eigenvalue decomposition of the Hermitian matrix P with unitary matrix U . Obviously, P X = X P holds if and only if (4) diag ( I k , - I n - k ) X ¯ = X ¯ diag ( I k , - I n - k ) , where X ¯ = U * X U . Partitioning (5) X ¯ = ( X 11 X 12 X 21 X 22 ) , X 11 k × k , X 22 ( n - k ) × ( n - k ) , (4) is equivalent to (6) X 12 = - X 12 , X 21 = - X 21 . Therefore, (7) X = U diag ( X 11 , X 22 ) U * , X 11 k × k , a a a a a a a a a a a a a a a a a a X 22 ( n - k ) × ( n - k ) . The constraint X * = X is equivalent to (8) X = U diag ( X 1 , X 2 ) U * , X i * = X i , i = 1,2 , with X 1 k × k , X 2 ( n - k ) × ( n - k ) .

Partition U = ( U 1 , U 2 ) and denote (9) A 1 = A U 1 , A 2 = A U 2 , B 1 = B U 1 , B 2 = B U 2 ; then assume that the singular value decomposition of A 1 and A 2 is as follows: (10) A 1 = M 1 [ Σ 1 0 0 0 ] N 1 * , A 2 = M 2 [ Σ 2 0 0 0 ] N 2 * , where M 1 , M 2 , N 1 , and N 2 are unitary matrices, Σ 1 = diag ( α 1 , , α r ) ,   α i > 0 ( i = 1 , , r ) ,   r = rank ( A 1 ) ,   Σ 2 = diag ( β 1 , , β l ) ,   β j > 0 ( j = 1 , , l ) ,   and   l = rank ( A 2 ) .

Theorem 1.

Given A , B m × n . Then the least squares solutions to the matrix equation (1) with the constraints P X = X P and X * = X can be expressed as (11) X = U diag ( N 1 ( Σ 1 - 1 B 11 + B 11 * Σ 1 - 1 2 Σ 1 - 1 B 12 B 12 * Σ 1 - 1 X 14 ) N 1 * 0 0 N 2 ( Σ 2 - 1 B 21 + B 21 * Σ 2 - 1 2 Σ 2 - 1 B 22 B 22 * Σ 2 - 1 X 24 ) N 2 * ) U * , where X 14 = X 14 * and X 24 = X 24 * are arbitrary matrix.

Proof.

According to (8) and the unitary invariance of Frobenius norm (12) A X - B = A U diag ( X 1 , X 2 ) U * - B = A U diag ( X 1 , X 2 ) - B U . By (9), the least squares problem is equivalent to (13) A X - B = ( A 1 X 1 - B 1 , A 2 X 2 - B 2 ) . We get (14) A X - B 2 = A 1 X 1 - B 1 2 + A 2 X 2 - B 2 2 . According to (10), the least squares problem is equivalent to (15) A X - B 2 = M 1 [ Σ 1 0 0 0 ] N 1 * X 1 - B 1 2 + M 2 [ Σ 2 0 0 0 ] N 2 * X 2 - B 2 2 = [ Σ 1 0 0 0 ] N 1 * X 1 N 1 - M 1 * B 1 N 1 2 + [ Σ 2 0 0 0 ] N 2 * X 2 N 2 - M 2 * B 2 N 2 2 . Assume that (16) N 1 * X 1 N 1 = [ X 11 X 12 X 13 X 14 ] , N 2 * X 2 N 2 = [ X 21 X 22 X 23 X 24 ] , M 1 * B 1 N 1 = [ B 11 B 12 B 13 B 14 ] , M 2 * B 2 N 2 = [ B 21 B 22 B 23 B 24 ] . Then we have (17) A X - B 2 = [ Σ 1 0 0 0 ] [ X 11 X 12 X 13 X 14 ] - [ B 11 B 12 B 13 B 14 ] 2 + [ Σ 2 0 0 0 ] [ X 21 X 22 X 23 X 24 ] - [ B 21 B 22 B 23 B 24 ] 2 = Σ 1 X 11 - B 11 2 + Σ 2 X 21 - B 21 2 + Σ 1 X 12 - B 12 2 + Σ 2 X 22 - B 22 2 + B 13 2 + B 14 2 + B 23 2 + B 24 2 . Hence (18) A X - B 2 = min is solvable if and only if there exist X 11 , X 12 , X 21 , X 22 such that (19) Σ 1 X 11 - B 11 2 = min , Σ 1 X 12 - B 12 2 = min , Σ 2 X 21 - B 21 2 = min , Σ 2 X 22 - B 22 2 = min . It follows from (19) that (20) X 11 = Σ 1 - 1 B 11 + B 11 * Σ 1 - 1 2 , X 12 = Σ 1 - 1 B 12 , X 21 = Σ 2 - 1 B 21 + B 21 * Σ 2 - 1 2 , X 22 = Σ 2 - 1 B 22 . Substituting (20) into (16) and then into (8), we can get that the form of X is (11).

3. Least Squares Solutions to the Matrix Equation (<xref ref-type="disp-formula" rid="EEq1.1">1</xref>) with the Constraints <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M76"> <mml:mi>P</mml:mi> <mml:mi>X</mml:mi> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mo mathvariant="bold">=</mml:mo> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mi>X</mml:mi> <mml:mi>G</mml:mi> <mml:mi>P</mml:mi> <mml:msup> <mml:mrow> <mml:mi>G</mml:mi></mml:mrow> <mml:mrow> <mml:mi>*</mml:mi></mml:mrow> </mml:msup></mml:math> </inline-formula> and <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M77"> <mml:msup> <mml:mrow> <mml:mi>X</mml:mi></mml:mrow> <mml:mrow> <mml:mi>*</mml:mi></mml:mrow> </mml:msup> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mo mathvariant="bold">=</mml:mo> <mml:mo> </mml:mo> <mml:mo> </mml:mo> <mml:mi>X</mml:mi></mml:math> </inline-formula>

In this section, we generalize the constraints P X = X P to P X = X G P G * , where G is a given unitary matrix of order n . Obviously, the constraint is equal to (21) P X G = X G P . Notice that (1) can be equivalently rewritten in (22) A X G = B G . Denoting by Y = X G and setting C = B G , the equation becomes (23) A Y = C , with the constraints P Y = Y P and Y * = Y .

Therefore, the least squares solutions to matrix equation (1) with the constraints P X = X G P G * and X * = X can be solved similar to Theorem 1.

Theorem 2.

Given A , B m × n . Then the least squares solutions to the matrix equation (1) with the constraints P X = X G P G * and X * = X can be expressed as (24) X = U diag ( N 1 ( Σ 1 - 1 C 11 + C 11 * Σ 1 - 1 2 Σ 1 - 1 C 12 C 12 * Σ 1 - 1 Y 14 ) N 1 * 0 0 N 2 ( Σ 2 - 1 C 21 + C 21 * Σ 2 - 1 2 Σ 2 - 1 C 22 C 22 * Σ 2 - 1 Y 24 ) N 2 * ) U * G * , where Y 14 = Y 14 * and Y 24 = Y 24 * are arbitrary matrix.

4. An Algorithm and Numerical Examples

Based on the main results of this paper, we in this section propose an algorithm for finding the least squares solutions to the matrix equation A X = B with the constraints P X = X P and X * = X . All the tests are performed by MATLAB 6.5 which has a machine precision of around 1 0 - 16 .

Algorithm 3.

( 1 ) Input A , B m × n , P n × n and compute U n × n ,   I k k × k ,   - I n - k ( n - k ) × ( n - k ) by the eigenvalue decomposition to P .

( 2 ) Compute A 1 , A 2 , B 1 , B 2 according to (9).

( 3 ) Compute N 1 , N 2 , M 1 , M 2 , Σ 1 , Σ 2 by the singular value decomposition of A 1 , A 2 .

( 4 ) Compute B 11 , B 12 , B 21 , B 22 according to (16).

( 5 ) Compute X by Theorem 1.

Example 4.

Suppose (25) A = [ 0 0 0 0 0 1.2 i 0 0 0 0 0 0.8 i ] , B = [ - 3 - 0.8 i - 1 - 3 i - 1 - 1 - i - 1 9 i - 7 - 2 - 2 2 i - 2 ] , P = [ 1 0 0 0 0 1 0 0 0 0 - 1 0 0 0 0 - 1 ] .

Applying Algorithm 3, we obtain the following: (26) U = [ 0 - i 0 0 i 0 0 0 0 0 0 1 0 0 1 0 ] , A 1 = [ 0 0 1.2 0 0 0 ] , A 2 = [ 0 0 0 0 0.8 i 0 ] , B 1 = [ 0.8 3 . i - i - 1 + i - 2 i 2 i ] , B 2 = [ - 1 - 1 - 3 i - 7 9 i - 2 2 i ] , M 1 = [ 0 i 0 - i 0 0 0 0 1 ] , M 2 = [ 0 1 0 0 0 i 1 0 0 ] , N 1 = [ i 0 0 1 ] , N 2 = [ - i 0 0 i ] , Σ 1 = [ 1.2 ] , Σ 2 = [ 0.8 ] , B 11 = [ i ] , B 12 = [ - 1 - i ] , B 21 = [ 2 i ] , B 22 = [ - 2 ] , X = [ 3 - 0.83 + 0.83 i 0 0 - 0.83 - 0.83 i 0 0 0 0 0 - 2 2.5 0 0 2.5 0 ] .

Conflict of Interests

The authors declare that they have no conflict of interests regarding the publication of this paper.

Acknowledgments

This research was supported by the Natural Science Foundation of Hebei province (A2012403013), the Natural Science Foundation of Hebei province (A2012205028), and the Education Department Foundation of Hebei province (Z2013110).

Wang Q.-W. He Z.-H. A system of matrix equations and its applications Science China Mathematics 2013 56 9 1795 1820 10.1007/s11425-013-4596-y MR3090856 ZBL06231939 Wang Q.-W. He Z.-H. Solvability conditions and general solution for mixed Sylvester equations Automatica 2013 49 9 2713 2719 10.1016/j.automatica.2013.06.009 MR3084457 ZBL06231939 Wang Q.-W. He Z.-H. Some matrix equations with applications Linear and Multilinear Algebra 2012 60 11-12 1327 1353 10.1080/03081087.2011.648635 MR2989766 ZBL06174747 Yuan S.-F. Wang Q.-W. Duan X.-F. On solutions of the quaternion matrix equation A X = B and their applications in color image restoration Applied Mathematics and Computation 2013 221 10 20 10.1016/j.amc.2013.05.069 MR3091901 ZBL0754.15003 Chu K.-E. Symmetric solutions of linear matrix equations by matrix decompositions Linear Algebra and Its Applications 1989 119 35 50 10.1016/0024-3795(89)90067-0 MR1005233 ZBL0688.15003 He Z.-H. Wang Q.-W. A real quaternion matrix equation with applications Linear and Multilinear Algebra 2013 61 6 725 740 10.1080/03081087.2012.703192 MR3005652 ZBL06180251 Kyrchei I. Explicit representation formulas for the minimum norm least squares solutions of some quaternion matrix equations Linear Algebra and Its Applications 2013 438 1 136 152 10.1016/j.laa.2012.07.049 MR2993371 ZBL1255.15022 Li Y. Gao Y. Guo W.-B. A Hermitian least squares solution of the matrix equation A X B = C subject to inequality restrictions Computers & Mathematics with Applications 2012 64 6 1752 1760 10.1016/j.camwa.2012.02.010 MR2960799 ZBL1268.15012 Wu L. The re-positive definite solutions to the matrix inverse problem A X = B Linear Algebra and Its Applications 1992 174 145 151 MR1176457 ZBL0754.15003 Chu K.-E. Singular value and generalized singular value decompositions and the solution of linear matrix equations Linear Algebra and Its Applications 1987 88-89 83 98 10.1016/0024-3795(87)90104-2 MR882442 Paige C.-C. Computing the generalized singular value decomposition SIAM Journal on Scientific and Statistical Computing 1986 7 4 1126 1146 10.1137/0907077 MR857786 ZBL0621.65030 Paige C.-C. Saunders M.-A. Towards a generalized singular value decomposition SIAM Journal on Numerical Analysis 1981 18 3 398 405 10.1137/0718026 MR615522 ZBL0621.65030 Meng C.-J. Hu X.-Y. Zhang L. The skew-symmetric orthogonal solutions of the matrix equation A X = B Linear Algebra and Its Applications 2005 402 303 318 10.1016/j.laa.2005.01.022 MR2141092 ZBL1128.15301 Meng C.-J. Hu X.-Y. An inverse problem for symmetric orthogonal matrices and its optimal approximation Mathematica Numerica Sinica 2006 28 3 269 280 MR2271800 ZBL1128.15301 Zhang Z.-Z. Hu X.-Y. Zhang L. On the Hermitian-generalized Hamiltonian solutions of linear matrix equations SIAM Journal on Matrix Analysis and Applications 2005 27 1 294 303 10.1137/S0895479801396725 MR2176822 ZBL0754.15003 Dai H. On the symmetric solutions of linear matrix equations Linear Algebra and Its Applications 1990 131 1 7 10.1016/0024-3795(90)90370-R MR1057060 ZBL0712.15009 Henk Don F.-J. On the symmetric solutions of a linear matrix equation Linear Algebra and Its Applications 1987 93 1 7 10.1016/S0024-3795(87)90308-9 MR898539 ZBL0622.15001 Peng Z.-Y. Hu X.-Y. The reflexive and anti-reflexive solutions of the matrix equation A X = B Linear Algebra and Its Applications 2003 375 147 155 10.1016/S0024-3795(03)00607-4 MR2013461 ZBL1050.15016 Wang Q.-W. Yu J. On the generalized bi (skew-) symmetric solutions of a linear matrix equation and its procrust problems Applied Mathematics and Computation 2013 219 19 9872 9884 10.1016/j.amc.2013.03.061 MR3055704 Qiu Y.-Y. Wang A.-D. Eigenvector-free solutions to A X = B with P X = X P and X H = s X constraints Applied Mathematics and Computation 2011 217 12 5650 5657 10.1016/j.amc.2010.12.043 MR2770185 ZBL1221.15025