Characterizations of matrix equalities involving the sums and products of multiple matrices and their generalized inverse

Yongge Tian; Yongge Tian

doi:10.3934/era.2023298

Electronic Research Archive

2023, Volume 31, Issue 9: 5866-5893. doi: 10.3934/era.2023298

Previous Article Next Article

Research article

Characterizations of matrix equalities involving the sums and products of multiple matrices and their generalized inverse

Yongge Tian ^,

CBE, Shanghai Business School, Shanghai, China

Received: 24 March 2023 Revised: 25 July 2023 Accepted: 09 August 2023 Published: 30 August 2023

It is common knowledge that matrix equalities involving ordinary algebraic operations of inverses or generalized inverses of given matrices can be constructed arbitrarily from theoretical and applied points of view because of the noncommutativity of the matrix algebra and singularity of given matrices. Two of such matrix equality examples are given by $A_1B_1^{g_1}C_1 + A_2B_2^{g_2}C_2 + \cdots + A_kB_k^{g_k}C_{k} = D$ and $A_1B_1^{g_1}A_2B_2^{g_2} \cdots A_kB_k^{g_k}A_{k+1} = A$ , where $A_1$ , $A_2$ , $\ldots$ , $A_{k+1}$ , $C_1$ , $C_2$ , $\ldots$ , $C_{k}$ and $A$ and $D$ are given, and $B_1^{g_1}$ , $B_2^{g_2}$ , $\ldots$ , $B_k^{g_k}$ are generalized inverses of matrices $B_1$ , $B_2$ , $\ldots$ , $B_k$ . These two matrix equalities include many concrete cases for different choices of the generalized inverses, and they have been attractive research topics in the area of generalized inverse theory. As an ongoing investigation of this subject, the present author presents in this article several groups of new results and facts on constructing and characterizing the above matrix equalities for the mixed combinations of $\{1\}$ - and $\{1, 2\}$ -generalized inverses of matrices with $k = 2, 3$ by using some elementary methods, including a series of explicit rank equalities for block matrices.

Keywords:

Citation: Yongge Tian. Characterizations of matrix equalities involving the sums and products of multiple matrices and their generalized inverse[J]. Electronic Research Archive, 2023, 31(9): 5866-5893. doi: 10.3934/era.2023298

Related Papers:

[1]	Qinghua Zhou, Li Wan, Hongbo Fu, Qunjiao Zhang . Exponential stability of stochastic Hopfield neural network with mixed multiple delays. AIMS Mathematics, 2021, 6(4): 4142-4155. doi: 10.3934/math.2021245
[2]	Huahai Qiu, Li Wan, Zhigang Zhou, Qunjiao Zhang, Qinghua Zhou . Global exponential periodicity of nonlinear neural networks with multiple time-varying delays. AIMS Mathematics, 2023, 8(5): 12472-12485. doi: 10.3934/math.2023626
[3]	Ravi P. Agarwal, Snezhana Hristova . Stability of delay Hopfield neural networks with generalized proportional Riemann-Liouville fractional derivative. AIMS Mathematics, 2023, 8(11): 26801-26820. doi: 10.3934/math.20231372
[4]	Qian Cao, Xiaojin Guo . Anti-periodic dynamics on high-order inertial Hopfield neural networks involving time-varying delays. AIMS Mathematics, 2020, 5(6): 5402-5421. doi: 10.3934/math.2020347
[5]	Zhigang Zhou, Li Wan, Qunjiao Zhang, Hongbo Fu, Huizhen Li, Qinghua Zhou . Exponential stability of periodic solution for stochastic neural networks involving multiple time-varying delays. AIMS Mathematics, 2024, 9(6): 14932-14948. doi: 10.3934/math.2024723
[6]	Er-Yong Cong, Yantao Wang, Xian Zhang, Li Zhu . A new approach for accuracy-preassigned finite-time exponential synchronization of neutral-type Cohen–Grossberg memristive neural networks involving multiple time-varying leakage delays. AIMS Mathematics, 2025, 10(5): 10917-10942. doi: 10.3934/math.2025496
[7]	Boonyachat Meesuptong, Peerapongpat Singkibud, Pantiwa Srisilp, Kanit Mukdasai . New delay-range-dependent exponential stability criterion and $ H_\infty $ performance for neutral-type nonlinear system with mixed time-varying delays. AIMS Mathematics, 2023, 8(1): 691-712. doi: 10.3934/math.2023033
[8]	Yijia Zhang, Tao Xie, Yunlong Ma . Robustness analysis of exponential stability of Cohen-Grossberg neural network with neutral terms. AIMS Mathematics, 2025, 10(3): 4938-4954. doi: 10.3934/math.2025226
[9]	Xiaofang Meng, Yongkun Li . Pseudo almost periodic solutions for quaternion-valued high-order Hopfield neural networks with time-varying delays and leakage delays on time scales. AIMS Mathematics, 2021, 6(9): 10070-10091. doi: 10.3934/math.2021585
[10]	Qinghua Zhou, Li Wan, Hongbo Fu, Qunjiao Zhang . Pullback attractor of Hopfield neural networks with multiple time-varying delays. AIMS Mathematics, 2021, 6(7): 7441-7455. doi: 10.3934/math.2021435

Abstract

1. Introduction

The stability of Hopfield neural networks is a necessary prerequisite in the practical applications of signal processing ^[1,2], pattern recognition ^[3], associative memory ^[4], nonlinear programming ^[5] and optimization ^[6]. Accordingly, the stability of various delayed Hopfield neural networks has been received enough attention ^{[7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33]}. In particular, the neutral delays have been introduced into the model of neural networks and the stability of neutral-type neural networks has become a hot topic ^{[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33]}. The mathematical model of neutral-type neural networks extends its application domain to a wider class of practical engineering problems, (for the detailed applications of such neural networks, the readers may refer to the references ^[34,35,36].)

The model of Hopfield neural networks of neutral type considered in this paper is expressed as

$\begin{eqnarray} \dot{x}_{i}(t) & = & - c_{i} x_{i}(t) +\sum\limits^{n}_{j = 1}a_{ij} f_j(x_{j}(t)) +\sum\limits^{n}_{j = 1}b_{ij} g_j(x_{j}(t-\tau_{ij}(t))) +\sum\limits^{n}_{j = 1}e_{ij}\dot{x}_{j}(t-\xi_{ij}(t)) +u_i , t\geq 0; \end{eqnarray}$

(1.1)

$x_i(t) = \varphi_i(t), \dot{x}_i(t) = \phi_i(t), t\in [- \max \{\tau , \xi \}, 0], i = 1, \cdots, n,$

where $\xi, \tau$ and $c_i$ are positive numbers, $\varphi_i(t)$ and $\phi_i(t)$ are continuous functions, $\xi_{ij}(t)$ and $\tau_{ij}(t)$ are delay functions, $f_j(\cdot)$ and $g_j(\cdot)$ are nonlinear continuous activation functions. These functions satisfy that for every $x, y\in R,$ $t\geq 0$ and $i, j = 1, \cdots, n,$

$\begin{eqnarray} 0\leq \xi_{ij}(t)\leq \xi, 0\leq \tau_{ij}(t)\leq \tau, \dot{\xi}_{ij}(t)\leq \overline{\xi}, \dot{ \tau}_{ij}(t)\leq \overline{\tau} ; \end{eqnarray}$

(1.2)

$\begin{eqnarray} | f_i(x)-f_i(y)| \leq l_i|x-y|, | g_i(x)-g_i(y)| \leq m_i|x-y|, \end{eqnarray}$

(1.3)

where $\overline{\xi}, \overline{\tau}, l_i$ and $m_i$ are some positive numbers. From ^[37,38], we know that $x(t) = (x_1(t), \cdots, x_n(t))^T$ of (1.1) is continuously differentiable.

The mathematical expression of system (1.1) includes some models studied in existing references. For example, when $\tau_{ij}(t) = \tau_{j}(t)$ and $\xi_{ij}(t) = \xi_{j}(t)$ , system (1.1) transforms into the following vector-matrix form studied ^[18,20]:

$\begin{eqnarray} \dot{x}(t) & = & - C x (t) +A f (x(t)) +B g(x(t-\tau(t)))+E\dot{x} (t-\xi(t)) +u , t\geq 0, \end{eqnarray}$

(1.4)

where $A = (a_{ij})_{n\times n}, B = (b_{ij})_{n\times n}, C = diag(c_1, \cdots, c_n), u = (u_1, \cdots, u_n)^T,$

$x(t) = (x_1(t), \cdots, x_n(t))^T, \dot{x}(t-\xi(t)) = (\dot{x}_1(t-\xi_1(t)), \cdots, \dot{x}_n(t-\xi_n(t)))^T,$

$f (x(t)) = (f_1(x_1(t)), \cdots, f_n(x_n(t)))^T , g (x(t-\tau(t))) = (g_1(x_1(t-\tau_1(t))), \cdots, g_n(x_n(t-\tau_n(t))))^T.$

When $\tau_{ij}(t) = \tau_{ij}, f_j = g_j$ and $\xi_{ij}(t) = \xi_{ij}$ (or $\xi_{j}$ ), system (1.1) transforms into the following system studied in ^[30]:

$\begin{eqnarray} \dot{x}_{i}(t) & = & -c_{i} x_{i}(t) +\sum\limits^{n}_{j = 1}a_{ij}f_j(x_{j}(t)) +\sum\limits^{n}_{j = 1}b_{ij} f_j(x_{j}(t-\tau_{ij} )) +\sum\limits^{n}_{j = 1}e_{ij}\dot{x}_{j}(t-\xi_{ ij} )+u_i, \end{eqnarray}$

(1.5)

or the following system studied in ^[29]:

(1.6)

Different from the systems studied in ^{[18,19,20,21,22,23,24,25,26,27,28]}, system (1.1) cannot be expressed in the vector-matrix form due to the existence of multiple delays $\tau_{ij}(t)$ and $\xi_{ij}(t)$ . Therefore, it is impossible to obtain the stability conditions of the linear matrix inequality form for the system with multiple delays ^[29,31]. In this case, it is necessary to develop new mathematical techniques and find more suitable Lyapunov-Krasovskii functional for the stability analysis of system (1.1).

On the other hand, the results of ^{[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33]} have only provided the sufficient conditions of global asymptotic stability and have not considered the exponential stability. Actually, in the stability analysis of neural networks, exponential stability is a better property than asymptotic stability since it can converge to the equilibrium point faster and provides information about the decay rate of the networks. Moreover, the exponential stability property can ensure that whatever transformation occurs, the network stability of fast storage activity mode remains unchanged by self-organization ^[39]. Therefore, the exponential stability analysis of neural networks with multiple time-varying delays is worth investigating.

In addition, based on our careful review of recently almost all the existing stability results for system (1.1), we have realized that the research on the exponential stability of system (1.1) has not received enough attention. These facts have been the main motivations of the current paper to focus on the exponential stability of system (1.1).

The primary contributions and innovations of this work are summarized as follows: (1) novel sufficient conditions of exponential stability are established for Hopfield neural networks of neutral type with multiple time-varying delays; (2) a modified and suitable Lyapunov-Krasovskii functional is provided to study the exponential stability; (3) novel sufficient conditions of global asymptotical stability are provided for the systems studied in the existing references; (4) compared with the existing results, the established conditions are less conservative.

2. Exponential stability

From ^[18,20], we know that system (1.1) has at least one equilibrium point $x^* = (x^*_1, \cdots, x^*_n)^T$ under the conditions (1.2) and (1.3). By employing the formula $y_i(t) = x_i(t)-x^*_i (i = 1, \cdots, n)$ , system (1.1) can be transformed into the following equivalent system

$\begin{eqnarray} \dot{y}_{i}(t) & = & -c_{i} y_{i}(t) +\sum\limits^{n}_{j = 1}a_{ij} \widetilde{ f}_j(y_{j}(t)) +\sum\limits^{n}_{j = 1}b_{ij} \widetilde{ g}_j(y_{j}(t-\tau_{ij}(t))) +\sum\limits^{n}_{j = 1}e_{ij}\dot{y}_{j}(t-\xi_{ij}(t)), \end{eqnarray}$

(2.1)

where

$\widetilde{ f}_{j}( y_{j}(t)) = f_{j}( y_{j}(t)+x^*_j)-f_{j}( x^*_j), \widetilde{g}_j(y_{j}(t-\tau_{ij}(t))) = g_j(y_{j}(t-\tau_{ij}(t))+x^*_j)-g_j( x^*_j ).$

Similarly, system (1.5) can be transformed into the following equivalent system

$\begin{eqnarray} \dot{y}_{i}(t) & = & -c_{i} y_{i}(t) +\sum\limits^{n}_{j = 1}a_{ij} \widetilde{ f}_j(y_{j}(t)) +\sum\limits^{n}_{j = 1}b_{ij} \widetilde{f}_j(y_{j}(t-\tau_{ij} )) +\sum\limits^{n}_{j = 1}e_{ij}\dot{y}_{j}(t-\xi_{ij} ). \end{eqnarray}$

(2.2)

It is obvious that if the origin of system (2.1) is exponentially stable, then the equilibrium point of system (1.1) is also exponentially stable. Meanwhile, the origin of system (2.1) and the equilibrium point of system (1.1) are also globally asymptotically stable. Now, we state the main stability result for system (2.1).

Theorem 1. Suppose that there exist some positive numbers $\gamma, p_1, p_2, \cdots, p_n$ such that $\gamma < 1, \max\{ \bar{\xi}, \; \bar{\tau}\} < 1-\gamma,$

$\gamma p_i {c}_{i} - \sum\limits^{n}_{j = 1} p_j (l_i |a_{ji}|\gamma+ m_i|b_{ji}| ) > 0, p_i\gamma- \sum\limits^{n}_{j = 1} p_j|e_{ji}| > 0, i = 1, 2, \cdots, n .$

Then the origin of system (2.1) is exponentially stable.

Proof. Let

$h_1(z) = 1 - e^{ \xi z}\overline{\xi}- e^{ \xi z}\gamma, h_2(z) = 1- e^{ \tau z}\overline{\tau}-\gamma,$

$h_3(z) = \gamma p_i {c}_{i}- z p_i \gamma -z \sum\limits^{n}_{j = 1}p_j |e_{ji}|- \sum\limits^{n}_{j = 1} p_j (l_i|a_{ji}|\gamma+ m_i|b_{ji}| e^{ \tau z} ) .$

Then, $\dot{h}_i(z) < 0, h_i(+\infty) < 0, i = 1, 2, 3,$ and

$h_1(0) = 1 - \overline{\xi}- \gamma > 0, h_2(0) = 1- \overline{\tau}-\gamma > 0,$

$h_3(0) = \gamma p_i {c}_{i}- \sum\limits^{n}_{j = 1} p_j (l_i|a_{ji}|\gamma+ m_i|b_{ji}| ) > 0.$

Therefore, there exist some positive numbers $\lambda_1, \lambda_2$ and $\lambda_3$ such that $h_1(\lambda_1) = h_2(\lambda_2) = h_3(\lambda_3) = 0$ , which implies that there must exist a positive number $\lambda\in (0, \min\{\lambda_1, \lambda_2, \lambda_3 \})$ such that

$\begin{eqnarray} 1 - e^{\lambda \xi}\overline{\xi}- e^{\lambda \xi}\gamma > 0, 1- e^{\lambda \tau }\overline{\tau}-\gamma > 0, \end{eqnarray}$

(2.3)

$\begin{eqnarray} \gamma p_i {c}_{i}- \lambda p_i \gamma -\lambda \sum\limits^{n}_{j = 1}p_j |e_{ji}|- \sum\limits^{n}_{j = 1} p_j (l_i|a_{ji}|\gamma+ m_i|b_{ji}| e^{\lambda \tau } ) > 0. \end{eqnarray}$

(2.4)

We construct the following Lyapunov-Krasovskii functional ^[40]

$\begin{eqnarray} V (t)& = & e^{\lambda (t+\xi)} \sum\limits^{n}_{i = 1} [p_i \gamma- \sum\limits^{n}_{j = 1}p_j |e_{ji}| sgn(y_i(t)) sgn(\dot{y}_i(t))]| y_i(t)| +\sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}p_i|e_{ij}| \int^t_{t-\xi_{ij}(t)}e^{\lambda (s+\xi)} |\dot{y}_{j}(s)|ds\\&& +e^{\lambda \xi} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j\int^t_{t-\tau_{ij}(t)} e^{\lambda (s+\tau)}|y_{j}(s)| ds, \end{eqnarray}$

(2.5)

and derive

$\begin{eqnarray} \min\limits_{1\leq i\leq n}\{ p_i \gamma-\sum\limits^{n}_{j = 1}p_j|e_{ji}|\} e^{\lambda (t+\xi)} \| y(t) \|_1 &\leq& e^{\lambda (t+\xi)} \sum\limits^{n}_{i = 1} [p_i \gamma-\sum\limits^{n}_{j = 1}p_j|e_{ji}| ]| y_i(t) | \\ &\leq& e^{\lambda (t+\xi)} \sum\limits^{n}_{i = 1} [p_i \gamma-\sum\limits^{n}_{j = 1}p_j |e_{ji}| sgn(y_i(t)) sgn(\dot{y}_i(t))]| y_i(t)| \leq V (t) , \end{eqnarray}$

(2.6)

$\begin{eqnarray} V(0) & = & \nonumber e^{\lambda \xi } \sum\limits^{n}_{i = 1} [p_i \gamma-\sum\limits^{n}_{j = 1}p_j |e_{ji}| sgn(y_i(0)) sgn(\dot{y}_i(0))]| y_i(0)|\\&&\nonumber + e^{\lambda \xi }\sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}p_i|e_{ij}| \int^0_{-\xi_{ij}(0)}e^{\lambda s} |\dot{y}_{j}(s)|ds + e^{\lambda (\xi+\tau)} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j\int^0_{-\tau_{ij}(0)} e^{\lambda s}|y_{j}(s)| ds \\ \nonumber &\leq& \nonumber e^{\lambda \xi }\max\limits_{1\leq i\leq n}\{p_i \gamma+\sum\limits^{n}_{j = 1}p_j |e_{ji}| \} \| y(0)\|_1 +e^{\lambda \xi }\sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}p_i|e_{ij}| \int^0_{-\xi} |\dot{y}_{j}(s)|ds \\&& +e^{\lambda (\xi+\tau)} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j\int^0_{-\tau} |y_{j}(s)| ds. \end{eqnarray}$

(2.7)

Taking the right upper Dini derivative of the first term and the time derivatives of the all other terms in the Lyapunov-Krasovskii functional $V(t)$ along the trajectories of system (2.1), we derive

$\begin{eqnarray} \dot{V}(t)& = &\nonumber \lambda e^{\lambda (t+\xi)}\nonumber \sum\limits^{n}_{i = 1} [p_i \gamma- \sum\limits^{n}_{j = 1}p_j |e_{ji}| sgn(y_i(t)) sgn(\dot{y}_i(t))] | y_i(t)|\\&&\nonumber +e^{\lambda t}\sum\limits^{n}_{i = 1} e^{\lambda \xi} [ p_i \gamma- \sum\limits^{n}_{j = 1}p_j |e_{ji}| sgn(y_i(t)) sgn(\dot{y}_i(t))]sgn(y_i(t)) \dot{ y}_i(t)\\&&\nonumber +e^{\lambda (t+\xi)}\sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}p_i|e_{ij}| |\dot{y}_{j}(t)| -\sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}(1-\dot{\xi}_{ij} (t)) e^{\lambda (t-\xi_{ij} (t)+\xi)}p_i|e_{ij}| |\dot{y}_{j}(t-\xi_{ij}(t))| \\ && \nonumber +e^{\lambda \xi} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j \bigg( e^{\lambda (t+\tau)}|y_{j}(t)| -(1-\dot{\tau}_{ij}(t) )e^{\lambda (t-\tau_{ij}(t)+\tau)}|y_{j}(t-\tau_{ij}(t))|\bigg) \\&\leq &\nonumber \lambda e^{\lambda (t+\xi)}\nonumber \sum\limits^{n}_{i = 1} (p_i \gamma+ \sum\limits^{n}_{j = 1}p_j |e_{ji}| ) | y_i(t)| +e^{\lambda t}\sum\limits^{n}_{i = 1} \bigg\{ e^{\lambda \xi} p_i \gamma sgn(y_i(t)) \dot{ y}_i(t) \\&&\nonumber- e^{\lambda \xi} \sum\limits^{n}_{j = 1}p_j |e_{ji}| (sgn(y_i(t)) )^2 | \dot{ y}_i(t)| +e^{\lambda \xi } \sum\limits^{n}_{j = 1}p_j|e_{ji}| |\dot{y}_{i}(t)|\bigg\}\\&& \nonumber + \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}( \dot{\xi}_{ij} (t)e^{\lambda (t-\xi_{ij} (t)+\xi)} -e^{\lambda (t-\xi_{ij} (t)+\xi)}) p_i|e_{ij}| |\dot{y}_{j}(t-\xi_{ij}(t))| \\ && \nonumber +e^{\lambda \xi} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j \bigg( e^{\lambda (t+\tau)}|y_{j}(t)| +[ \dot{\tau}_{ij}(t) e^{\lambda (t-\tau_{ij}(t)+\tau)}- e^{\lambda (t-\tau_{ij}(t)+\tau)} ]|y_{j}(t-\tau_{ij}(t))|\bigg) \\&\leq &\nonumber \lambda e^{\lambda (t+\xi)}\nonumber \sum\limits^{n}_{i = 1} (p_i \gamma+ \sum\limits^{n}_{j = 1}p_j |e_{ji}| ) | y_i(t)| +e^{\lambda t}\sum\limits^{n}_{i = 1} \bigg\{ e^{\lambda \xi} p_i \gamma sgn(y_i(t)) \dot{ y}_i(t) \\&&\nonumber - e^{\lambda \xi} \sum\limits^{n}_{j = 1}p_j |e_{ji}| (sgn(y_i(t)) )^2 | \dot{ y}_i(t)| +e^{\lambda \xi } \sum\limits^{n}_{j = 1}p_j|e_{ji}| |\dot{y}_{i}(t)|\bigg\}\\&& \nonumber + \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}( \bar{\xi} e^{\lambda (t+\xi)} -e^{\lambda t}) p_i|e_{ij}| |\dot{y}_{j}(t-\xi_{ij}(t))| \\ && +e^{\lambda \xi} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j \bigg( e^{\lambda (t+\tau)}|y_{j}(t)| +( \bar{\tau} e^{\lambda (t +\tau)}- e^{\lambda t})|y_{j}(t-\tau_{ij}(t))|\bigg) \nonumber \\ & = &\nonumber\lambda e^{\lambda (t+\xi)}\nonumber \sum\limits^{n}_{i = 1} (p_i \gamma+ \sum\limits^{n}_{j = 1}p_j |e_{ji}| ) | y_i(t)| +e^{\lambda t}\sum\limits^{n}_{i = 1} \bigg\{ e^{\lambda \xi} p_i \gamma sgn(y_i(t)) \dot{ y}_i(t) \\&&\nonumber- e^{\lambda \xi} \sum\limits^{n}_{j = 1}p_j |e_{ji}| (sgn(y_i(t)) )^2 | \dot{ y}_i(t)| +e^{\lambda \xi } \sum\limits^{n}_{j = 1}p_j|e_{ji}| |\dot{y}_{i}(t)|\bigg\} \\ &&\nonumber + e^{\lambda t}( \overline{\xi}e^{\lambda \xi}-1)\sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}p_i|e_{ij}| |\dot{y}_{j}(t-\xi_{ij}(t))| \\&& +e^{\lambda \xi} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j \bigg ( e^{\lambda (t+\tau)}|y_{j}(t)| +( \overline{\tau}e^{\lambda \tau} -1)e^{\lambda t}|y_{j}(t-\tau_{ij}(t))|\bigg). %\\&& +\eta \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1}(e^{\lambda (t+\tau)}|y_{j}(t)|-(1-\overline{\tau})e^{\lambda t}| y_{j}( t-\tau_{ij}(t))| ). \end{eqnarray}$

(2.8)

It is noted that for $y_i(t)\neq 0,$

and for $y_i(t) = 0,$

$\begin{eqnarray*} && \nonumber e^{\lambda \xi} p_i \gamma sgn(y_i(t)) \dot{ y}_i(t)-e^{\lambda \xi} \sum\limits^{n}_{j = 1}p_j |e_{ji}| (sgn(y_i(t)) )^2 | \dot{ y}_i(t)| +e^{\lambda \xi } \sum\limits^{n}_{j = 1}p_j|e_{ji}| |\dot{y}_{i}(t)| \\& = & e^{\lambda \xi } \sum\limits^{n}_{j = 1}p_j|e_{ji}|\nonumber | \dot{ y}_i(t)| \\& = &\nonumber e^{\lambda \xi}\bigg\{ - \sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t)) {c}_{i} y_{i}(t) + \sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t)) \sum\limits^{n}_{j = 1}a_{ij} \widetilde{f}_j(y_{j}(t)) \\&& \nonumber + \sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t)) \sum\limits^{n}_{j = 1}b_{ij} \widetilde{g}_j(y_{j}(t-\tau_{ij}(t))) + \sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t))\sum\limits^{n}_{j = 1}e_{ij}\dot{y}_{j}(t-\xi_{ij}(t)) \bigg\} \\&\leq &\nonumber e^{\lambda \xi}\bigg\{ - \gamma p_i {c}_{i} |y_i(t)| + |\sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t)) \sum\limits^{n}_{j = 1}a_{ij} \widetilde{f}_j(y_{j}(t))| \\&& \nonumber + | \sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t)) \sum\limits^{n}_{j = 1}b_{ij} \widetilde{g}_j(y_{j}(t-\tau_{ij}(t)))| + |\sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t))\sum\limits^{n}_{j = 1}e_{ij}\dot{y}_{j}(t-\xi_{ij}(t)) |\bigg\} \\&\leq&\nonumber e^{\lambda \xi}\bigg\{ - \gamma p_i {c}_{i} |y_i(t)|+ \sum\limits^{n}_{j = 1}p_j|e_{ji}| \sum\limits^{n}_{j = 1}|a_{ij}| l_j|y_{j}(t) | + \sum\limits^{n}_{j = 1}p_j|e_{ji}| \sum\limits^{n}_{j = 1}|b_{ij}|m_j| y_{j}(t-\tau_{ij}(t)) |\bigg\} \\&& \nonumber + e^{\lambda \xi} \sum\limits^{n}_{j = 1}p_j|e_{ji}|\sum\limits^{n}_{j = 1}|e_{ij}||\dot{y}_{j}(t-\xi_{ij}(t))| \\ &\leq&\nonumber e^{\lambda \xi}\bigg\{ - \gamma p_i {c}_{i} |y_i(t)|+ \gamma p_i \sum\limits^{n}_{j = 1}|a_{ij}| l_j|y_{j}(t) | + \gamma p_i \sum\limits^{n}_{j = 1}|b_{ij}|m_j| y_{j}(t-\tau_{ij}(t)) |\bigg\}\\&& \nonumber + e^{\lambda \xi}\gamma p_i\sum\limits^{n}_{j = 1}|e_{ij}||\dot{y}_{j}(t-\xi_{ij}(t))|, \end{eqnarray*}$

where we use $\sum\limits^{n}_{j = 1} p_j|e_{ji}| < p_i\gamma,$ and $\sum\limits^{n}_{j = 1}p_j|e_{ji}| sgn(\dot{y}_i(t)) {c}_{i} y_{i}(t) = \gamma p_i {c}_{i} |y_i(t)| = 0$ when $y_i(t) = 0.$

Therefore, for every $y_i(t) \in R$ , we have

$\begin{eqnarray} && e^{\lambda \xi} p_i \gamma sgn(y_i(t)) \dot{ y}_i(t)-e^{\lambda \xi} \sum\limits^{n}_{j = 1}p_j |e_{ji}| (sgn(y_i(t)) )^2 | \dot{ y}_i(t)| +e^{\lambda \xi } \sum\limits^{n}_{j = 1}p_j|e_{ji}| |\dot{y}_{i}(t)| \\ &\leq& e^{\lambda \xi}\bigg\{ - \gamma p_i {c}_{i} |y_i(t)|+ \gamma p_i \sum\limits^{n}_{j = 1}|a_{ij}| l_j|y_{j}(t) | + \gamma p_i \sum\limits^{n}_{j = 1}|b_{ij}|m_j| y_{j}(t-\tau_{ij}(t)) |\bigg\}\\&& + e^{\lambda \xi}\gamma p_i\sum\limits^{n}_{j = 1}|e_{ij}||\dot{y}_{j}(t-\xi_{ij}(t))|. \end{eqnarray}$

(2.9)

Then, from (2.3), (2.4), (2.8) and (2.9), we have

$\begin{eqnarray} \dot{V}(t)&\leq & \lambda e^{\lambda (t+\xi)} \sum\limits^{n}_{i = 1} (p_i \gamma+ \sum\limits^{n}_{j = 1}p_j |e_{ji}| ) | y_i(t)| +e^{\lambda t}\sum\limits^{n}_{i = 1} \bigg\{ e^{\lambda \xi}\bigg( - \gamma p_i {c}_{i} |y_i(t)| + \gamma p_i \sum\limits^{n}_{j = 1}|a_{ij}| l_j|y_{j}(t) | \\&& + \gamma p_i \sum\limits^{n}_{j = 1}|b_{ij}|m_j| y_{j}(t-\tau_{ij}(t)) |\bigg) +( e^{\lambda \xi}\gamma + e^{\lambda \xi}\overline{\xi} -1) p_i\sum\limits^{n}_{j = 1}|e_{ij}||\dot{y}_{j}(t-\xi_{ij}(t))|\bigg\} \\&& +e^{\lambda \xi} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}|m_j \bigg ( e^{\lambda (t+\tau)}|y_{j}(t)| +( \overline{\tau}e^{\lambda \tau}-1)e^{\lambda t}|y_{j}(t-\tau_{ij}(t))|\bigg) \\& = & -e^{\lambda (t+\xi)} \sum\limits^{n}_{i = 1} \bigg( \gamma p_i {c}_{i}- \lambda p_i \gamma -\lambda \sum\limits^{n}_{j = 1}p_j |e_{ji}|- \sum\limits^{n}_{j = 1} p_j (l_i |a_{ji}|\gamma+m_i|b_{ji}| e^{\lambda \tau }) \bigg) | y_i(t)| \\&& -e^{\lambda t}(1 - e^{\lambda \xi}\overline{\xi}- e^{\lambda \xi}\gamma )\sum\limits^{n}_{i = 1}\sum\limits^{n}_{j = 1}p_i|e_{ij}||\dot{y}_{j}(t-\xi_{ij}(t))| \\&& - (1- e^{\lambda \tau }\overline{\tau}-\gamma) e^{\lambda (t+\xi)} \sum\limits^{n}_{i = 1} \sum\limits^{n}_{j = 1} p_i |b_{ij}| m_j |y_{j}(t-\tau_{ij}(t))| \leq 0. \end{eqnarray}$

(2.10)

Finally, (2.6), (2.7) and (2.10) imply that there must exist a number $\delta > 1$ such that

$\| y(t) \|_1 \leq \delta e^{- \lambda t} (\|\varphi\|_1+\|\phi\|_1), t\geq 0,$

where

$\|\varphi\|_1 = \sup\limits_{t\in [-\max \{\tau , \; \xi \}, \;0]}\|\varphi(t)\|_1, \|\phi\|_1 = \sup\limits_{t\in [- \max \{\tau , \; \xi \}, \; 0]}\|\phi(t)\|_1.$

Generally speaking, it is not easy to find the values of the positive constants $p_1, \cdots, p_n$ . Therefore, it is necessary to give a result without involving the constants $p_1, \cdots, p_n.$ The following result is a special case of Theorem 1 for $p_1 = \cdots = p_n$ , which is easier to validate.

Theorem 2. Suppose that there exists a positive number $\gamma$ such that

$\max\{ \bar{\xi} , \bar{\tau}\} < 1-\gamma, \sum\limits^{n}_{j = 1} |e_{ji}| < \gamma < 1, \gamma {c}_{i} - \sum\limits^{n}_{j = 1} (l_i |a_{ji}|\gamma+m_i |b_{ji}| ) > 0, i = 1, 2, \cdots, n.$

Then the equilibrium point of system (1.1) is globally asymptotically stable, even exponentially stable.

Remark 1. Theorem 1 and Theorem 2 give novel sufficient conditions of global asymptotical stability for the systems studied in ^{[18,20,29,30]} since these systems are some special cases of system (1.1).

For system (1.5) or (1.6), Theorem 1 and Theorem 2 give the following results.

Corollary 1. Suppose that there exist some positive numbers $\gamma, p_1, p_2, \cdots, p_n$ such that $\gamma < 1,$

$\gamma p_i {c}_{i} - \sum\limits^{n}_{j = 1} p_j l_i ( |a_{ji}|\gamma+ |b_{ji}| ) > 0, p_i\gamma- \sum\limits^{n}_{j = 1} p_j|e_{ji}| > 0, i = 1, 2, \cdots, n .$

Then the equilibrium point of system (1.5) (or (1.6)) is globally asymptotically stable, even exponentially stable.

Corollary 2. Suppose that there exists a positive number $\gamma$ such that $\gamma < 1,$

$\gamma {c}_{i} - \sum\limits^{n}_{j = 1} l_i ( |a_{ji}|\gamma+ |b_{ji}| ) > 0, \gamma- \sum\limits^{n}_{j = 1} |e_{ji}| > 0, i = 1, 2, \cdots, n .$

Then the equilibrium point of system (1.5) (or (1.6)) is globally asymptotically stable, even exponentially stable.

Remark 2. It is noted that $\gamma p_i {c}_{i} - \sum\limits^{n}_{j = 1} p_j l_i (|a_{ji}|\gamma+ |b_{ji}|) > 0$ and $0 < \gamma < 1$ can deduce that $p_i {c}_{i} - \sum\limits^{n}_{j = 1} p_j l_i (|a_{ji}| + |b_{ji}|) > 0.$ Therefore, the conditions of Corollary 2 are less conservative than those of the result in ^[29].

Remark 3. For system (1.5), the conditions of Theorem 1 in ^[30] are as follows:

$\begin{eqnarray*} \alpha_i& = & \nonumber c^2_i-l^2_i\sum\limits^{n}_{j = 1}|\sum\limits^{n}_{k = 1}a_{ki}a_{kj}|-l^2_i\sum\limits^{n}_{j = 1} \sum\limits^{n}_{k = 1}(|a_{ji}||b_{jk}|+|b_{ji}||a_{jk}| +|b_{ji}||b_{jk}|)\\&&\nonumber -l^2_i\sum\limits^{n}_{j = 1} \sum\limits^{n}_{k = 1}c_j |e_{jk}|(|a_{ji}|+|b_{ji}|)-\sum\limits^{n}_{j = 1} \sum\limits^{n}_{k = 1}c_j |e_{ji}|(|a_{jk}|+|b_{jk}|) \\&&-\sum\limits^{n}_{j = 1}c^2_i |e_{ij}|-\sum\limits^{n}_{j = 1}c^2_j |e_{ji}| > 0, \end{eqnarray*}$

and $1-\sum^{n}_{j = 1} |e_{ji}| > 0, i = 1, \cdots, n.$ Example 2 demonstrates the above conditions are not satisfied while the conditions of Corollary 1 and Corollary 2 can be satisfied.

Remark 4. Global asymptotical stability of a more general class of multiple delayed neutral type neural network model has been studied in ^[31,32]. The results of ^[31,32] can be particularized for system (2.2) as follows:

i) The sufficient conditions of global asymptotical stability are given in ^[31]:

$\begin{eqnarray} \sigma_i& = & 2c_i- \sum\limits^{n}_{j = 1}(l_j| a_{ij}|+l_i|a_{ji}|)- \sum\limits^{n}_{j = 1}(l_j| b_{ij}|+l_i|b_{ji}|)- \sum\limits^{n}_{j = 1}(| e_{ij}| + |e_{ji}| ) \\&& -\sum\limits^{n}_{j = 1}\sum\limits^{n}_{k = 1}(l_i|a_{ki}| |e_{kj}| +l_i|b_{ki}||e_{kj}|)-\sum\limits^{n}_{j = 1}\sum\limits^{n}_{k = 1}(l_k|a_{jk}| |e_{ji}| +l_k|b_{jk}||e_{ji}|) > 0, \end{eqnarray}$

and $1-\sum^{n}_{j = 1} |e_{ji}| > 0, i = 1, \cdots, n.$

ii) The sufficient conditions of global asymptotical stability are given in ^[32]:

$\begin{eqnarray*} \epsilon_i& = & \nonumber c^2_i- \sum\limits^{n}_{j = 1}|\sum\limits^{n}_{k = 1}a_{ki}a_{kj}|- \sum\limits^{n}_{j = 1} \sum\limits^{n}_{k = 1}(|a_{ji}||b_{jk}|\\&&\nonumber+|a_{ji}||e_{jk}| +|b_{ji}||a_{jk}| +|b_{ji}||e_{jk}|+|b_{ji}||b_{jk}|) > 0, \end{eqnarray*}$

$\begin{eqnarray*} \varepsilon_{ij} = \frac{1}{n}- \sum\limits^{n}_{k = 1}(|e_{ji}||e_{jk}| +|a_{jk}||e_{ji}|+|b_{jk}| |e_{ji}|) > 0, i, j = 1, \cdots, n. \end{eqnarray*}$

Example 3 demonstrates the above conditions are not satisfied while the conditions of Corollary 1 and Corollary 2 can be satisfied.

Example 1. Consider system (1.1) with the following conditions:

$A = \left(\begin{array}{cccc} 1 & -1 & 1 & -1\\ -1 & 1& 1 & -1\\ 1 & 1& 1& 1 \\ 1 & -1 &-1 & -1 \end{array}\right), B = \left(\begin{array}{cccc} 1 & 1 &-1 & 1\\ -1 & -1 &1 &-1\\ 1 & 1& 1 & 1\\ -1 & -1 &-1 & 1 \end{array}\right), E = \left(\begin{array}{cccc} 0.1 & -0.1 &-0.1 & -0.1\\ -0.1 & 0.1 &0.1 &0.1\\0.1 & -0.1& 0.1 & -0.1\\ -0.1 & -0.1 &-0.1 & 0.1 \end{array}\right),$

$c_1 = c_2 = c_3 = 9, c_4 = 8, f_i(x) = tanh(x), g_i(x) = 0.5tanh(x), \xi_{ii}(t) = 0.1sin t+0.1, \tau_{ii}(t) = 0.1cos t+0.1; \xi_{ij}(t) = 0.1cos t+0.1, \tau_{ij}(t) = 0.1sin t+0.1, i\neq j; i, j = 1, 2, 3, 4.$

We calculate $\bar{\xi} = \bar{\tau} = 0.1, l_i = 1, m_i = 0.5, \sum\limits^{4}_{j = 1} |e_{ji}| = 0.4$ and

$\begin{eqnarray*} {c}_{i}\gamma - \sum\limits^{4}_{j = 1} (l_i |a_{ji}|\gamma +m_i |b_{ji}| ) = \begin{cases} 5\gamma-2 , & i = 1, 2, 3; \\ 4\gamma-2 , & i = 4. \end{cases} \end{eqnarray*}$

It is clear that when $\gamma \in (0.5, 0.9),$ Theorem 2 holds.

Now we choose $\gamma = 0.424$ and require that

$\gamma p_i {c}_{i} - \sum\limits^{4}_{j = 1} p_j (l_i |a_{ji}|\gamma+ m_i|b_{ji}| ) = 0.01 , i = 1, 2, 3, 4.$

Then, we calculate $p_1 = p_2 = p_3 = 2.2222, p_4 = 2.5000,$ and

$\begin{eqnarray*} p_i\gamma- \sum\limits^{4}_{j = 1} p_j|e_{ji}| = 0.424 p_i - 0.1\sum\limits^{4}_{j = 1} p_j = \begin{cases} 0.025552, & i = 1, 2, 3; \\ 0.14334, & i = 4. \end{cases} \end{eqnarray*}$

Thus, all conditions of Theorem 1 are satisfied.

If we choose $\gamma = 0.423$ and still require that

$\gamma p_i {c}_{i} - \sum\limits^{4}_{j = 1} p_j (l_i |a_{ji}|\gamma+ m_i|b_{ji}| ) = 0.01 , i = 1, 2, 3, 4.$

Then $p_i < 0, i = 1, 2, 3, 4.$ Therefore, when $\gamma\in [0.424, 0.9),$ the system (1.1) in this example is globally asymptotically stable, even exponentially stable.

Example 2. Consider system (1.5) with the conditions: $c_1 = c_2 = c_3 = 9, c_4 = 8.5, f_i(x) = tanh(x), \xi_{ii} = 0.2, \tau_{ii} = 0.1; \xi_{ij} = 0.3, \tau_{ij} = 0.4, i\neq j; i, j = 1, 2, 3, 4,$ the matrices $A, B, E$ are the same as in Example 1.

Then, we calculate $l_i = 1, \sum\limits^{4}_{j = 1} |e_{ji}| = 0.4$ and

$\begin{eqnarray*} {c}_{i}\gamma - \sum\limits^{4}_{j = 1} l_i ( |a_{ji}|\gamma + |b_{ji}| ) = \begin{cases} 5\gamma-4 , & i = 1, 2, 3; \\ 4.5\gamma-4 , & i = 4. \end{cases} \end{eqnarray*}$

It is clear that when $\gamma \in (8/9, 1),$ Corollary 2 holds.

Now we choose $\gamma = 0.85$ and require that

$\gamma p_i {c}_{i} - \sum\limits^{4}_{j = 1} p_j l_i (|a_{ji}|\gamma+|b_{ji}| ) = 0.01 , i = 1, 2, 3, 4.$

Then, we calculate $p_1 = p_2 = p_3 = 0.0708, p_4 = 0.0750,$ and

$\begin{eqnarray*} p_i\gamma- \sum\limits^{4}_{j = 1} p_j|e_{ji}| = 0.85 p_i - 0.1\sum\limits^{4}_{j = 1} p_j = \begin{cases} 0.03144, & i = 1, 2, 3; \\ 0.03501, & i = 4. \end{cases} \end{eqnarray*}$

Thus, all conditions of Corollary 1 are satisfied.

On the other hand, we calculate

$\begin{eqnarray*} \alpha_1& = & \nonumber 81- \sum\limits^{4}_{j = 1}|\sum\limits^{4}_{k = 1}a_{k1}a_{kj}|- 3\times 4\times 4 - 35.5\times 4 \times 0.2 -35.5\times 4 \times 0.2 \\&&-81 \times 0.4-315.25 \times 0.1 < 0, \end{eqnarray*}$

which shows the conditions of Theorem 1 in ^[30] are not satisfied. Therefore, Theorem 1 in ^[30] is invalid for the system (1.5) in this example.

Example 3. Consider system (2.2) with all parameters and functions are the same as in Example 2. Then, we calculate $l_i = 1, \sum^{4}_{j = 1} |e_{ji}| = 0.4,$

$\varepsilon_{ij} = \frac{1}{4}-0.21\times 4 = -0.59 < 0,$

$\sigma_{1} = 2\times 9-2\times 4-2\times 4-0.2\times 4-0.2\times 4^2-0.2\times 4^2 = -5.2 < 0.$

Therefore, the sufficient conditions of the results in ^[31,32] are not satisfied. Meanwhile, all conditions of Corollary 1 and Corollary 2 are staisfied for system (2.2) since system (1.5) is equivalent to system (2.2).

3. Conclusions

This paper has investigated the exponential stability of neutral-type Hopfield neural networks involving multiple time-varying delays. Different from some existing results, linear matrix inequality approach cannot be used to determine the stability conditions of such networks since the networks studied here can not be expressed in vector-matrix form. By using a modified and suitable Lyapunov-Krasovskii functional and inequality techniques, novel algebraic conditions are established to ensure the exponential stability and the global asymptotic stability of neutral-type Hopfield neural networks involving multiple time-varying delays. Compared with some references, the networks studied here is more general and the established conditions are less conservative. Three examples are given to demonstrate the effectiveness of the theoretical results and compare the established stability conditions to the previous results.

Acknowledgments

The authors would like to thank the editor and the reviewers for their detailed comments and valuable suggestions. This work was supported by the National Natural Science Foundation of China (No: 11971367, 11826209, 11501499, 61573011 and 11271295), the Natural Science Foundation of Guangdong Province (2018A030313536).

Conflict of interest

All authors declare no conflicts of interest in this paper.

References

[1]	A. Ben-Israel, T. N. E. Greville, Generalized Inverses: Theory and Applications, 2nd edition, Springer-Verlag, New York, 2003.
[2]	D. S. Bernstein, Scalar, Vector, and Matrix Mathematics: Theory, Facts, and Formulas-Revised and Expanded Edition, Princeton University Press, Princeton, NJ, 2018. https://doi.org/10.1515/9781400888252
[3]	S. L. Campbell, C. D. Meyer, Generalized Inverses of Linear Transformations, SIAM, Philadelphia, 2009.
[4]	D. Carlson, Generalized inverse invariance, partial orders, and rank-minimization problems for matrices, in Current Trends in Matrix Theory, North-Holland, New York, (1987), 81–87.
[5]	R. E. Hartwig, 1–2 inverses and the invariance of $BA^{+}C$ , Linear Algebra Appl., 11 (1975), 271–275. https://doi.org/10.1016/0024-3795(75)90026-9 doi: 10.1016/0024-3795(75)90026-9
[6]	X. Liu, Z. Miao, Y. Yu, Note on the invariance properties of operator products involving generalized inverses, Abstr. Appl. Anal., 2014 (2014), 213458. https://doi.org/10.1155/2014/213458 doi: 10.1155/2014/213458
[7]	Z. Xiong, Y. Qin, Invariance properties of an operator product involving generalized inverses, Electron. J. Linear Algebra, 22 (2011), 694–703. https://doi.org/10.13001/1081-3810.1467 doi: 10.13001/1081-3810.1467
[8]	J. Groß, Y. Tian, Invariance properties of a triple matrix product involving generalized inverses, Linear Algebra Appl., 417 (2006), 94–107. https://doi.org/10.1016/j.laa.2006.03.026 doi: 10.1016/j.laa.2006.03.026
[9]	B. Jiang, Y. Tian, Necessary and sufficient conditions for nonlinear matrix identities to always hold, Aequat. Math., 93 (2019), 587–600. https://doi.org/10.1007/s00010-018-0610-3 doi: 10.1007/s00010-018-0610-3
[10]	B. Jiang, Y. Tian, Linear and multilinear functional identities in a prime ring with applications, J. Algebra Appl., 20 (2021), 2150212. https://doi.org/10.1142/S0219498821502121 doi: 10.1142/S0219498821502121
[11]	B. Jiang, Y. Tian, Invariance property of a five matrix product involving two generalized inverses, Anal. Sti. Univ. Ovid. Const. Ser. Mat., 29 (2021), 83–92. https://doi.org/10.2478/auom-2021-0006 doi: 10.2478/auom-2021-0006
[12]	Y. Tian, B. Jiang, Closed-form formulas for calculating the max-min ranks of a triple matrix product composed by generalized inverses, Comput. Appl. Math., 37 (2018), 5876–5919. https://doi.org/10.1007/s40314-018-0668-5 doi: 10.1007/s40314-018-0668-5
[13]	Y. Tian, Miscellaneous reverse order laws for generalized inverses of matrix products with applications, Adv. Oper. Theory, 5 (2020), 1889–1942. https://doi.org/10.1007/s43036-020-00072-8 doi: 10.1007/s43036-020-00072-8
[14]	Y. Tian, A family of 512 reverse order laws for generalized inverses of a matrix product: a review, Heliyon, 6 (2020), e04924. https://doi.org/10.1016/j.heliyon.2020.e04924 doi: 10.1016/j.heliyon.2020.e04924
[15]	Y. Tian, Characterizations of matrix equalities for generalized inverses of matrix products, Axioms, 11 (2022), 291. https://doi.org/10.3390/axioms11060291 doi: 10.3390/axioms11060291
[16]	Y. Tian, On a family of matrix equalities that involve multiple products of generalized inverses, Aequat. Math., 97 (2023), 563–574. https://doi.org/10.1007/s00010-023-00940-7 doi: 10.1007/s00010-023-00940-7
[17]	R. Penrose, A generalized inverse for matrices, Proc. Cambridge Philos. Soc., 51 (1955), 406–413. https://doi.org/10.1017/S0305004100030401 doi: 10.1017/S0305004100030401
[18]	G. Marsaglia, G. P. H. Styan, Equalities and inequalities for ranks of matrices, Linear Multilinear Algebra, 2 (1974), 269–292.
[19]	Y. Tian, Upper and lower bounds for ranks of matrix expressions using generalized inverses, Linear Algebra Appl., 355 (2002), 187–214. https://doi.org/10.1016/S0024-3795(02)00345-2 doi: 10.1016/S0024-3795(02)00345-2
[20]	Y. Tian, More on maximal and minimal ranks of Schur complements with applications, Appl. Math. Comput., 152 (2004), 675–692. https://doi.org/10.1016/S0096-3003(03)00585-X doi: 10.1016/S0096-3003(03)00585-X

This article has been cited by:

1.	Leonid Berezansky, Josef Diblík, Zdeněk Svoboda, Zdeněk Šmarda, Exponential stability criteria for linear neutral systems with applications to neural networks of neutral type, 2023, 360, 00160032, 301, 10.1016/j.jfranklin.2022.11.012
2.	Syed Anayet Karim, Mohd Shareduwan Mohd Kasihmuddin, Saratha Sathasivam, Mohd. Asyraf Mansor, Siti Zulaikha Mohd Jamaludin, Md Rabiol Amin, A Novel Multi-Objective Hybrid Election Algorithm for Higher-Order Random Satisfiability in Discrete Hopfield Neural Network, 2022, 10, 2227-7390, 1963, 10.3390/math10121963
3.	Zhongjie Zhang, Xian Zhang, Tingting Yu, Global exponential stability of neutral-type Cohen–Grossberg neural networks with multiple time-varying neutral and discrete delays, 2022, 490, 09252312, 124, 10.1016/j.neucom.2022.03.068
4.	Marat Akhmet, Madina Tleubergenova, Zakhira Nugayeva, Unpredictable and Poisson Stable Oscillations of Inertial Neural Networks with Generalized Piecewise Constant Argument, 2023, 25, 1099-4300, 620, 10.3390/e25040620
5.	Qinghua Zhou, Li Wan, Hongshan Wang, Hongbo Fu, Qunjiao Zhang, Exponential stability of Cohen-Grossberg neural networks with multiple time-varying delays and distributed delays, 2023, 8, 2473-6988, 19161, 10.3934/math.2023978
6.	Hongjun Qiu, Yanhong Zhang, Fanchao Kong, Stability analysis of periodic solutions of the neutral-type neural networks with impulses and time-varying delays, 2023, 37, 0354-5180, 6021, 10.2298/FIL2318021Q
7.	Xian Zhang, Zhongjie Zhang, Tingting Yu, Xin Wang, Global Results on Exponential Stability of Neutral Cohen–Grossberg Neural Networks Involving Multiple Neutral and Discrete Time-Varying Delays: A Method Based on System Solutions, 2023, 55, 1370-4621, 11273, 10.1007/s11063-023-11375-1
8.	Ravi P. Agarwal, Snezhana Hristova, Stability of delay Hopfield neural networks with generalized proportional Riemann-Liouville fractional derivative, 2023, 8, 2473-6988, 26801, 10.3934/math.20231372
9.	Dongmei Xia, Kaiyuan Chen, Lin Sun, Guojin Qin, Research on reachable set boundary of neutral system with various types of disturbances, 2025, 20, 1932-6203, e0317398, 10.1371/journal.pone.0317398
10.	N. Mohamed Thoiyab, Mostafa Fazly, R. Vadivel, Nallappan Gunasekaran, Stability analysis for bidirectional associative memory neural networks: A new global asymptotic approach, 2025, 10, 2473-6988, 3910, 10.3934/math.2025182
11.	Xin Wang, Yu Xue, Xian Zhang, Yantao Wang, 2025, Chapter 6, 978-981-96-1176-8, 119, 10.1007/978-981-96-1177-5_6

Reader Comments

Your name:*

Email:*
© 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

Electronic Research Archive

1.1 1.7

Metrics

Article views(1228) PDF downloads(35) Cited by(2)

Preview PDF

Download XML

Export Citation

Article outline

Show full outline

Electronic Research Archive

Characterizations of matrix equalities involving the sums and products of multiple matrices and their generalized inverse

Related Papers:

Abstract

1. Introduction

2. Exponential stability

3. Conclusions

Acknowledgments

Conflict of interest

References

This article has been cited by:

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Other Articles By Authors

Catalog

Electronic Research Archive

Characterizations of matrix equalities involving the sums and products of multiple matrices and their generalized inverse

Related Papers:

Abstract

1. Introduction

2. Exponential stability

3. Conclusions

Acknowledgments

Conflict of interest

References

This article has been cited by:

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Other Articles By Authors

Related pages

Tools

Export File

Citation

Format

Content

Catalog