Citation: Ernest Greene, Michael J. Hautus. Demonstrating Invariant Encoding of Shapes Using A Matching Judgment Protocol[J]. AIMS Neuroscience, 2017, 4(3): 120-147. doi: 10.3934/Neuroscience.2017.3.120
[1] | Mahmoud S. Mehany, Faizah D. Alanazi . An η-Hermitian solution to a two-sided matrix equation and a system of matrix equations over the skew-field of quaternions. AIMS Mathematics, 2025, 10(4): 7684-7705. doi: 10.3934/math.2025352 |
[2] | Abdur Rehman, Muhammad Zia Ur Rahman, Asim Ghaffar, Carlos Martin-Barreiro, Cecilia Castro, Víctor Leiva, Xavier Cabezas . Systems of quaternionic linear matrix equations: solution, computation, algorithm, and applications. AIMS Mathematics, 2024, 9(10): 26371-26402. doi: 10.3934/math.20241284 |
[3] | Vladislav N. Kovalnogov, Ruslan V. Fedorov, Denis A. Demidov, Malyoshina A. Malyoshina, Theodore E. Simos, Spyridon D. Mourtas, Vasilios N. Katsikis . Computing quaternion matrix pseudoinverse with zeroing neural networks. AIMS Mathematics, 2023, 8(10): 22875-22895. doi: 10.3934/math.20231164 |
[4] | Wenxv Ding, Ying Li, Anli Wei, Zhihong Liu . Solving reduced biquaternion matrices equation k∑i=1AiXBi=C with special structure based on semi-tensor product of matrices. AIMS Mathematics, 2022, 7(3): 3258-3276. doi: 10.3934/math.2022181 |
[5] | Abdur Rehman, Cecilia Castro, Víctor Leiva, Muhammad Zia Ur Rahman, Carlos Martin-Barreiro . Solving two-sided Sylvester quaternionic matrix equations: Theoretical insights, computational implementation, and practical applications. AIMS Mathematics, 2025, 10(7): 15663-15697. doi: 10.3934/math.2025702 |
[6] | Yang Chen, Kezheng Zuo, Zhimei Fu . New characterizations of the generalized Moore-Penrose inverse of matrices. AIMS Mathematics, 2022, 7(3): 4359-4375. doi: 10.3934/math.2022242 |
[7] | Abdur Rehman, Ivan Kyrchei, Muhammad Zia Ur Rahman, Víctor Leiva, Cecilia Castro . Solvability and algorithm for Sylvester-type quaternion matrix equations with potential applications. AIMS Mathematics, 2024, 9(8): 19967-19996. doi: 10.3934/math.2024974 |
[8] | Anli Wei, Ying Li, Wenxv Ding, Jianli Zhao . Three special kinds of least squares solutions for the quaternion generalized Sylvester matrix equation. AIMS Mathematics, 2022, 7(4): 5029-5048. doi: 10.3934/math.2022280 |
[9] | Dong Wang, Ying Li, Wenxv Ding . The least squares Bisymmetric solution of quaternion matrix equation AXB=C. AIMS Mathematics, 2021, 6(12): 13247-13257. doi: 10.3934/math.2021766 |
[10] | Qi Xiao, Jin Zhong . Characterizations and properties of hyper-dual Moore-Penrose generalized inverse. AIMS Mathematics, 2024, 9(12): 35125-35150. doi: 10.3934/math.20241670 |
In this paper, we establish the following four symmetric quaternion matrix systems:
{A11X1=B11,C11X1D11=E11,X2A22=B22,C22X2D22=E22,F11X1H11+X2F22=G11, | (1.1) |
{A11X1=B11,C11X1D11=E11,X2A22=B22,C22X2D22=E22,F11X1+H11X2F22=G11, | (1.2) |
{A11X1=B11,C11X1D11=E11,A22X2=B22,C22X2D22=E22,F11X1+H11X2F22=G11, | (1.3) |
{A11X1=B11,C11X1D11=E11,A22X2=B22,C22X2D22=E22,F11X1+X2F22=G11, | (1.4) |
where Aii, Bii, Cii, Dii, Eii, Fii(i=¯1,2), H11, and G11 are known matrices, while Xi(i=¯1,2) are unknown.
In this paper, R and Hm×n denote the real number field and the set of all quaternion matrices of order m×n, respectively.
H={v0+v1i+v2j+v3k|i2=j2=k2=ijk=−1,v0,v1,v2,v3∈R}. |
Moreover, r(A), 0 and I represent the rank of matrix A, the zero matrix of suitable size, and the identity matrix of suitable size, respectively. The conjugate transpose of A is A∗. For any matrix A, if there exists a unique solution X such that
AXA=A,XAX=X,(AX)∗=AX,(XA)∗=XA, |
then X is called the Moore-Penrose (M−P) inverse. It should be noted that A† is used to represent the M−P inverse of A. Additionally, LA=I−A†A and RA=I−AA† denote two projectors toward A.
H is known to be an associative noncommutative division algebra over R with extensive applications in computer science, orbital mechanics, signal and color image processing, control theory, and so on (see [1,2,3,4]).
Matrix equations, significant in the domains of descriptor systems control theory [5], nerve networks [6], back feed [7], and graph theory [8], are one of the key research topics in mathematics.
The study of matrix equations in H has garnered the attention of various researchers; consequently they have been analyzed by many studies (see, e.g., [9,10,11,12]). Among these the system of symmetric matrix equations is a crucial research object. For instance, Mahmoud and Wang [13] established some necessary and sufficient conditions for the three symmetric matrix systems in terms of M−P inverses and rank equalities:
{A1V=C1, VB1=C2,A3X+YB3=C3,A2Y+ZB2+A5VB5=C5,A4W+ZB4=C4,{A1V=C1, VB1=C2,A3X+YB3=C3,A2Z+YB2+A5VB5=C5,A4Z+WB4=C4,{A1V=C1, VB1=C2,A3X+YB3=C3,A2Y+ZB2+A5VB5=C5,A4Z+WB4=C4. | (1.5) |
Wang and He [14] established the sufficient and necessary conditions for the existence of solutions to the following three symmetric coupled matrix equations and the expressions for their general solutions:
{A1X+YB1=C1,A2Y+ZB2=C2,A3W+ZB3=C3,{A1X+YB1=C1,A2Z+YB2=C2,A3Z+WB3=C3,{A1X+YB1=C1,A2Y+ZB2=C2,A3Z+WB3=C3. | (1.6) |
It is noteworthy that the following matrix equation plays an important role in the analysis of the solvability conditions of systems (1.1)–(1.4):
A1U+VB1+A2XB2+A3YB3+A4ZB4=B. | (1.7) |
Liu et al. [15] derived some necessary and sufficient conditions to solve the quaternion matrix equation (1.7) using the ranks of coefficient matrices and M−P inverses. Wang et al. [16] derived the following quaternion equations after obtaining some solvability conditions for the quaternion equation presented in Eq (1.8) in terms of M−P inverses:
{A11X1=B11,C11X1D11=E11,X2A22=B22,C22X2D22=E22,F11X1+X2F22=G11. | (1.8) |
To our knowledge, so far, there has been little information on the solvability conditions and an expression of the general solution to systems (1.1)–(1.4).
In mathematical research and applications, the concept of η-Hermitian matrices has gained significant attention [17]. An η-Hermitian matrix, for η∈{i,j,k}, is defined as a matrix A such that A=Aη∗, where Aη∗=−ηA∗η. These matrices have found applications in various fields including linear modeling and the statistics of random signals [1,17]. As an application of (1.1), this paper establishes some necessary and sufficient conditions for the following matrix equation:
{A11X1=B11,C11X1Cη∗11=E11,F11X1Fη∗11+(F22X1)η∗=G11 | (1.9) |
to be solvable.
Motivated by the study of Systems (1.8), symmetric matrix equations, η-Hermitian matrices, and the widespread use of matrix equations and quaternions as well as the need for their theoretical advancements, we examine the solvability conditions of the quaternion systems presented in systems (1.1)–(1.4) by utilizing the rank equalities and the M−P inverses of coefficient matrices. We then obtain the general solutions for the solvable quaternion equations in systems (1.1)–(1.4). As an application of (1.1), we utilize the M−P inverse and the rank equality of matrices to investigate the necessary and sufficient conditions for the solvability of quaternion matrix equations involving η-Hermicity matrices. It is evident that System (1.8) is a specific instance of System (1.1).
The remainder of this article is structured as follows. Section 2 outlines the basics. Section 3 examines some solvability conditions of the quaternion equation presented in System (1.1) using the M−P inverses and rank equalities of the matrices, and derives the solution of System (1.1). Section 4 establishes some solvability conditions for matrix systems (1.2)–(1.4) to be solvable. Section 5 investigates some necessary and sufficient conditions for matrix equation (1.9) to have common solutions. Section 6 concludes the paper.
Marsaglia and Styan [18] presented the following rank equality lemma over the complex field, which can be generalized to H.
Lemma 2.1. [18] Let A∈Hm×n, B∈Hm×k, C∈Hl×n, D∈Hj×k, and E∈Hl×i be given. Then, the following rank equality holds:
r(ABLDREC0)=r(AB0C0E0D0)−r(D)−r(E). |
Lemma 2.2. [19] Let A∈Hm×n be given. Then,
(1)(Aη)†=(A†)η,(Aη∗)†=(A†)η∗;(2)r(A)=r(Aη∗)=r(Aη);(3)(LA)η∗=−η(LA)η=(LA)η=LAη∗=RAη∗,(4)(RA)η∗=−η(RA)η=(RA)η=RAη∗=LAη∗;(5)(AA†)η∗=(A†)η∗Aη∗=(A†A)η=Aη(A†)η;(6)(A†A)η∗=Aη∗(A†)η∗=(AA†)η=(A†)ηAη. |
Lemma 2.3. [20] Let A1 and A2 be given quaternion matrices with adequate shapes. Then, the equation A1X=A2 is solvable if, and only if, A2=A1A†1A2. In this case, the general solution to this equation can be expressed as
X=A†1A2+LA1U1, |
where U1 is any matrix with appropriate size.
Lemma 2.4. [20] Let A1 and A2 be given quaternion matrices with adequate shapes. Then, the equation XA1=A2 is solvable if, and only if, A2=A2A†1A1. In this case, the general solution to this equation can be expressed as
X=A2A†1+U1RA1, |
where U1 is any matrix with appropriate size.
Lemma 2.5. [21] Let A,B, and C be known quaternion matrices with appropriate sizes. Then, the matrix equation
AXB=C |
is consistent if, and only if,
RAC=0,CLB=0. |
In this case, the general solution to this equation can be expressed as
X=A†CB†+LAU+VRB, |
where U and V are any quaternion matrices with appropriate sizes.
Lemma 2.6. [15] Let Ci,Di, and Z(i=¯1,4) be known quaternion matrices with appropriate sizes.
C1X1+X2D1+C2Y1D2+C3Y2D3+C4Y3D4=Z. | (2.1) |
Denote
RC1C2=C12,RC1C3=C13,RC1C4=C14,D2LD1=D21,D31LD21=N32,D3LD1=D31,D4LD1=D41,RC12C13=M23,S12=C13LM23,RC1ZLD1=T1,C32=RM23RC12,A1=C32C14,A2=RC12C14,A3=RC13C14,A4=C14,D13=LD21LN32,B1=D41,B2=D41LD31,B3=D41LD21,B4=D41D13,E1=C32T1,E2=RC12T1LD31,E3=RC13T1LD21,E4=T1D13,A24=(LA2,LA4),B13=(RB1RB3),A11=LA1,B22=RB2,A33=LA3,B44=RB4,E11=RA24A11,E22=RA24A33,E33=B22LB13,E44=B44LB13,N=RE11E22,M=E44LE33,K=K2−K1,E=RA24KLB13,S=E22LN,K11=A2LA1,G1=E2−A2A†1E1B†1B2,K22=A4LA3,G2=E4−A4A†3E3B†3B4,K1=A†1E1B†1+LA1A†2E2B†2,K2=A†3E3B†3+LA3A†4E4B†4. |
Then, the following statements are equivalent:
(1) Equation (2.1) is consistent.
(2)
RAiEi=0,EiLBi=0(i=¯1,4),RE11ELE44=0. |
(3)
r(ZC2C3C4C1D10000)=r(D1)+r(C2,C3,C4,C1),r(ZC2C4C1D3000D1000)=r(C2,C4,C1)+r(D3D1),r(ZC3C4C1D2000D1000)=r(C3,C4,C1)+r(D2D1),r(ZC4C1D200D300D100)=r(D2D3D1)+r(C4,C1),r(ZC2C3C1D4000D1000)=r(C2,C3,C1)+r(D4D1),r(ZC2C1D300D400D100)=r(D3D4D1)+r(C2,C1),r(ZC3C1D200D400D100)=r(D2D4D1)+r(C3,C1),r(ZC1D20D30D40D10)=r(D2D3D4D1)+r(C1),r(ZC2C1000C4D3000000D1000000000−ZC3C1C4000D2000000D1000D400D4000)=r(D30D100D20D1D4D4)+r(C2C100C400C3C1C4). |
Under these conditions, the general solution to the matrix equation (2.1) is
X1=C†1(Z−C2Y1D2−C3Y2D3−C4Y3D4)−C†1U1D1+LC1U2,X2=RC1(Z−C2Y1D2−C3Y2D3−C4Y3D4)D†1+C1C†1U1+U3RD1,Y1=C†12TD†21−C†12C13M†23TD†21−C†12S12C†13TN†32D31D†21−C†12S12U4RN32D31D†21+LC12U5+U6RD21,Y2=M†23TD†31+S†12S12C†13TN†32+LM23LS12U7+U8RD31+LM23U4RN32,Y3=K1+LA2V1+V2RB1+LA1V3RB2, or Y3=K2−LA4W1−W2RB3−LA3W3RB4, |
where T=T1−C4Y3D4,Ui(i=¯1,8) are arbitrary matrices with appropriate sizes over H,
V1=(Im,0)[A†24(K−A11V3B22−A33W3B44)−A†24U11B13+LA24U12],W1=(0,Im)[A†24(K−A11V3B22−A33W3B44)−A†24U11B13+LA24U12],W2=[RA24(K−A11V3B22−A33W3B44)B†13+A24A†24U11+U21RB13](0In),V2=[RA24(K−A11V3B22−A33W3B44)B†13+A24A†24U11+U21RB13](In0),V3=E†11KE†33−E†11E22N†KE†33−E†11SE†22KM†E44E†33−E†11SU31RME44E†33+LE11U32+U33RE33,W3=N†KE44+S†SE†22KM†+LNLSU41+LNU31RM−U42RE44, |
U11,U12,U21,U31,U32,U33,U41, and U42 are arbitrary quaternion matrices with appropriate sizes, and m and n denote the column number of C4 and the row number of D4, respectively.
Some necessary and sufficient conditions for System (1.1) to be solvable will be established in this section. The general solution of System (1.1) will also be derived in this section. Moreover, we provide an example to illustrate our main results.
Theorem 3.1. Let Aii,Bii,Cii,Dii,Eii,Fii,H11, and G11 (i = 1, 2) be given quaternion matrices. Put
{A1=C11LA11,P1=E11−C11A†11B11D11,B2=RA22D22,P2=E22−C22B22A†22D22,^B1=RB2RA22F22,^A2=F11LA11LA1,^A3=F11LA11,^B3=RD11H11,^A4=LC22,^B4=RA22F22,H11L^B1=^B11,P=G11−F11A†11B11H11−F11LA11A†1P1D†11H11−B22A†22F22−C†22P2B†2RA22F22, | (3.1) |
{^B22L^B11=N1,^B3L^B1=^B22,^B4L^B1=^B33,R^A2^A3=^M1,S1=^A3L^M1,T1=PL^B1,C=R^M1R^A2,C1=C^A4,C2=R^A2^A4,C3=R^A3^A4,C4=^A4,D=L^B11LN1,D1=^B33,D2=^B33L^B22,D4=^B33D,E1=CT1,E2=R^A2T1L^B22,E3=R^A3T1L^B11,E4=T1D,^C11=(LC2,LC4),D3=^B33L^B11,^D11=(RD1RD3),^C22=LC1,^D22=RD2,^C33=LC3,^D33=RD4,^E11=R^C11^C22,^E22=R^C11^C22,^E33=^D22L^D11,^E44=^D33L^D11,M=R^E11^E22,N=^E44L^E33,F=F2−F1,E=R^C11FL^D11,S=^E22LM,^F11=C2LC1,G1=E2−C2C†1E1D†1D2,^F22=C4LC3,G2=E4−C4C†3E3D†3D4,F1=C†1E1D†1+LC1C†2E2D†2,F2=C†3E3D†3+LC3C†4E4D†4. | (3.2) |
Then, the following statements are equivalent:
(1) System (1.1) is solvable.
(2)
RA11B11=0,RA1P1=0,P1LD11=0,B22LA22=0,RC22P2=0,P2LB2=0,RCiEi=0,EiLDi=0(i=¯1,4),R^E11EL^E44=0. |
(3)
r(B11,A11)=r(A11),r(E11C11B11D11A11)=r(C11A11), | (3.3) |
r(E11D11)=r(D11),r(B22A22)=r(A22), | (3.4) |
r(E22,C22)=r(C22),r(E22C22B22D22A22)=r(D22,A22), | (3.5) |
r(F220D22A22B11H11A1100C22G11C22F11E22C22B22)=r(F22,D22,A22)+r(A11C22F11), | (3.6) |
r(H110−D1100F2200D22A220C11E11000A11B11D1100C22G11C22F110E22C22B22)=r(C11A11C22F11)+r(H11D1100F220D22A22), | (3.7) |
r(H11000F220D22A220A1100C22G11C22F11E22C22B22)=r(H1100F22D22A22)+r(A11C22F11), | (3.8) |
r(H1100F22D22A22C22G11E22C22B22)=r(H1100F22D22A22), | (3.9) |
r(G11F11B22F220A22B11H11A110)=r(F11A11)+r(F22,A22), | (3.10) |
r(G11F110B22H110−D110F2200A220C11E1100A11B11D110)=r(H11D110F220A22)+r(F11C11A11), | (3.11) |
r(G11F11B22H1100F220A220A110)=r(H110F22A22)+r(F11A11), | (3.12) |
r(G11B22H110F22A22)=r(H110F22A22), | (3.13) |
r(H11000000D110F220000D22A220000H1100000000F22D22A220000F220F2200000A220C1100000−E1100A1100000−B11D110C22G11C22F11000E22C22B2200)=r(H1100000D110F22000D22A22000H110000000F22D22A220000F22F2200000A22)+r(C11A11C22F11). | (3.14) |
Proof. (1)⇔(2): The System (1.1) can be written as follows.
A11X1=B11, X2A22=B22, | (3.15) |
C11X1D11=E11, C22X2D22=E22, | (3.16) |
and
F11X1H11+X2F22=G11. | (3.17) |
Next, the solvability conditions and the expression for the general solutions of Eq (3.15) to Eq (3.17) are given by the following steps:
Step 1: According to Lemma 2.3 and Lemma 2.4, the system (3.15) is solvable if, and only if,
RA11B11=0, B22LA22=0. | (3.18) |
When condition (3.18) holds, the general solution of System (3.15) is
X1=A†11B11+LA11U1, X2=B22A†22+U2RA22. | (3.19) |
Step 2: Substituting (3.19) into (3.16) yields,
A1U1D11=P1, C22U2B2=P2, | (3.20) |
where A1,P1,B2,P2 are defined by (3.1). By Lemma 2.5, the system (3.20) is consistent if, and only if,
RA1P1=0, P1LD11=0, RC22P2=0, P2LB2=0. | (3.21) |
When (3.21) holds, the general solution to System (3.20) is
U1=A†1P1D†11+LA1W1+W2RD11,U2=C†22P2B†2+LC22W3+W4RB2. | (3.22) |
Comparing (3.22) and (3.19), hence,
X1=A†11B11+LA11A†1P1D†11+LA11LA1W1+LA11W2RD11,X2=B22A†22+C†22P2B†2RA22+LC22W3RA22+W4RB2RA22. | (3.23) |
Step 3: Substituting (3.23) into (3.17) yields
W4^B1+^A2W1H11+^A3W2^B3+^A4W3^B4=P, | (3.24) |
where ^Bi,^Aj(i=¯1,4,j=¯2,4) are defined by (3.1). It follows from Lemma 2.6 that Eq (3.24) is solvable if, and only if,
RCiEi=0,EiLDi=0(i=¯1,4),R^E11EL^E44=0. | (3.25) |
When (3.25) holds, the general solution to matrix equation (3.24) is
W1=^A2†T^B11†−^A2†^A3^M1†T^B11†−^A2†S1^A3†TN†1^B22^B11†−^A2†S1V4RN1^B22^B11†+L^A2V5+V6R^B11,W2=^M1†T^B22†+S†1S1^A3†TN†1+L^M1LS1V7+V8R^B22+L^M1V4RN1,W3=F1+LC2^V1+^V2RD1+LC1^V3RD2, or W3=F2−LC4V1−V2RD3−LC3V3RD4,W4=(P−^A2W1H11−^A3W2^B3−^A4W3^B4)^B1†+V3R^B1, |
where Ci,Ei,Di(i=¯1,4),^E11,^E44 are defined as (3.2), T=T1−^A4W3^B4,Vi(i=¯1,8) are arbitrary matrices with appropriate sizes over H,
^V1=(Im,0)[^C11†(F−^C22V3^D22−^C33^V3^D33)−^C11†U11^D11+L^C11U12],V1=(0,Im)[^C11†(F−^C22V3^D22−^C33^V3^D33)−^C11†U11^D11+L^C11U12],V2=[R^C11(F−^C22V3^D22−^C33^V3^D33)^D11†+^C11^C11†U11+U21R^D11](0In),^V2=[R^C11(F−^C22V3^D22−^C33^V3^D33)^D11†+^C11^C11†U11+U21R^D11](In0),^V3=^E11†F^E33†−^E11†^E22M†F^E33†−^E11†S^E22†FN†^E44^E33†−^E11†SU31RN^E44^E33†+L^E11U32+U33R^E33,V3=M†F^E44†+S†S^E22†FN†+LMLSU41+LMU31RN−U42R^E44, |
U11,U12,U21,U31,U32,U33,U41, and U42 are any quaternion matrices with appropriate sizes, and m and n denote the column number of C22 and the row number of A22, respectively. We summarize that System (1.1) has a solution if, and only if, (3.18), (3.21), and (3.25) hold, i.e., the System (1.1) has a solution if, and only if, (2) holds.
(2)⇔(3): We prove the equivalence in two parts. In the first part, we want to show that (3.18) and (3.21) are equivalent to (3.3) to (3.5), respectively. In the second part, we want to show that (3.25) is equivalent to (3.6) to (3.14). It is easy to know that there exist X01,X02,U01, and U02 such that
A11X01=B11, X02A22=B22,A1U01D11=P1, C22U02B2=P2 |
holds, where
X01=A†11B11,U01=A†1P1D†11,X02=B22A†22,U02=C†22P2B†2, |
P1=E11−C11X01D11,P2=E22−C22X02D22, and P=G11−F11X01H11−F11LA11U01H11−X02F22−U02RA22F22.
Part 1: We want to show that (3.18) and (3.21) are equivalent to (3.3) to (3.5), respectively. It follows from Lemma 2.1 and elementary transformations that
(3.18)⇔r(RA11B11)=0⇔r(B11,A11)=r(A11)⇔(3.3),(3.21)⇔r(RA1P1)=0⇔r(P1,A1)=r(A1)⇔r(E11−C11A†11B11D11,C11LA11)=r(C11LA11)⇔r(E11C11B11D11A11)=r(C11A11)⇔(3.3),(3.21)⇔r(P1LD11)=0⇔r(P1D11)=r(D11)⇔r(E11−C11A†11B11D11D11)=r(D11)⇔r(E11D11)=r(D11)⇔(3.4),(3.18)⇔r(B22LA22)=0⇔r(B22A22)=r(A22)⇔(3.4). |
Similarly, we can show that (3.21) is equivalent to (3.5). Hence, (3.18) and (3.21) are equivalent to (3.3) and (3.5), respectively.
Part 2: In this part, we want to show that (3.25) is equivalent to (3.6) and (3.14). According to Lemma 2.6, we have that (3.25) is equivalent to the following:
r(P^A2^A3^A4^B1000)=r(^B1)+r(^A2,^A3,^A4), | (3.26) |
r(P^A2^A4^B300^B100)=r(^A2,^A4)+r(^B3^B1), | (3.27) |
r(P^A3^A4H1100^B100)=r(^A3,^A4)+r(H11^B1), | (3.28) |
r(P^A4H110^B30^B10)=r(H11^B3^B1)+r(^A4), | (3.29) |
r(P^A2^A3^B400^B100)=r(^A2,^A3)+r(^B4^B1), | (3.30) |
r(P^A2^B30^B40^B10)=r(^B3^B4^B1)+r(^A2), | (3.31) |
r(P^A3H110^B40^B10)=r(H11^B4^B1)+r(^A3), | (3.32) |
r(PH11^B3^B4^B1)=r(H11^B3^B4^B1), | (3.33) |
r(P^A200^A4^B30000^B1000000−P^A3^A400H110000^B100^B40^B400)=r(^B30^B100H110^B1^B4^B4)+r(^A20^A40^A3^A4), | (3.34) |
respectively. Hence, we only prove that (3.26)–(3.34) are equivalent to (3.6)–(3.14) when we prove that (3.25) is equivalent to (3.6)–(3.14). Now, we prove that (3.26)–(3.34) are equivalent to (3.6)–(3.14). In fact, we only prove that (3.26), (3.30), and (3.34) are equivalent to (3.6), (3.10), and (3.14); the remaining part can be proved similarly. According to Lemma 2.1 and elementary transformations, we have that
(3.26)=r(P^A2^A3^A4^B1000)=r(^B1)+r(^A2,^A3,^A4)⇔r(G11−F11X01H11−F11LA11U01H11−X02F22−U02RA22F22F11LA11LA1F11LA11LC22RB2RA22F22000)=r(RB2RA22F22)+r(F11LA11LA1,F11LA11,LC22)⇔r(G11−F11X01H11−X02F22−U02RA22F22F11I0RA22F2200B20A110000C220)=r(RA22F22,B2)+r(F11IA1100C22)⇔r(G11F11IU02B20F2200B2A22B11H11A11000C22X02F220C2200)=r(F22,D22,A22)+r(F11IA1100C22)⇔r(F220D22A22B11H11A1100C22G11C22F11E22C22B22)=r(F22,D22,A22)+r(A11F11C22)⇔(3.6). |
Similarly, we have that (3.27)⇔(3.7),(3.28)⇔(3.8),(3.29)⇔(3.9).
(3.30)=r(P^A2^A3^B400^B100)=r(^A2,^A3)+r(^B4^B1)⇔r(G11−F11X01H11−F11LA11U01H11−X02F22−U02RA22F22F11LA11LA1F11LA1RA22F2200RB2RA22F2200)=r(F11LA11LA1,F11LA11)+r(RA22F22RB2RA22F22)⇔r(G11−F11X01H11F11B22F220A220A110)=r(F11A11)+r(F22,A22)⇔r(G11F11B22F220A22B11H11A110)=r(F11A11)+r(F22,A22)⇔(3.10). |
Similarly, we have that (3.31)⇔(3.11),(3.32)⇔(3.12),(3.33)⇔(3.13).
(3.34)=r(P^A200^A4^B30000^B1000000−P^A3^A400H110000^B100^B40^B400)=r(^B30^B100H110^B1^B4^B4)+r(^A20^A40^A3^A4)⇔r(PF11LA11LA100LC22RD11H110000RB2RA22F22000000−PF11LA11LC2200H110000RB2RA22F2200RA22F220RA22F2200)=r(RD11H110RB2RA22F2200H110RB2RA22F22RA22F22RA22F22)+r(F11LA11LA10LC220F11LA11LC22)⇔r(PF11LA1100LC22000H110000D1100RA22F2200000B2000−G11+X02F22+U02RA22F22F11LA11LC2200000H110000000RA22F220000B2RA22F220RA22F22000000A1000000)=r(H110D1100RA2200B200H110000RA22F2200B2RA22F22RA22F22000)+r(F11LA110LC220F11LA11LC22A100)⇔r(H11000000D110F220000D22A220000H1100000000F22D22A220000F220F2200000A220C1100000−E1100A1100000−B11D110C22G11C22F11000E22C22B2200)=r(H1100000D110F22000D22A22000H110000000F22D22A220000F22F2200000A22)+r(C11A11C22F11)⇔(3.14). |
Theorem 3.2. Let System (1.1) be solvable. Then, the general solution of System (1.1) is
X1=A†11B11+LA11A†1P1D†11+LA11LA1W1+LA11W2RD11,X2=B22A†22+C†22P2B†2RA22+LC22W3RA22+W4RB2RA22, |
where
W1=^A2†T^B11†−^A2†^A3^M1†T^B11†−^A2†S1^A3†TN†1^B22^B11†−^A2†S1V4RN1^B22^B11†+L^A2V5+V6R^B11,W2=^M1†T^B22†+S†1S1^A3†TN†1+L^M1LS1V7+V8R^B22+L^M1V4RN1,W3=F1+LC2^V1+^V2RD1+LC1^V3RD2, or W3=F2−LC4V1−V2RD3−LC3V3RD4,W4=(P−^A2W1H11−^A3W2^B3−^A4W3^B4)^B1†+V3R^B1,^V1=(Im,0)[^C11†(F−^C22V3^D22−^C33^V3^D33)−^C11†U11^D11+L^C11U12],V1=(0,Im)[^C11†(F−^C22V3^D22−^C33^V3^D33)−^C11†U11^D11+L^C11U12],V2=[R^C11(F−^C22V3^D22−^C33^V3^D33)^D11†+^C11^C11†U11+U21R^D11](0In),^V2=[R^C11(F−^C22V3^D22−^C33^V3^D33)^D11†+^C11^C11†U11+U21R^D11](In0),^V3=^E11†F^E33†−^E11†^E22M†F^E33†−^E11†S^E22†FN†^E44^E33†−^E11†SU31RN^E44^E33†+L^E11U32+U33R^E33,V3=M†F^E44†+S†S^E22†FN†+LMLSU41+LMU31RN−U42R^E44, |
T=T1−^A4W3^B4,Vi(i=¯4,8) are arbitrary matrices with appropriate sizes over H, U11,U12,U21, U31,U32,U33,U41, and U42 are any quaternion matrices with appropriate sizes, and m and n denote the column number of C22 and the row number of A22, respectively.
Next, we consider a special case of the System (1.1).
Corollary 3.3. [16] Let Aii,Bii,Cii,Dii,Eii,Fii(i=1,2), and G11 be given matrices with appropriate dimensions over H. Denote
T=C11LA11,K=RA22D22, B1=RKRA22F22,A1=F11LA11LT,C3=F11LA11,D3=RD11,C4=LC22,D4=RA22F22,Aα=RA1C3,Bβ=D3LB1,Cc=RAαC4,Dd=D4LB1,E=RA1E1LB1,A=A†11B11+LA11T†(E11−C11A†11B11D11)D†,B=B22A†22+C†22(E22−C22B22A†22D22)K†RA22,E1=G11−F11A−BF22,M=RAαCc,N=DdLBβ,S=CcLM. |
Then, the following statements are equivalent:
(1) Equation (1.8) is consistent.
(2)
RA11B11=0,B22LA22=0,RC22E22=0,E11LD11=0,RT(E11−C11A†11B11D11)=0,(E22−C22B22A†22D22)LK=0,RMRAαE=0,ELBβLN=0,RAαELDd=0,RCcELBβ=0. |
(3)
r(B11,A11)=r(A11),r(E11C11B11D11A11)=r(C11A11),r(E11D11)=r(D11),r(B22A22)=r(A22),r(E22,C22)=r(C22),r(E22C22B22D22A22)=r(D22,A22),r(F220D22A22B11A1100C22G11C22F11E22C22B22)=r(F22,D22,A22)+r(A11C22F11),r(0F22D11D22A22C11E1100A11B11D1100C22F11C22G11D11E22C22B22)=r(C11A11C22F11)+r(F22D11,D22,A22),r(G11F11B22F220A22B11A110)=r(F11A11)+r(F22,A22),r(F11G11D11B220F22D11A22C11E110A11B11D110)=r(F22D11,A22)+r(F11C11A11). |
Finally, we provide an example to illustrate the main results of this paper.
Example 3.4. Conside the matrix equation (1.1)
A11=(a111a121),B11=(b111b112b121b122),C11=(c111c121),D11=(d111d121),E11=(e111e121),A22=(a211a212),B22=(b211b212b221b222),C22=(c211c212c221c222),D22=(d211),E22=(e211e221),F11=(f111f121),H11=(h111h112h121h122),F22=(f211f212),G11=(g111g112g121g122), |
where
a111=0.9787+0.5005i+0.0596j+0.0424k,a121=0.7127+0.4711i+0.6820j+0.0714k,b111=0.5216+0.8181i+0.7224j+0.6596k,b112=0.9730+0.8003i+0.4324j+0.0835k,b121=0.0967+0.8175i+0.1499j+0.5186k,b122=0.6490+0.4538i0.8253j+0.1332k,c111=0.1734+0.8314i+0.0605j+0.5269k,c121=0.3909+0.8034i+0.3993j+0.4168k,d111=0.6569+0.2920i+0.0159j+0.1671k,d121=0.6280+0.4317i+0.9841j+0.1062k,e111=0.3724+0.4897i+0.9516j+0.0527k,e121=0.1981+0.3395i+0.9203j+0.7379k,a211=0.2691+0.4228i+0.5479j+0.9427k,a212=0.4177+0.9831i+0.3015j+0.7011k,b211=0.6663+0.6981i+0.1781j+0.9991k,b212=0.0326+0.8819i+0.1904j+0.4607k,b221=0.5391+0.6665i+0.1280j+0.1711k,b222=0.5612+0.6692i+0.3689j+0.9816k,c211=0.1564+0.6448i+0.1909j+0.4820k,c212=0.5895+0.3846i+0.2518j+0.6171k,c221=0.8555+0.3763i+0.4283j+0.1206k,c222=0.2262+0.5830i+0.2904j+0.2653k,d211=0.8244+0.9827i+0.7302j+0.3439k,e211=0.5847+0.9063i+0.8178j+0.5944k,e221=0.1078+0.8797i+0.2607j+0.0225k,f111=0.4253+0.1615i+0.4229j+0.5985k,f121=0.3127+0.1788i+0.0942j+0.4709k,h111=0.6959+0.6385i+0.0688j+0.5309k,h112=0.4076+0.7184i+0.5313j+0.1056k,h121=0.6999+0.0336i+0.3196j+0.6544k,h122=0.8200+0.9686i+0.3251j+0.6110k,f211=0.7788+0.4235i+0.0908j+0.2665k,f212=0.1537+0.2810i+0.4401j+0.5271k,g111=0.4574+0.5181i+0.6377j+0.2407k,g112=0.2891+0.6951i+0.2548j+0.6678k,g121=0.8754+0.9436i+0.9577j+0.6761k,g122=0.6718+0.0680i+0.2240j+0.8444k. |
Computing directly yields the following:
r(B11A11)=r(A11)=2,r(E11C11B11D11A11)=r(C11A11)=2,r(E11D11)=r(D11)=1,r(B22A22)=r(A22)=2,r(E22C22)=r(C22)=2,r(E22C22B22D22A22)=r(D22A22)=3,r(F220D22A22B11H11A1100C22G11C22F11E22C22B22)=r(F22D22A22)+r(A11C22F11)=5,r(H110−D1100F2200D22A220C11E11000A11B11D1100C22G11C22F110E22C22B22)=r(C11A11C22F11)+r(H11D1100F220D22A22)=7,r(H11000F220D22A220A1100C22G11C22F11E22C22B22)=r(H1100F22D22A22)+r(A11C22F11)=6,r(H1100F22D22A22C22G11E22C22B22)=r(H1100F22D22A22)=5,r(G11F11B22F220A22B11H11A110)=r(F11A11)+r(F22,A22)=5,r(G11F110B22H110−D110F2200A220C11E1100A11B11D110)=r(H11D110F220A22)+r(F11C11A11)=6,r(G11F11B22H1100F220A220A110)=r(H110F22A22)+r(F11A11)=5, r(G11B22H110F22A22)=r(H110F22A22)=4,r(H11000000D110F220000D22A220000H1100000000F22D22A220000F220F2200000A220C1100000−E1100A1100000−B11D110C22G11C22F11000E22C22B2200)=r(H1100000D110F22000D22A22000H110000000F22D22A220000F22F2200000A22)+r(C11A11C22F11)=11. |
All rank equations in (3.3) to (3.14) hold. So, according to Theorem 3.1, the system of matrix equation (1.1) has a solution. By Theorem 3.2, the solution of System (1.1) can be expressed as
X1=(0.4946+0.1700i−0.1182j−0.3692k0.4051−0.0631i−0.2403j−0.1875k),X2=(−0.0122+0.2540i−0.3398j−0.3918k0.7002−0.3481i−0.2169j+0.0079k). |
In this section, we use the same method and technique as in Theorem 3.1 to study the three systems of Eqs (1.2)–(1.4). We only present their results and omit their proof.
Theorem 4.1. Consider the matrix equation (1.2) over H, where Aii,Bii,Cii,Dii,Eii,Fii,G11, and H11(i=¯1,2) are given. Put
A1=C11LA11,P1=E11−C11A†11B11D11,B2=RA22D22,P2=E22−C22B22A†22D22,^A1=F11LA11LA1,^A2=F11LA1,^B2=RD11,^A3=H11LC22,^B3=RA22F22,^B4=RB2RA22F22,B=G11−F11A†11B11−F11LA11A†1P1D†11−H11B22A†22F22−H11C†22P2B†2RA22F22,R^A1^A2=A12,R^A1^A3=A13,R^A1H11=A14,^B3L^B2=N1,RA12A13=M1,S1=A13LM1,R^A1B=T1,C=RM1RA12,^C1=CA14,^C2=RA12A14,^C3=RA13A14,^C4=A14,D=L^B2LN1,^D1=^B4,^D2=^B4L^B3,^D3=^B4L^B2,^D4=^B4D,^E1=CT1,^E2=RA12T1L^B3,^E3=RA13T1L^B2,^E4=T1D,C24=(L^C2,L^C4),D13=(R^D1R^D3),C12=L^C1,D12=R^D2,C33=L^C3,D33=R^D4,E24=RC24C12,E13=RC24C33,E33=D12LD13,E44=D33LD13,M=RE24E13,N=E44LE33,F=F2−F1,E=RC24FLD13,S=E13LM,^F11=^C2L^C1,^G1=^E2−^C2^C1†^E1^D1†^D2,F33=^C4L^C3,^G2=^E4−^C4^C3†^E3^D3†^D4,F1=^C1†^E1^D1†+L^C1^C2†^E2^D2†,F2=^C3†^E3^D3†+L^C3^C4†^E4^D4†. |
Then, the following statements are equivalent:
(1) System (1.2) is consistent.
(2)
RA11B11=0,RA1P1=0,P1LD11=0,B22LA22=0,RC22P2=0,P2LB2=0,R^Ci^Ei=0,^EiL^Di=0(i=¯1,4),RE24ELE44=0. |
(3)
r(B11,A11)=r(A11),r(E11C11B11D11A11)=r(C11A11),r(E11D11)=r(D11),r(B22,A22)=r(A22),r(E22,C22)=r(C22),r(E22C22B22D22A22)=r(D22,A22),r(G11D11F11H11E11C110B11D11A110)=r(F11H11C110A110),r(G11D11F11H110F22D1100A22E11C1100B11D11A1100)=r(F22,A22)+r(F11H11C110A110),r(H11F11G11D110C11E110A11B11D11)=r(H11F110C110A11),r(H11F110G11D1100A22F22D110C110E110A110B11D11)=r(F22D11,A22)+r(H11F110C110A11),r(G11D11F11H1100F22D1100D22A22E11C1100000C22−E22−C22B22B11D11A11000)=r(F11H11C1100C22A110)+r(F22,D22,A22),r(G11D11F11H11B22F22D110A22E11C110B11D11A110)=r(F11C11A11)+r(F22,A22),r(H11F1100G11D1100D22A22F22D110C1100E110A1100B11D11C220−E22−C22B220)=r(H11F110C110A11C220)+r(D22,A22,F22D11),r(F11H11B22G11D110A22F22D11C110E11A110B11D11)=r(F11C11A11)+r(A22,F22D11),r(G11F1100H1100H5B220F22000000A22000H11F11H110−H11B220G11D1100000D22A220−F22D1100C2200E22000000C110000E11000A110000B11D11B11A110000000)=r(F2200A2200D22A220F22D11)+r(F1100H110H11F11H110C220000C11000A110A11000). |
Under these conditions, the general solution of System (1.2) is
X1=A†11B11+LA11A†1P1D†11+LA11LA1W1+LA11W2RD11,X2=B22A†22+C†22P2B†2RA22+LC22W3RA22+W4RB2RA22, |
where
W1=^A1†(B−^A2W1^B2−^A3W3^B3−H11W4^B4)+L^A1U1,W2=A†12T^B2†−A†12A13M†1T^B2†−A†12S1A†13TN†1^B3^B2†−A†12S1U2RN1^B3^B2†+LA12U3+U4R^B2,W3=M†1T^B3†+S†1S1A†13TN†1+LM1LS1U5+U6R^B3+LM1U2RN1,W4=F1+L^C2V1+V2R^D1+L^C1V3R^D2, or W4=F2−L^C4^V1−^V2R^D3−L^C3^V3R^D4, |
where T=T1−H11W4^B4,Ui(i=¯1,6) are arbitrary matrices with appropriate sizes over H,
V1=(Im,0)[C†24(F−C12V3D12−C33^V3D33)−C†24U11D13+LC24U12],^V1=(0,Im)[C†24(F−C12V3D12−C33^V3D33)−C†24U11D13+LC24U12],^V2=[RC24(F−C12V3D12−C33^V3D33)D†13+C24C†24U11+U21RD13](0In),V2=[RC24(F−C12V3D12−C33^V3D33)D†13+C24C†24U11+U21RD13](In0),V3=E†24FE†33−E†24E13M†FE†33−E†24SE†13FN†E44E†33−E†24SU31RNE44E†33+LE24U32+U33RE33,^V3=M†FE†44+S†SE†13FN†+LMLSU41+LMU31RN−U42RE44, |
U11,U12,U21,U31,U32,U33,U41, and U42 are any quaternion matrices with appropriate sizes, and m and n denote the column number of H11 and the row number of A22, respectively.
Theorem 4.2. Consider the matrix equation (1.3) over H, where Aii,Bii,Cii,Dii,Eii,Fii,G11H11(i=¯1,2) are given. Put
A1=C11LA11,P1=E11−C11A†11B11D11,A2=C22LA22,P2=E22−C22A†22B22D22,^A1=F11LA11LA1,^A2=F11LA11,^B2=RD11,^A11=H11LA22LA2,^A22=H11LA22,^B4=RD22F22,B=G11−F11A†11B11−F11LA11A†1P1D†11−H11A†22B22F22−H11LA22A†2P2D†22F22,R^A1^A2=A12,R^A1^A11=A13,R^A1^A22=A33,F22L^B2=N1,RA12A13=M1,S1=A13LM1,R^A1B=T1,C=RM1RA12,^C1=CA33,^C2=RA12A33,^C11=RA13A33,^C22=A33,D=L^B2LN1,^D1=^B4,^D2=^B4LF22,^D11=^B4L^B2,^D22=^B4D,^E1=CT1,^E2=RA12T1LF22,^E11=RA13T1L^B2,^E4=T1D,C24=(L^C2,L^C22),D13=(R^D1R^D11),C21=L^C1,D12=R^D2,C33=L^C11,D33=R^D22,E11=RC24C21,E22=RC24C33,E33=D12LD13,E44=D33LD13,M=RE11E22,N=E44LE33,F=F2−F1,E=RC24FLD13,S=E22LM,^F11=^C2L^C1,^G1=^E2−^C2^C1†^E1^D1†^D2,^F22=^C22L^C11,^G2=^E4−^C22^C11†^E11^D11†^D22,F1=^C1†^E1^D1†+L^C1^C2†^E2^D2†,F2=^C11†^E11^D11†+L^C11^C22†^E4^D22†. |
Then, the following statements are equivalent:
(1) System (1.3) is consistent.
(2)
RA11B11=0,RA1P1=0,P1LD11=0,RA22B22=0,RA2P2=0,P2LD22=0,R^Ci^Ei=0,R^C11^E11=0,R^C22^E4=0,^EiL^Di=0(i=¯1,2),^E11L^D11=0,^E4L^D22=0,RE11ELE44=0. |
(3)
r(B11,A11)=r(A11),r(E11C11B11D11A11)=r(C11A11),r(E11D11)=r(D11),r(B22,A22)=r(A22),r(E22C22B4D22A22)=r(C22A22),r(E22D22)=r(D22),r(G11F11H11B11A110B22F220A22)=r(F11H11A1100A22),r(G11F11H11F2200B11A11000A22)=r(F22)+r(F11H11A1100A22),r(H11F11G11D11A220B22F22D110C11E110A11B11D11)=r(H11F110C110A11A220),r(H11F11G11D1100F22D110C11E110A11B11D11A2200)=r(H11F110C110A11A220)+r(F22D11),r(G11F11H110F2200D22B11A110000C22−E2200A22−B22D22)=r(F11H11A1100C220A22)+r(F22,D22),r(G11F11F220B11A11)=r(F11A11)+r(F22),r(H11F110G11D1100D22F22D11C220−E2200C110E11A2200B22F22D110A110B11D11)=r(H11F11C2200C11A2200A11)+r(D22,F22D11),r(F11G11D110F22D11C11E11A11B11D11)=r(F11C11A11)+r(F22D11),r(G11F11000H110F2200000000−G11D11H11F11H11000F22D11000B22B11A1100000000C2200E2200−E110C110000−B22F22D11A2200000−B11D110A110000000A220)=r(F22000D22F22D11)+r(F1100H110H11F11H110C22000A220000C11000A110A11000000A22). |
Under these conditions, the general solution of System (1.3) is
X1=A†11B11+LA11A†1P1D†11+LA11LA1W1+LA11W2RD11,X2=A†22B4+LA22A†2P2D†22+LA22LA2W3+LA22W4RD22, |
where
W1=^A1†(B−^A2W1^B2−^A11W3F22−^A22W4^B4)+L^A1U1,W2=A†12T^B2†−A†12A13M†1T^B2†−A†12S1A†13TN†1F22^B2†−A†12S1U2RN1F22^B2†+LA12U3+U4R^B2,W3=M†1TF†22+S†1S1A†13TN†1+LM1LS1U5+U6RF22+LM1U2RN1,W4=F1+L^C2V1+V2R^D1+L^C1V3R^D2, or W4=F2−L^C22^V1−^V2R^D11−L^C11^V3R^D22, |
where T=T1−^A22W4^B4,Ui(i=¯1,6) are arbitrary matrices with appropriate sizes over H,
V1=(Im,0)[C†24(F−C21V3D12−C33^V3D33)−C†24U11D13+LC24U12],^V1=(0,Im)[C†24(F−C21V3D12−C33^V3D33)−C†24U11D13+LC24U12],^V2=[RC24(F−C21V3D12−C33^V3D33)D†13+C24C†24U11+U21RD13](0In),V2=[RC24(F−C21V3D12−C33^V3D33)D†13+C24C†24U11+U21RD13](In0),V3=E†11FE†33−E†11E22M†FE†33−E†11SE†22FN†E44E†33−E†11SU31RNE44E†33+LE11U32+U33RE33,^V3=M†FE†44+S†SE†22FN†+LMLSU41+LMU31RN−U42RE44, |
U11,U12,U21,U31,U32,U33,U41, and U42 are any matrices with appropriate sizes, and m and n denote the column number of H11 and the row number of D22, respectively.
Theorem 4.3. Consider the matrix equation (1.4) over H, where Aii,Bii,Cii,Dii,Eii,Fii(i=¯1,2), and G11 are given. Put
^A1=C11LA11,P1=E11−C11A†11B11D11,^A2=C22LA22,P2=E22−C22A†22B22D22,A5=F11LA1L^A1,A6=F11LA11,A7=LA22L^A2,A8=LA22,B5=RD11,B7=RD22F22,B=G11−F11A†11B11−F11LA1^A1†P1D†11−A†22B22F22−LA22^A2†P2D†22F22,RA5A6=A11,RA5A7=A2,RA5A8=A33,F22LB5=N1,RA11A2=M1,S1=A2LM1,RA5B=T1,C=RM1RA11,^C1=CA33,^C2=RA11A33,^C11=RA2A33,^C4=A33,D=LB5LN1,^D1=B7,^D2=B7LF22,^D3=B7LB5,^D4=B7D,^E1=CT1,^E2=RA11T1LF22,^E11=RA2T1LB5,^E4=T1D,C1=(L^C2,L^C4),D13=(R^D1R^D3),D1=L^C1,D2=R^D2,C33=L^C11,D33=R^D4,E11=RC1D1,E2=RC1C33,E33=D2LD13,E44=D33LD13,M=RE11E2,N=E44LE33,F=^F2−^F1,E=RC1FLD13,S=E2LM,F11=^C2L^C1,^G1=^E2−^C2^C1†^E1^D1†^D2,F33=^C4L^C11,^G2=^E4−^C4^C11†^E11^D3†^D4,^F1=^C1†^E1^D1†+L^C1^C2†^E2^D2†,^F2=^C11†^E11^D3†+L^C11^C4†^E4^D4†. |
Then, the following statements are equivalent:
(1) Equation (1.4) is consistent.
(2)
RA11B11=0,R^A1P1=0,P1LD11=0,RA22B22=0,R^A2P2=0,P2LD22=0, R^Ci^Ei=0,^EiL^Di=0(i=¯1,2),R^C11^E11=0,R^C4^E4=0,^E11L^D3=0,^E4L^D4=0,RE11ELE44=0. |
(3)
r(B11,A11)=r(A11),r(E11C11B11D11A11)=r(C11A11),r(E11D11)=r(D11), r(B22,A22)=r(A22),r(E22C22B22D22A22)=r(C22A22),r(E22D22)=r(D22),r(B11A11A22G11−B22F22A22F11)=r(A11A22F11),r(F220B11A11A22G11A22F11)=r(F22)+r(A11A22F11),r(C11E11A11B11D11−A22F11B22F22D11−A22G11D11)=r(C11A11A22F11),r(0F22D11C11E11A11B11D11A22F11A22G11D11)=r(C11A11A22F11)+r(F22D11),r(F220D22C22G11C22F11E22B11A110A22G11A22F11B22D22)=r(F22,D22)+r(C22F11A22F11A11),r(G11F11F220B11A11)=r(F11A11)+r(F22),r(0D22F22D11C22F11E22C22G11D11C110E22A22F110A22G11D11−B22F22D11A110B11D11)=r(C22F11C11A22F11A11)+r(D22,F22D11),r(F11G11D110F22D11C11E11A11B11D11)=r(F11C11A11)+r(F22D11), |
r(F22000000F22D110B22B11A11000C22G11C22F11C22G11D11−C22F11E2200−E11C110A22G11A22F11A22G11D11−B22F22D11−A22F11000−B11D11A110A22G11A22F11000)=r(F22000F22D11D22)+r(−C22F11C22F11−A22F11A22F110C110A11A110A110−A22F110). |
Under these conditions, the general solution of System (1.4) is
X1=A†11B11+LA1^A1†P1^B1†+LA1L^A1W1+LA1W2R^B1,X2=A†2B22+LA2^A2†P2^B2†+LA2L^A2W3+LA3W4R^B2, |
where
W1=A†5(B−A6W1B5−A7W3F22−A8W4B7)+LA5U1,W2=A†1TB†5−A†1A2M†1TB†5−A†1S1A†2TN†1F22B†5−A†1S1U2RN1F22B†5+LA1U3+U4RB5,W3=M†1TF†22+S†1S1A†2TN†1+LM1LS1U5+U6RF22+LM1U2RN1,W4=^F1+L^C2V1+V2R^D1+L^C1V3R^D2, or W4=^F2−L^C4^V1−^V2R^D3−L^C11^V3R^D4, |
where T=T1−A8W4B7,Ui(i=¯1,6) are arbitrary matrices with appropriate sizes over H,
V1=(Im,0)[C†1(F−D1V3D2−C33^V3D33)−C†1U11D1+LC1U12],^V1=(0,Im)[C†1(F−D1V3D2−C33^V3D33)−C†1U11D1+LC1U12],^V2=[RC1(F−D1V3D2−C33^V3D33)D†1+C1C†1U11+U21RD1](0In),V2=[RC1(F−C2V3D2−C33^V3D33)D†1+C1C†1U11+U21RD1](In0),V3=E†11FE†33−E†11E2M†FE†33−E†11SE†2FN†E44E†33−E†11SU31RNE44E†33+LE11U32+U33RE33,^V3=M†FE†44+S†SE†2FN†+LMLSU41+LMU31RN−U42RE44, |
U11,U12,U21,U31,U32,U33,U41, and U42 are any quaternion matrices with appropriate sizes, and m and n denote the column number of A22 and the row number of D22, respectively.
In this section, we use the Lemma 2.2 and the Theorem 3.1 to study matrix equation (1.9) involving η-Hermicity matrices.
Theorem 5.1. Let A11,B11,C11,E11,F11,F22, and G11(G11=Gη∗11) be given matrices. Put
A1=C11LA11,P1=E11−C11A†11B11Cη∗11,B2=Aη∗1,P2=Pη∗1,ˆB1=RB2(F22LA11)η∗,ˆA3=F11LA11,ˆA2=ˆA3LA1,ˆA4=LC11,ˆB3=(F11ˆA4)η∗,ˆB4=(F22LA11)η∗,Fη∗11LˆB1=ˆB11,P=G11−F11A†11B11Fη∗11−ˆA3A†1P1(F11C†11)η∗−(F22A†11B11)η∗−C†11P2B†2ˆB4,ˆB22LB11=N1,ˆB3LˆB1=ˆB22,ˆB4LˆB1=ˆB33,RˆA2ˆA3=ˆM1,S1=ˆA3LM1,T1=PL^B1,C=RM1RˆA2,C1=CˆA4,C2=RˆA2ˆA4,C3=RˆA3ˆA4,C4=ˆA4,D=LˆB11LN1,D1=ˆB33,D2=ˆB33LˆB22,D4=ˆB33D,E1=CT1,E2=RˆA2T1LˆB11,E4=T1D,ˆC11=(LC2,LC4),D3=ˆB33LˆB11,ˆD11=(RD1RD3),ˆC22=LC1,ˆD22=RD2, ˆC33=LC3,ˆD33=RD4,ˆE11=RˆC11ˆC22,ˆE22=RˆC11ˆC33,ˆE33=ˆD22LˆD11,ˆE44=ˆD33LˆD11,M=RˆE11ˆE22,N=ˆE44LˆE33, F=F2−F1,E=RˆC11FLˆD11,S=ˆE22LM,^F11=C2LC1,G1=E2−C2C†1E1D†1D2,^F22=C4LC3,G2=E4−C4C†3E3D†3D4,F1=C†1E1D†1+L†C1C†2E2D†2,F2=C†3E3D†3+LC3C†4E4D†4. |
Then, the following statements are equivalent:
(1) System (1.9) is solvable.
(2)
RA11B11=0,RA1P1=0,P1(RC11)η∗=0,RCiEi=0,EiLDi=0(i=¯1,4),RˆE11ELˆE44=0. |
(3)
r(B11,A11)=r(A11),r(E11C11B11Cη∗11A11)=r(C11A11), r(E11Cη∗11)=r(C11),r(Fη∗220Cη∗11Aηη∗11B11Fη∗11A1100C11G11C11F11Eη∗11C11Bη∗11)=r(Fη∗22,Cη∗11,Aη∗11)+r(A11C11F11),r(Fη∗110−Cη∗1100Fη∗2200Cη∗11Aη∗110C11E11000A11B11Cη∗1100C11G11C11F110Eη∗11C11Bη∗11)=r(C11A110)+r(Fη∗11Cη∗1100Fη∗220Cη∗11Aη∗11),r(Fη∗11000Fη∗220Cη∗11Aη∗110A1100C11G11C11F11Eη∗11C11Bη∗11)=r(Fη∗1100Fη∗22Cη∗11Aη∗11)+r(A11C11F11),r(Fη∗1100Fη∗22Cη∗11Aη∗11C11G11Eη∗11C11Bη∗11)=r(Fη∗1100Fη∗22Cη∗11Aη∗11,),r(G11F11Bη∗11Fη∗220Aη∗11B11Fη∗11A110)=r(F11A11)+r(Fη∗22,Aη∗11),r(G11F110Bη∗11Fη∗110−Cη∗110Fη∗2200Aη∗110C11E1100A11B11Cη∗110)=r(Fη∗11Cη∗110Fη∗220Aη∗11)+r(F11C11A11),r(G11F11Bη∗11Fη∗1100Fη∗220Aη∗110A110)=r(Fη∗110Fη∗22Aη∗11)+r(F11A11),r(G11Bη∗11Fη∗110Fη∗22Aη∗11)=r(Fη∗110Fη∗22Aη∗11),r(Fη∗11000000Cη∗110Fη∗220000Cη∗11Aη∗110000Fη∗1100000000Fη∗22Cη∗11Aη∗110000Fη∗220Fη∗2200000Aη∗110C1100000−E1100A1100000−B11Cη∗110C11G11C11F11000Eη∗11C11Bη∗1100)=r(Fη∗1100000Cη∗110Fη∗22000Cη∗11Aη∗11000Fη∗110000000Fη∗22Cη∗11Aη∗110000Fη∗22Fη∗2200000Aη∗11)+r(C11A11C11F11). |
Proof. Evidently, the system of Eq (1.9) has a solution if and only if the following matrix equation has a solution:
A11^X1=B11,C11^X1Cη∗11=E11,^X2Aη∗11=Bη∗11,C11^X2Cη∗11=Eη∗11,F11X1Fη∗11+^X2η∗Fη∗22=G11. | (5.1) |
If (1.9) has a solution, say, X1, then (^X1, ^X2):=(X1, Xη∗1) is a solution of (5.1). Conversely, if (5.1) has a solution, say (^X1, ^X2), then it is easy to show that (1.5) has a solution
X1:=^X1+Xη∗22. |
According to Theorem 3.1, we can deduce that this theorem holds.
We have established the solvability conditions and the expression of the general solutions to some constrained systems (1.1)–(1.4). As an application, we have investigated some necessary and sufficient conditions for Eq (1.9) to be consistent. It should be noted that the results of this paper are valid for the real number field and the complex number field as special number fields.
Long-Sheng Liu, Shuo Zhang and Hai-Xia Chang: Conceptualization, formal analysis, investigation, methodology, software, validation, writing an original draft, writing a review, and editing. All authors of this article have contributed equally. All authors have read and approved the final version of the manuscript for publication.
This work is supported by the National Natural Science Foundation(No. 11601328) and Key scientific research projects of univesities in Anhui province(No. 2023AH050476).
The authors declare that they have no conflicts of interest.
[1] | Kohler W (1938) Gestaltprobleme und Anfänge einer Gestalttheorie. Jahresbericht über der Gesellshaft Physiologie. Translated by: Ellis WD, A Source Book of Gestalt Psychology, London: Routledge & Kegan Paul: 58. |
[2] | Goldmeier E (1936/1972) Similarity in visually perceived forms. Psychol Issues 8: 1-135. |
[3] | Greene E (2007) Recognition of objects displayed with incomplete sets of discrete boundary dots. Percept Mot Skills 104: 1043-1059. |
[4] |
Greene E, Visani A (2015) Recognition of letters displayed as briefly flashed dot patterns. Atten Percept Psychophys 77: 1955-1969. doi: 10.3758/s13414-015-0913-6
![]() |
[5] | Greene E (2016) Information persistence evaluated with low-density dot patterns. Acta Psychol 170: 215-225. |
[6] | Green DM, Swets JA (1966) Signal Detection Theory and Psychophysics. New York; Wiley. |
[7] |
Hautus MJ, van Hout D, Lee HS (2009) Variants of A Not-A and 2AFC tests: Signal Detection Theory models. Food Qual Prefer 20: 222-229. doi: 10.1016/j.foodqual.2008.10.002
![]() |
[8] | Macmillan NA, Creelman CD (2005) Detection Theory: A User's Guide, New Jersey: Lawrence Erlbaum. |
[9] |
Hautus MJ (1995) Corrections for extreme proportions and their biasing effects on estimated values of d´. Behav Res Methods Instrum Comput 27: 46-51. doi: 10.3758/BF03203619
![]() |
[10] |
Miller J (1996) The sampling distribution of d'. Percept Psychophys 58: 65-72. doi: 10.3758/BF03205476
![]() |
[11] | Hautus J (2012) SDT Assistant (version 1.0) [Software]. Available from http://hautus.org. |
[12] | Hautus MJ (1997) Calculating estimates of sensitivity from group data: pooled versus averaged estimators. Behav Res Methods Instrum Comput 29: 556-562. |
[13] | Hubel DH, Wiesel TN (1959) Receptive fields of single neurons in the cat's striate cortex. J Physiol 148: 574-591. |
[14] | Hubel DH, Wiesel TN (1968) Receptive fields and functional architecture of monkey striate cortex. J Physiol 195: 215-243. |
[15] | Selfridge OG (1959) Pandemonium: A Paradigm for Learning in the Mechanization of Thought Process, London; HM Stationary Office. |
[16] |
Sutherland NS (1968) Outlines of a theory of visual pattern recognition in animals and man. Proc R Soc Lond B Biol Sci 171: 297-317. doi: 10.1098/rspb.1968.0072
![]() |
[17] | Binford TO (1971) Visual perception by computer. Proc IEEE Conf Syst Control, Miami, FL. |
[18] |
Barlow HB (1972) Single units and sensation: a neuron doctrine for perceptual psychology. Perception 1: 371-394. doi: 10.1068/p010371
![]() |
[19] | Milner PM (1974) A model for visual shape recognition. Psychol Rev 81: 521-535. |
[20] | Palmer SE (1975) Visual perception and world knowledge: notes on a model of sensory-cognitive interaction. In: Norman DA, Rumelhart DE eds, Explorations in Cognitions, San Francisco: WH Freeman & Co.: 279-307. |
[21] | Marr D (1982) Vision: A Computational Investigation into the Human Representation and Processing of Information, New York: Freemen: 51-79. |
[22] | DeValois RL, Devalois KK (1991) Vernier acuity with stationary moving Gabors. Vision Res 31: 1619-1626. |
[23] |
Panda R, Chatterji BN (1996) Gabor function: an efficient tool for digital image processing. IETE Tech Rev 13: 225-231. doi: 10.1080/02564602.1996.11416611
![]() |
[24] | Kohonen T, Oja E (1998) Visual feature analysis by the self-organizing maps. Neural Comput Appl 7: 273-286. |
[25] |
Pettet MW, McKee SP, Grzywacz NM (1998) Constraints on long range interactions mediating contour detection. Vision Res 38: 865-879. doi: 10.1016/S0042-6989(97)00238-1
![]() |
[26] |
Pennefather PM, Chandna A, Kovacs I, et al. (1999) Contour detection threshold: repeatabililty and learning with 'contour cards.' Spat Vis 12: 257-266. doi: 10.1163/156856899X00157
![]() |
[27] | Taylor G, Hipp D, Moser A, et al. (2014) The development of contour processing: evidence from physiology and psychophysics. Front Psychol 5: e719 |
[28] | Dong X, Chantlier MJ (2016) Perceptually motivated image features using contours. IEEE Trans Image Process 25: 5050-5062. |
[29] | Edelman S (1999) Representation and Recognition in Vision, MIT Press. |
[30] |
Cooke T, Jakel F, Wallraven C, et al. (2007) Multimodal similarity and categorization of novel, three-dimensional objects. Neuropsychologia 45: 484-495. doi: 10.1016/j.neuropsychologia.2006.02.009
![]() |
[31] | Hayworth KJ (2012) Dynamically partitionable autoassociative networks as a solution to the neural binding problem. Front Comput Neurosci 6: e73. |
[32] |
Rodriguez-Sanchez AJ, Tsotsos JK (2012) The roles of endstopped and curvature tuned computations in a hierarchical representation of 2D shape. PLoS ONE 7: e42058. doi: 10.1371/journal.pone.0042058
![]() |
[33] | Hopfield JJ (1995) Pattern recognition computation using action potential timing for stimulus representation. Nature 376: 33-36. |
[34] | Hopfield JJ (1996) Transforming neural computations and representing time. Proc Nat Acad Sci 83: 15440-15444. |
[35] |
Maass W (1997) Fast sigmoidal networks via spiking neurons. Neural Comput 9: 279-304. doi: 10.1162/neco.1997.9.2.279
![]() |
[36] | McClelland JL (2013) Integrating probabilistic models of perception and interactive neural networks: a historical and tutorial review. Front Psychol 4: e503. |
[37] | Guan T, Wang Y, Duan L, et al. (2015) On-device mobile landmark recognition using binarized descriptor and multifeature fusion. ACM Trans Intell Syst Technol 7: e12. |
[38] |
Wei B, Guan T, Duan L, et al. (2015) Wide area localization and tracking on camera phones for mobile segmented reality systems. Multimed Syst 21: 381-399. doi: 10.1007/s00530-014-0364-2
![]() |
[39] | Pan H, Guan T, Luo Y, et al. (2016) Dense 3D reconstruction combining depth and RGB information. Neurocomputing 175: 644-651. |
[40] | Ullman S (1976) Filling-in the gaps: the shape of subjective contours and a model for their generation. Biol Cybern 25: 1-6. |
[41] | Sha'ashua A, Ullman S (1988) Structural saliency: the detection of globally salient structures using a locally connected network. In: Proc 2nd Intern Conf Comput Vision, Clearwater FL: 321-327. |
[42] |
Kellman PJ, Shipley TF (1991) A theory of visual interpolation in object rperception. Cogn Psychol 23: 141-221. doi: 10.1016/0010-0285(91)90009-D
![]() |
[43] | Shipley TF, Kellman PJ (1992) Strength of visual interpolation depends on the ratio of physically specified to total edge length. Percept Psychophys 52: 97-106. |
[44] | Cormen TH, Leherson CE, Rivest RL, et al. (2001) Single-source shortest paths and all pairs shortest paths. In: Introduction to Algorithms, MIT Press & McGraw-Hill: 580-642. |
[45] | Field DJ, Hayes A, Hess RF (1993) Contour integration by the human visual system: evidence for a local "association field." Vision Res 33: 173-193. |
[46] | Kwon TK, Agrawal K, Li Y, et al. (2016) Spatially-global integration of closed, fragmented contours by finding the shortest-path in a log-polar representation. Vision Res 126: 143-163. |
[47] | Sceniak MP, Hawken MJ, Shapley R (2001) Visual spatial characterization of Macaque V1 neurons. J Neurophysiol 85: 1873-1887. |
[48] | Bowers JS (2009) On the biological plausibility of grandmother cells: implications for neural network theories in psychology and neuroscience. Psychol Rev 116: 220-251. |
[49] | Greene E (2008) Additional evidence that contour attributes are not essential cues for object recognition. Behav Brain Funct 4: e26. |
[50] | Greene E (2016) How do we know whether three dots form an equilateral triangle? JSM Brain Sci 1: 1002. |
[51] |
Lichtsteiner P, Posch C, Delbruck T (2008) A 128x128 120 dB 15 µs latency asynchronous temporal contrast vision sensor. IEEE J Solid-State Circuits 43: 566-576. doi: 10.1109/JSSC.2007.914337
![]() |
[52] | Robinson DA (1964) The mechanisms of human saccadic eye movement. J Physiol 174: 245-264. |
[53] | Zimmermann E, Lappe M (2016) Visual space constructed by saccadic motor maps. Front Human Neurosci 10: e225. |
[54] |
McSorlely E, McCloy R, Williams L (2016) The concurrent programming of saccades. PLoS ONE 11: e0168724. doi: 10.1371/journal.pone.0168724
![]() |
[55] | Bhutani N, Sengupta S, Basu D, et al. (2017) Parallel activation of prospective motor plans during visually-guided sequential saccades. Neurosci 45: 631-642. |
[56] |
Feldman JA, Ballard DH (1982) Connectionist models and their properties. Cogn Sci 6: 205-254. doi: 10.1207/s15516709cog0603_1
![]() |
[57] |
Fukushima K (1980) Neocognitron: a self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biol Cybern 36: 193-202. doi: 10.1007/BF00344251
![]() |
[58] |
Rolls ET (1992) Neurophysiological mechanisms underlying face processing within and beyond the temporal cortical visual areas. Phil Trans R Soc 335: 11-21. doi: 10.1098/rstb.1992.0002
![]() |
[59] | Wallis G, Rolls ET (1997) Invariant face and object recognition in the visual system. Prog Neurobiol 51: 167-194. |
[60] | Riesenhuber M, Poggio T (2000) Models of object recognition. Nature Neurosci (S3): 1199-1204. |
[61] | Pasupathy A, Connor CE (2001) Shape representation in area V4: position-specific tuning for boundary conformation. J Neurophysiol 86: 2505-2519. |
[62] |
Suzuki N, Hashimoto N, Kashimori Y, et al. (2004) A neural model of predictive recognition in form pathway of visual cortex. BioSystems 76: 33-42. doi: 10.1016/j.biosystems.2004.05.004
![]() |
[63] |
Pinto N, Cox DD, DeCarlo JJ (2008) Why is real-world visual object recognition hard? PLoS Comput Biol 4: e27. doi: 10.1371/journal.pcbi.0040027
![]() |
[64] | Hancock PJB, Walton L, Mitchell G, et al. (2008) Segregation by onset asynchrony. J Vision 8: 1-21. |
[65] | Karplus I, Goren M, Algorn D (1982) A preliminary experimental analysis of predator face recognition by Chromis caenuleus (Pisces, Pomacentridae). Z Tierpsychol 58: 53-65. |
[66] | Siebeck UE, Parker AN, Sprenger D, et al. (2010) A species of reef fish that uses untraviolet patterns for covert face recognition. Curr Biol 20: 407-410. |
[67] | Karplus I, Katzenstein R, Goren M (2006) Predator recognition and social facilitation of predator avoidance in coral reef fish Dascyllus marginatus juveniles. Mar Ecol Prog Ser 319: 215-223. |
[68] | Siebeck UE, Litherland L, Wallis GM (2009) Shape learning and discrimination in reef fish. J Exp Biol 212: 2113-2119. |
[69] |
Newport C, Wallis G, Reshitnyk Y, et al. (2016) Discrimination of human faces by archerfish (Toxotes catareus). Sci Rep 6: e27523. doi: 10.1038/srep27523
![]() |
[70] |
Greschner M, Field GD, Li PH, et al. (2014) A polyaxonal amacrine cell population in the primate retina. J Neurosci 34: 3597-3606. doi: 10.1523/JNEUROSCI.3359-13.2014
![]() |
[71] |
Greschner M, Heitman AK, Field GD, et al. (2016) Identification of a retinal circuit for recurrent suppression using indirect electrical imaging. Curr Biol 26: 1935-1942. doi: 10.1016/j.cub.2016.05.051
![]() |
[72] | Greene E (2007) Retinal encoding of ultrabrief shape recognition cues. PLoS One 2: e871. |
[73] | Greene E (2016) Retinal encoding of shape boundaries. JSM Anat Physiol 1: 1002. |
1. | Mahmoud S. Mehany, Faizah D. Alanazi, An η-Hermitian solution to a two-sided matrix equation and a system of matrix equations over the skew-field of quaternions, 2025, 10, 2473-6988, 7684, 10.3934/math.2025352 |