
Citation: Yanjie Hong, Wanbiao Ma. Sufficient and necessary conditions for global attractivity and stability of a class of discrete Hopfield-type neural networks with time delays[J]. Mathematical Biosciences and Engineering, 2019, 16(5): 4936-4946. doi: 10.3934/mbe.2019249
[1] | Gonzalo Robledo . Feedback stabilization for a chemostat with delayed output. Mathematical Biosciences and Engineering, 2009, 6(3): 629-647. doi: 10.3934/mbe.2009.6.629 |
[2] | Yun Hu, Qianqian Duan . Solving the TSP by the AALHNN algorithm. Mathematical Biosciences and Engineering, 2022, 19(4): 3427-3448. doi: 10.3934/mbe.2022158 |
[3] | Masaki Sekiguchi, Emiko Ishiwata, Yukihiko Nakata . Dynamics of an ultra-discrete SIR epidemic model with time delay. Mathematical Biosciences and Engineering, 2018, 15(3): 653-666. doi: 10.3934/mbe.2018029 |
[4] | Luyao Li, Licheng Fang, Huan Liang, Tengda Wei . Observer-based event-triggered impulsive control of delayed reaction-diffusion neural networks. Mathematical Biosciences and Engineering, 2025, 22(7): 1634-1652. doi: 10.3934/mbe.2025060 |
[5] | Yuan Ma, Yunxian Dai . Stability and Hopf bifurcation analysis of a fractional-order ring-hub structure neural network with delays under parameters delay feedback control. Mathematical Biosciences and Engineering, 2023, 20(11): 20093-20115. doi: 10.3934/mbe.2023890 |
[6] | Huaying Liao, Zhengqiu Zhang . Finite-time stability of equilibrium solutions for the inertial neural networks via the figure analysis method. Mathematical Biosciences and Engineering, 2023, 20(2): 3379-3395. doi: 10.3934/mbe.2023159 |
[7] | Bingrui Zhang, Jin-E Zhang . Fixed-deviation stabilization and synchronization for delayed fractional-order complex-valued neural networks. Mathematical Biosciences and Engineering, 2023, 20(6): 10244-10263. doi: 10.3934/mbe.2023449 |
[8] | Sheng Wang, Lijuan Dong, Zeyan Yue . Optimal harvesting strategy for stochastic hybrid delay Lotka-Volterra systems with Lévy noise in a polluted environment. Mathematical Biosciences and Engineering, 2023, 20(4): 6084-6109. doi: 10.3934/mbe.2023263 |
[9] | Toshikazu Kuniya, Tarik Mohammed Touaoula . Global stability for a class of functional differential equations with distributed delay and non-monotone bistable nonlinearity. Mathematical Biosciences and Engineering, 2020, 17(6): 7332-7352. doi: 10.3934/mbe.2020375 |
[10] | Yuanpei Xia, Weisong Zhou, Zhichun Yang . Global analysis and optimal harvesting for a hybrid stochastic phytoplankton-zooplankton-fish model with distributed delays. Mathematical Biosciences and Engineering, 2020, 17(5): 6149-6180. doi: 10.3934/mbe.2020326 |
It is well known that stability properties of Hopfield-type neural networks (see, for example, [1], [2]) have been widely used in optimization problems, associative memory, pattern recognition and signal processing (see, for example, [3], [4], [5], [6], [7]). Early researches on stability of the equilibria of Hopfield-type neural networks mainly involves ordinary differential equations. However, time delays are inevitable in the propagation of electrical signals between neurons or in complex signal processing. Therefore, time delays are commonly found in artificial neural networks. At the same time, a large number of theoretical studies show that time delays may have complicated and unpredictable impacts on local or global dynamics of neural networks. Time delays may destroy stability properties of the equilibria of neural networks and cause their orbits to oscillate and even become chaotic etc.
Due to both the universality of time delays in the real world and the importance of constructing the neural networks, many mathematical researchers have paid great attention to Hopfield-type neural networks with time delays and a great amount of results have been achieved (see, for example, [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18]).
Compared with continuous Hopfield-type neural networks, theoretical studies on discrete Hopfield-type neural networks are much less. However, discrete Hopfield-type neural networks may have rich kinetic behavior.
In an ideal case, the discrete Hopfield-type neural networks should be produced in a way to reflect the dynamics of the continuous counterparts. Specifically, the discrete networks should inherit the dynamical characteristics of the continuous networks, and also maintain functional similarity to the continuous system and any physical or biological reality that the continuous has[19]. However, unfortunately, as pointed out in [20], discrete neural networks can hardly maintain the dynamics of continuous counterparts even during very short sampling periods. Therefore, it is crucial to study the similarity of the dynamic properties between the discrete neural networks and the continuous counterparts.
It should be mentioned that, based on the analytical method on the distribution of the roots of transcendental functions developed in [24], some sharp conditions are obtained for delay independent stability of a class of additive neural networks are given in [25].
In [26], [27], [28] and [29], by constructing suitable nonlinear algebraic equations and using the results in [25] and the properties of M-matrix in [30], some sufficient conditions and necessary conditions for global attractivity of retarded and neutral Hopfield-type neural networks are obtained.
The purpose of the short paper is to extend the methods in [24], [25], [26], [27], [28] and [29] to the analysis for global attractivity of a class of discrete Hopfield-type neural networks with time delays. This paper is organized as follows. In Section 2, a class of discrete Hopfield-type neural networks with time delays and some basic assumptions are given. Then, ultimate boundedness of the solutions of the discrete neural networks is considered by establishing a class of related nonlinear algebraic equations. At the same time, for convenience, some equivalent conditions of M-matrix are introduced. In Section 3, global attractivity of the discrete neural networks is discussed, and a sufficient condition is given. In Section 4, necessary conditions for stability of the discrete neural networks are obtained. To prove the necessary conditions, several important lemmas similar to [24] and [25] have been extended to the discrete neural networks. Finally, in Section 5, a 3-dimensional discrete Hopfield-type neural networks with time delays is considered by the method of numerical simulations. The numerical simulations show that, under certain conditions, the sufficient condition is very close to the necessary condition.
The following n-dimensional discrete Hopfield-type neural networks with time delays is considered,
Ui(k+1)=biUi(k)+n∑j=1aijfj(Uj(k−τij))(i=1,2,⋯,n;k∈Z+). | (2.1) |
In (2.1), Z+ denotes the set of all nonnegative integers. Ui(k) denotes the potential(or voltage) of the ith neuron at time k. bi is the real constant which represents the ith neuron receives itself's feedback. aij is the real constant which represents the synaptic connection weight between the jth neuron and the ith neuron. fj(U) denotes the nonlinear processing function in the dendrites of neurons. The time delay τij is the nonnegative integer which represents the signal propagation delay from jth neuron to ith neuron. As pointed out in [1], fj(U) can be regarded as a smooth input-output function because the biological information in neurons often lies in a short-time average of the firing rate[21]. Hence, fj(U) can be regarded as a continuously differentiable function with respect to the variable U. Based on biological meanings, it is assumed that the following conditions are satisfied.
(H1) |bi|<1(i=1,2,...,n).
(H2) fi(0)=0, |fi(U)|≤1 (U∈R), limU→−∞fi(U)=−1, limU→+∞fi(U)=1 (i=1,2,...,n).
(H3) f′i(U)>0 (U∈R), f′i(0)=supU∈Rf′i(U)=1 (i=1,2,...,n).
Remark 2.1. Based on [22] and [23], the constant bi represents the rate with which the ith neuron will reset its potential(or voltage) to the resting state in isolation when disconnected from the networks. Hence, the assumption (H1) is reasonable in biology. Moreover, according to [2] the nonlinear processing function fj can be considered as some sigmoidal function with saturation properties. Therefore, the assumptions (H2) and (H3) have rationality in biology.
Remark 2.2 For convenience, and without losing the generality of theoretical analysis, it is assumed that the external inputs are equal to 0 in (2.1). When there are constant external inputs in (2.1), similar to [22] and [25], the neural networks can be transformed to a system similar to (2.1) without external inputs.
The initial condition for (2.1) is given as ui(s)=ϕi(s),(s=−r,−r+1,⋯,0,i=1,2,⋯,n), where ϕi(s) are given real constants, r=max{τij|i,j=1,2,⋯,n}. It is clear that the solution U(k)=(U1(k),U2(k),⋯,Un(k))T of (2.1) with the above initial condition is existent and unique for all k∈Z+.
(2.1) always has the equilibrium U=(0,0,⋯,0)T≡0. This paper mainly discusses sufficient conditions for global attractivity of U=0 and necessary conditions for stability of U=0.
First, similar to [26] and [27], we have the following important result on dissipation of (2.1).
Lemma 2.1. If (H1) - (H3) hold, then any solution U(k) of (2.1) satisfies lim supk→+∞ |Ui(k)|≤Mi, where each Mi satisfies the following nonlinear algebraic equations,
Mi=11−|bi|n∑j=1|aij|¯fj(Mj)(i=1,2,⋯,n), |
where ¯fj(Mj)=max{fj(Mj),−fj(−Mj)}(i=1,2,⋯,n).
Proof. By (2.1) and (H2), it has
|Ui(k+1)|≤|bi||Ui(k)|+n∑j=1|aij|(i=1,2,⋯,n,k∈Z+). |
Note that |bi|<1 by (H1), it further has
lim supk→+∞|Ui(k)|≤11−|bi|n∑j=1|aij|≜Mi,0(i=1,2,⋯,n). |
For ∀ε>0, there is a positive integer K0, such that |Ui(k−τij)|≤Mi,0+ε (i=1,2,⋯,n,k>K0). According to (H3), it has
|Ui(k+1)|≤|bi||Ui(k)|+n∑j=1|aij|¯fj(Mj,0+ε)(i=1,2,⋯,n,k>K0). |
Hence,
lim supk→+∞|Ui(k)|≤11−|bi|n∑j=1|aij|¯fj(Mj,0+ε)(i=1,2,⋯,n). |
Let ε→0. It has lim supk→+∞|Ui(k)|≤Mi,1, where
Mi,1=11−|bi|n∑j=1|aij|¯fj(Mj,0)≤Mi,0(i=1,2,⋯,n). |
Repeating the above process, there has a decreasing sequence {Mi,p}, which satisfies lim supk→+∞|Ui(k)|≤Mi,p, where
Mi,p=11−|bi|n∑j=1|aij|¯fj(Mj,p−1)(i=1,2,⋯,n,p=1,2,⋯). |
Denote lim supp→+∞Mi,p=Mi (i=1,2,⋯,n). It has limk→+∞supk|Ui(k)|≤Mi, where
Mi=11−|bi|n∑j=1|aij|¯fj(Mj)(i=1,2,⋯,n). |
Lemma 2.2. ([30]) Define the set of real matrices Zn×n≜{Z=(zij)|zij≤0,i≠j, i,j=1,2,⋯,n}. For any C∈Zn×n, the following conditions are equivalent:
(i) All the principle minors of C are nonnegative.
(ii) For any vector x≠0, let y=Cx, then there is some i∈{1,2,⋯,n} such that xi≠0 and xiyi≥0.
If C∈Zn×n satisfies any of the above conditions, then C is called an M-matrix.
In this section, we will give sufficient conditions for global attractivity of U=0 of (2.1).
Define the matrix C∈Zn×n as follows:
C=[1−|b1|−|a11|−|a12|⋯−|a1n|−|a21|1−|b2|−|a22|⋯−|a2n|⋮⋮⋱⋮−|an1|−|an2|⋯1−|bn|−|ann|]=(cij)n×n. |
Using the similar method in [28], it has the following result for global attractivity of U=0 of (2.1).
Theorem 3.1. If C is an M-matrix, then U=0 of (2.1) is globally attractive for any time delays τij (i,j=1,2,⋯,n).
Proof. According to Lemma 2.1, any solution U(k) of (2.1) satisfies lim supk→+∞|Ui(k)|≤Mi, where
Mi=11−|bi|n∑j=1|aij|¯fj(Mj)(i=1,2,⋯,n). | (3.1) |
Subsequently, it will be shown that there must be Mi=0 for all i=1,2,⋯,n.
Firstly, it assumes that Mi>0 for all i=1,2,⋯,n. Then according to (3.1), (H2) and (H3), it has that n∑j=1|aij|>0 (i=1,2,...,n). Hence, it has that ¯fj(Mj)<Mj(j=1,2,...,n). which further implies that
Mi(1−|bi)=n∑j=1|aij|¯fj(Mj)<n∑j=1|aij|Mj(i=1,2,⋯,n), |
i.e., CM<0, here M=(M1,M2,⋯,Mn)T>0.
Denote x=M>0 and y=CM. Then for all i∈{1,2,⋯,n}, it has
xiyi=Min∑j=1cijMj<0, |
which contradicts to the condition (ⅱ) in Lemma 2.2. Therefore, it must have that there is at least some j such that Mj=0.
Without loss of generality, it is assumed that Mn=0. Afterwards, it can be shown from (3.1) that
Mi=11−|bi|n−1∑j=1|aij|¯fj(Mj)(i=1,2,⋯,n−1). |
With the same method as discussed above, it has that M2=M3=⋯=Mn−1=Mn=0. Finally, it has from (3.1) that
M1=11−|b1||a11|¯f1(M1). |
If a11=0, it is clear that M1=0. If a11≠0, it has that
M1=11−|b1||a11|¯f1(M1)<11−|b1||a11|M1, |
i.e., (1−|b1|−|a11|)M1<0, which is a contradiction to C being an M-matrix (hence 1−|b1|−|a11|≥0). Therefore, there must be M1=0. In summary, it has that M1=M2=⋯=Mn=0.
In this section, we will consider the necessary conditions for stability of U=0 of (2.1). First of all, from (H2) and (H3), it has the following linearized system of (2.1) at U=0,
Ui(k+1)=biUi(k)+n∑j=1aijUj(k−τij)(i=1,2,⋯,n;k∈Z+). | (4.1) |
The characteristic equation of (4.1) is
P(λ)≜det[λ−b1−a11λ−τ11−a12λ−τ12⋯−a1nλ−τ1n−a21λ−τ21λ−b2−a22λ−τ22⋯−a2nλ−τ2n⋮⋮⋱⋮−an1λ−τn1−an2λ−τn2⋯λ−bn−annλ−τnn]=0. | (4.2) |
Motivated by the results in [24] and [25], it has the following important Lemmas 4.1, 4.2 and 4.3.
Lemma 4.1. Assume that bi≤0 (i=1,2,⋯,n). If det(C)<0, then there exist the time delays τij∈N (i,j=1,2,⋯,n) such that the characteristic equation (4.2) has a root λ with |λ|>1.
Proof. Let us assume that λ≠0. Consider the following equation,
λnP(λ)=λndet[λ−b1−a11λ−τ11−a12λ−τ12⋯−a1nλ−τ1n−a21λ−τ21λ−b2−a22λ−τ22⋯−a2nλ−τ2n⋮⋮⋱⋮−an1λ−τn1−an2λ−τn2⋯λ−bn−annλ−τnn]=det[λ2−b1λ−a11λ−τ11+1−a12λ−τ12+1⋯−a1nλ−τ1n+1−a21λ−τ21+1λ2−b2λ−a22λ−τ22+1⋯−a2nλ−τ2n+1⋮⋮⋱⋮−an1λ−τn1+1−an2λ−τn2+1⋯λ2−bnλ−annλ−τnn+1]. |
Let us define the function,
Fε(z)≜det[(z+ε)2−b1(z+ε)−a11(z+ε)−τ11+1⋯−a1n(z+ε)−τ1n+1−a21(z+ε)−τ21+1⋯−a2n(z+ε)−τ2n+1⋮⋱⋮−an1(z+ε)−τn1+1⋯(z+ε)2−bn(z+ε)−ann(z+ε)−τnn+1], | (4.3) |
where
τij={even number,aij<0,odd number,aij≥0(i,j=1,2,...,n). | (4.4) |
For z=reπi (r∈R+) and ε=0, (4.3) becomes
R(r)≜F0(reπi)=det[r2+b1r−|a11|r−τ11+1−|a12|r−τ12+1⋯−|a1n|r−τ1n+1−|a21|r−τ21+1r2+b2r−|a22|r−τ22+1⋯−|a2n|r−τ2n+1⋮⋮⋱⋮−|an1|r−τn1+1−|an2|r−τn2+1⋯r2+bnr−|ann|r−τnn+1]=det[r2−|b1|r−|a11|r−τ11+1−|a12|r−τ12+1⋯−|a1n|r−τ1n+1−|a21|r−τ21+1r2−|b2|r−|a22|r−τ22+1⋯−|a2n|r−τ2n+1⋮⋮⋱⋮−|an1|r−τn1+1−|an2|r−τn2+1⋯r2−|bn|r−|ann|r−τnn+1]. |
According to bi≤0, it has R(1)=detC<0. Furthermore, R(+∞)=+∞>0. Since R(r) is continuous in R, it has from Lagrangian mean value theorem that there exists r∗>1 such that R(r∗)=0. Therefore, z∗=r∗eπi is a root of F0(z)=0. Obviously, for small ε>0, |Fε(z)−F0(z)| is also small enough. By Rouché's Theorem, Fε(z)=0 has a root ^z∗(ε) in the neighborhood of z∗ such that |^z∗|>1.
Denote λ∗≜^z∗+ε. Obviously, it has |λ∗|>1 as ε is small enough. According to the selection method of τij in (4.4), it has from (4.3) that
0=Fε(^z∗)=det[(^z∗+ε)2−b1(^z∗+ε)−a11(^z∗+ε)−τ11+1⋯−a1n(^z∗+ε)−τ1n+1−a21(^z∗+ε)−τ21+1⋯−a2n(^z∗+ε)−τ2n+1⋮⋱⋮−an1(^z∗+ε)−τn1+1⋯(^z∗+ε)2−bn(^z∗+ε)−ann(^z∗+ε)−τnn+1]=det[(λ∗)2−b1λ∗−a11(λ∗)−τ11+1⋯−a1n(λ∗)−τ1n+1−a21(λ∗)−τ21+1⋯−a2n(λ∗)−τ2n+1⋮⋱⋮−an1(λ∗)−τn1+1⋯(λ∗)2−bnλ∗−ann(λ∗)−τnn+1]=(λ∗)nP(λ∗). |
This proves that λ∗ is a root of the characteristic equation (4.2) such that |z∗|>1.
Lemma 4.2. Assume that bi≤0 (i=1,2,⋯,n). If C is not an M-matrix, then there exist the time delays τij∈N (i,j=1,2,⋯,n) such that the characteristic equation (4.2) has a root λ with |λ|>1.
Proof. From Lemma 2.2, if C is not an M-matrix, then there exists at least one negative principal minor of C. Without loss of generality, it is assumed that the sequential principal minor Cl of order l of C is negative, namely, det(Cl)<0.
The characteristic equation of Cl is as follows,
Pl(λ)≜det[λ−b1−a11λ−τ11−a12λ−τ12⋯−a1lλ−τ1l−a21λ−τ21λ−b2−a22λ−τ22⋯−a2lλ−τ2l⋮⋮⋱⋮−al1λ−τl1−al2λ−τl2⋯λ−bl−allλ−τll]=0. | (4.5) |
By method similar to the proof of Lemma 4.1, there exist time delays τij∈N+ (i,j=1,2,⋯,l) which satisfy (4.4) such that the characteristic equation (4.5) has a root satisfying |λ∗|>1.
Applying Laplace expansion, P(λ) can be written as
P(λ)=Pl(λ)[n∏m=l(λ−bm−ammλ−τmm)+S(λ,λ−τij)]+T(λ,λ−τij). | (4.6) |
Observing (4.2), it is easy to see that, both S(λ,λ−τij) and T(λ,λ−τij) contain the terms aijλ−τij(i,j=l+1,⋯,n). Therefore, when |λ|>1, |S(λ,λ−τij)| and |T(λ,λ−τij)| can become sufficiently small by choosing τij (i,j=l+1,l+2,⋯,n) to be large enough. Again using Rouché's Theorem, it has that (4.2) has a root ˜λ∗ in the neighborhood of λ∗ satisfying |˜λ∗|>1.
Lemma 4.3. ([31]) If the system (4.1) has a characteristic eigenvalue λ satisfying |λ|>1, then U=0 of (2.1) is not stable.
From Lemma 4.2, Lemma 4.3 and Lemma 4.3, it has the following result.
Theorem 4.1. Assume that bi≤0 (i=1,2,⋯,n). If C is not an M-matrix, then there exist time delays τij∈N+ (i,j=1,2,⋯,n) such that U=0 of (2.1) is not stable.
Remark 4.1. Based on [2], the Ui(k) of (2.1) refers to the potential of the ith neuron at time k. The sign of Ui(k) indicates the neuron being at the action potential or the resting potential. In terms of a single neuron (ignoring the interactions among neurons), when a neuron at resting potential receives a stimulus input, the neuron's potential will be at an action potential at next time. Similarly, when neuron's stimulation at the action potential is terminated, the neuron will be resting at next time. Therefore, to a certain extent, this explains the biological significance of the condition bi≤0 (i=1,2,...,n) in Theorem 4.1.
In this paper, global attractivity and stability of the equilibrium U=0 of (2.1) have been considered. In Theorem 3.1, sufficient conditions are given for global attractivity of U=0 for any time delays τij (i,j=1,2,...,n). In Theorem 4.1, under certain conditions, necessary conditions are presented for stability of U=0 for any time delays τij (i,j=1,2,...,n).
Theorem 3.1 shows that, under certain conditions, for global attractivity of the equilibrium, the discrete Hopfield-type neural networks (2.1) can maintain consistent conditions with continuous counterparts in [25] and [28]. Meanwhile, it is also confirmed that the discrete Hopfield-type neural networks can inherit some of biological dynamic properties of the continuous ones. Furthermore, based on [1] and [2], when the discrete Hopfield-type neural networks (2.1) behaves as an associative memory or a content addressable memory, the state near the equilibrium contains information about the memory. Hence, under the conditions in Theorem 3.1, the memory is truly addressable and allows information to be stored around the equilibrium. On the other hand, according to the following numerical simulations, it can be showed that when a discrete Hopfield-type neural networks does not satisfy the conditions in Theorem 4.1, the potential (voltage) of the neurons will not reset in the resting state, but becomes oscillatory.
It should be pointed out that the condition "bi≤0 (i=1,2,...,n)" in Theorem 4.1 may be a technical condition and is likely to be further improved.
Finally, as a simple application of the conclusions of this paper, let us consider the following 3-dimensional discrete Hopfield neural networks with time delays,
{U1(k+1)=−310U1(k)−130f(U1(k−τ11))−13f(U2(k−τ12))+13f(U3(k−τ13)),U2(k+1)=b2U2(k)+13f(U1(k−τ21))+130f(U2(k−τ22))−13f(U3(k−τ23)),U3(k+1)=−310U3(k)−13f(U1(k−τ31))−13f(U2(k−τ32))−130f(U3(k−τ33)), | (5.1) |
where b2≤0, f(x)=tanh(x).
Let b2=−0.3, (5.1) satisfies the conditions of Theorem 3.1, and detC=0. Hence, U=0 is globally attractive for any time delays τij (i,j=1,2,3). Figure 1 is corresponding numerical simulation.
On the other hand, let us change the value of b2 from b2=−0.3 to b2=−0.33, it has detC=−0.01<0. According to (4.4), the time delays τij (i,j=1,2,3) are selected as follows,
τij={30,aij<0,29,aij≥0(i,j=1,2,3). |
From Theorem 4.1, it has that U=0 is unstable. Figure 2 is corresponding numerical simulation, which shows that the solution becomes oscillatory.
This work is supported by National Key R & D Program of China (2017YFF0207401) and NNSF of China (No.11471034) for W. Ma.
The author declares no conflicts of interest in this paper.
[1] | J. J. Hopfield, Neural networks and physical systems with emergent collective computational abilities, Proc. Natl. Acad. Sci., 79 (1982), 2554–2558. |
[2] | J. J. Hopfield, Neurons with graded response have collective computational properties like those of two-state neurons, Proc. Natl. Acad. Sci., 81 (1984), 3088–3092. |
[3] | J. E. Steck and S. N. Balakrishnan, Use of Hopfield neural networks in optimal guidance, IEEE Trans. Aero. Elec. Sys., 30 (1994), 287–293. |
[4] | J. Lin, K. Cheng and C. Mao, Multispectral magnetic resonance images segmentation using fuzzy Hopfield neural network, Int. J. Bio-Med. Comput., 42 (1996), 205–214. |
[5] | J. Lin, Image vector quantization using an annealed Hopfield neural network, Opt. Eng., 38 (1999), 599–605. |
[6] | J.Lin, Segmentation of medical images through a penalized fuzzy Hopfield network with moments preservation, J. Chinese Inst. Eng., 23 (2000), 633–643. |
[7] | S. Liu and J. Lin, A compensated fuzzy hopfield neural network for codebook design in vector quantization, Int. J. Patt. Recogn. Artifi. Intell., 14 (2000), 1067–1079. |
[8] | C. M. Marcus and R. M. Westervelt, Stability of analog neural networks with delay, Phys. Rev. A., 39 (1989), 347. |
[9] | J. Blair, Stability in a model of a delayed neural network, J. Dyn. Differ. Equ., 5 (1993), 607–623. |
[10] | K. Gopalsamy and X. He, Stability in asymmetric Hopfield nets with transmission delays, Physica D., 76 (1994), 344–358. |
[11] | J. Blair, S. A. Campbell and P. van den Driessche, Frustration, stability and delay-induced oscil-lation in a neural network model, SIAM J. Appl. Math., 56 (1996), 245–255. |
[12] | Y. Zhang, Global exponential stability and periodic solutions of delay Hopfield neural networks, Int. J. Sys. Sci., 27 (1996), 227–231. |
[13] | P. van den Driessche and X. Zou, Global attractivity in delayed Hopfield neural network models, SIAM J. Appl. Math., 58 (1998), 1878–1890. |
[14] | T. Chen, Global convergence of delayed dynamical systems, IEEE Trans. Neural Networks, 10 (1999), 267–271. |
[15] | J. Wei and S. Ruan, Stability and bifurcation in a neural network model with two delays, Physica D., 130 (1999), 255–272. |
[16] | T. Chu, An exponential convergence estimate for analog neural networks with delay, Phys. Lett. A., 283 (2001), 113–118. |
[17] | P. van den Driessche, J. Wu and X. Zou, Stabilization role of inhibitory self-connections in a delayed neural network, Physica D., 150 (2001), 84–90. |
[18] | T. Chu, Z. Zhang and Z. Wang, A decomposition approach to analysis of competitive-cooperative neural networks with delay, Phys. Lett. A., 312 (2003), 339–347. |
[19] | Y. Liu, Z. Wang, A. Serrano, et al., Discrete-time recurrent neural networks with time-varying delays: exponential stability analysis, Phys. Lett. A., 362 (2007), 480–488. |
[20] | S. Mohamad and K. Gopalsamy, Exponential stability of continuous-time and discrete-time cel-lular neural networks with delays, Appl. Math. Comput., 135 (2003), 17–38. |
[21] | D. H. Perkel and T. H. Bullock, Neural coding, Neurosci. Res. Symp. Summ., 6 (1968), 221–348. |
[22] | Z. Wu, H. Su, J. Chu, et al., Improved delay-dependent stability condition of discrete recurrent neural networks with time-varying delays, IEEE Trans. Neural Networks, 21 (2010), 692–697. |
[23] | J. Bai, R. Lu, A. Xue, et al., Finite-time stability analysis of discrete-time fuzzy Hopfield neural network, Neurocomputing, 159 (2015), 263–267. |
[24] | J. Hofbauer and J. So, Diagonal dominance and harmless off-diagonal delays, Proc. Amer. Math. Soc., 128 (2000), 2675–2682. |
[25] | S. A. Campbell, Delay independent stability for additive neural networks, Differ. Equat. Dynam. Systems, 9 (2001), 115–138. |
[26] | W. Ma, T. Hara and Y. Takeuchi, Stability of a 2-dimensional neural network with time delays, J. Biol. Syst., 8 (2000), 177–193. |
[27] | S. Zhang, W. Ma and Y. Kuang, Necessary and sufficient conditions for global attractivity of Hopfield-type neural networks with time delays, Rocky Mountain J. Math., 38 (2008), 1829–1840. |
[28] | W. Ma, Y. Saito and Y. Takeuchi, M-matrix structure and harmless delays in a Hopfield-type neural network, Appl. Math. Lett., 22 (2009), 1066–1070. |
[29] | J. Dong and W. Ma, Sufficient conditions for global attractivity of a class of neutral Hopfield-type neural networks, Neurocomputing, 153 (2015), 89–95. |
[30] | A. Berman and R. J. Plemmons, Nonnegative Matrices in the Mathematical Sciences, SIAM, Philadelphia, 1994. |
[31] | S. Elaydi, An Introduction to Difference Equations, 3rd edition, Springer, New York, 2005. |
1. | Lu Junxiang, Hong Xue, Adaptive synchronization for fractional stochastic neural network with delay, 2021, 2021, 1687-1847, 10.1186/s13662-020-03170-2 | |
2. | Duc Thuan Do, Thai Son Doan, Viet Cuong Le, A characterization of delay independent stability for linear off-diagonal delay difference equations, 2023, 171, 01676911, 105428, 10.1016/j.sysconle.2022.105428 | |
3. | Sourav Kumar Sasmal, Yasuhiro Takeuchi, Editorial: Mathematical Modeling to Solve the Problems in Life Sciences, 2020, 17, 1551-0018, 2967, 10.3934/mbe.2020167 | |
4. | José J. Oliveira, Global exponential stability of discrete-time Hopfield neural network models with unbounded delays, 2022, 28, 1023-6198, 725, 10.1080/10236198.2022.2073820 | |
5. | António J.G. Bento, José J. Oliveira, César M. Silva, Existence and stability of a periodic solution of a general difference equation with applications to neural networks with a delay in the leakage terms, 2023, 126, 10075704, 107429, 10.1016/j.cnsns.2023.107429 | |
6. | A. Elmwafy, José J. Oliveira, César M. Silva, Existence and exponential stability of a periodic solution of an infinite delay differential system with applications to Cohen–Grossberg neural networks, 2024, 135, 10075704, 108053, 10.1016/j.cnsns.2024.108053 |