Research article

Exponential stability of Hopfield neural networks of neutral type with multiple time-varying delays

  • Received: 01 February 2021 Accepted: 22 April 2021 Published: 21 May 2021
  • MSC : 34D20

  • This paper investigates the problem for exponential stability of Hopfield neural networks of neutral type with multiple time-varying delays. Different from the existing results, the states of the neurons involve multiple time-varying delays and time derivative of states of neurons also include multiple time-varying delays. The exponential stability of such neutral-type system has not been received enough attention since it is not easy to construct a suitable Lyapunov-Krasovskii functional to analyze the exponential stability of this type of neural system. Novel sufficient conditions of the exponential stability are established by using Lyapunov method and inequality techniques. Compared with some references, the mathematical expression of the neutral-type system is more general and the established algebraic 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.

    Citation: Li Wan, Qinghua Zhou, Hongbo Fu, Qunjiao Zhang. Exponential stability of Hopfield neural networks of neutral type with multiple time-varying delays[J]. AIMS Mathematics, 2021, 6(8): 8030-8043. doi: 10.3934/math.2021466

    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] 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
    [7] 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
    [8] 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
    [9] 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
    [10] Nina Huo, Bing Li, Yongkun Li . Global exponential stability and existence of almost periodic solutions in distribution for Clifford-valued stochastic high-order Hopfield neural networks with time-varying delays. AIMS Mathematics, 2022, 7(3): 3653-3679. doi: 10.3934/math.2022202
  • This paper investigates the problem for exponential stability of Hopfield neural networks of neutral type with multiple time-varying delays. Different from the existing results, the states of the neurons involve multiple time-varying delays and time derivative of states of neurons also include multiple time-varying delays. The exponential stability of such neutral-type system has not been received enough attention since it is not easy to construct a suitable Lyapunov-Krasovskii functional to analyze the exponential stability of this type of neural system. Novel sufficient conditions of the exponential stability are established by using Lyapunov method and inequality techniques. Compared with some references, the mathematical expression of the neutral-type system is more general and the established algebraic 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.



    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

    ˙xi(t)=cixi(t)+nj=1aijfj(xj(t))+nj=1bijgj(xj(tτij(t)))+nj=1eij˙xj(tξij(t))+ui,t0; (1.1)
    xi(t)=φi(t),˙xi(t)=ϕi(t),t[max{τ,ξ},0],i=1,,n,

    where ξ,τ and ci are positive numbers, φi(t) and ϕi(t) are continuous functions, ξij(t) and τij(t) are delay functions, fj() and gj() are nonlinear continuous activation functions. These functions satisfy that for every x,yR, t0 and i,j=1,,n,

    0ξij(t)ξ,0τij(t)τ,˙ξij(t)¯ξ,˙τij(t)¯τ; (1.2)
    |fi(x)fi(y)|li|xy|,|gi(x)gi(y)|mi|xy|, (1.3)

    where ¯ξ,¯τ,li and mi are some positive numbers. From [37,38], we know that x(t)=(x1(t),,xn(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 τij(t)=τj(t) and ξij(t)=ξj(t), system (1.1) transforms into the following vector-matrix form studied [18,20]:

    ˙x(t)=Cx(t)+Af(x(t))+Bg(x(tτ(t)))+E˙x(tξ(t))+u,t0, (1.4)

    whereA=(aij)n×n,B=(bij)n×n,C=diag(c1,,cn),u=(u1,,un)T,

    x(t)=(x1(t),,xn(t))T,˙x(tξ(t))=(˙x1(tξ1(t)),,˙xn(tξn(t)))T,
    f(x(t))=(f1(x1(t)),,fn(xn(t)))T,g(x(tτ(t)))=(g1(x1(tτ1(t))),,gn(xn(tτn(t))))T.

    When τij(t)=τij,fj=gj and ξij(t)=ξij (or ξj), system (1.1) transforms into the following system studied in [30]:

    ˙xi(t)=cixi(t)+nj=1aijfj(xj(t))+nj=1bijfj(xj(tτij))+nj=1eij˙xj(tξij)+ui, (1.5)

    or the following system studied in [29]:

    ˙xi(t)=cixi(t)+nj=1aijfj(xj(t))+nj=1bijfj(xj(tτij))+nj=1eij˙xj(tξj)+ui. (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 τij(t) and ξij(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.

    From [18,20], we know that system (1.1) has at least one equilibrium point x=(x1,,xn)T under the conditions (1.2) and (1.3). By employing the formula yi(t)=xi(t)xi(i=1,,n), system (1.1) can be transformed into the following equivalent system

    ˙yi(t)=ciyi(t)+nj=1aij˜fj(yj(t))+nj=1bij˜gj(yj(tτij(t)))+nj=1eij˙yj(tξij(t)), (2.1)

    where

    ˜fj(yj(t))=fj(yj(t)+xj)fj(xj),˜gj(yj(tτij(t)))=gj(yj(tτij(t))+xj)gj(xj).

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

    ˙yi(t)=ciyi(t)+nj=1aij˜fj(yj(t))+nj=1bij˜fj(yj(tτij))+nj=1eij˙yj(tξij). (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 γ,p1,p2,,pn such that γ<1,max{ˉξ,ˉτ}<1γ,

    γpicinj=1pj(li|aji|γ+mi|bji|)>0,piγnj=1pj|eji|>0,i=1,2,,n.

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

    Proof. Let

    h1(z)=1eξz¯ξeξzγ,h2(z)=1eτz¯τγ,
    h3(z)=γpicizpiγznj=1pj|eji|nj=1pj(li|aji|γ+mi|bji|eτz).

    Then, ˙hi(z)<0,hi(+)<0,i=1,2,3, and

    h1(0)=1¯ξγ>0,h2(0)=1¯τγ>0,
    h3(0)=γpicinj=1pj(li|aji|γ+mi|bji|)>0.

    Therefore, there exist some positive numbers λ1,λ2 and λ3 such that h1(λ1)=h2(λ2)=h3(λ3)=0, which implies that there must exist a positive number λ(0,min{λ1,λ2,λ3}) such that

    1eλξ¯ξeλξγ>0,1eλτ¯τγ>0, (2.3)
    γpiciλpiγλnj=1pj|eji|nj=1pj(li|aji|γ+mi|bji|eλτ)>0. (2.4)

    We construct the following Lyapunov-Krasovskii functional [40]

    V(t)=eλ(t+ξ)ni=1[piγnj=1pj|eji|sgn(yi(t))sgn(˙yi(t))]|yi(t)|+ni=1nj=1pi|eij|ttξij(t)eλ(s+ξ)|˙yj(s)|ds+eλξni=1nj=1pi|bij|mjttτij(t)eλ(s+τ)|yj(s)|ds, (2.5)

    and derive

    min1in{piγnj=1pj|eji|}eλ(t+ξ)y(t)1eλ(t+ξ)ni=1[piγnj=1pj|eji|]|yi(t)|eλ(t+ξ)ni=1[piγnj=1pj|eji|sgn(yi(t))sgn(˙yi(t))]|yi(t)|V(t), (2.6)
    V(0)=eλξni=1[piγnj=1pj|eji|sgn(yi(0))sgn(˙yi(0))]|yi(0)|+eλξni=1nj=1pi|eij|0ξij(0)eλs|˙yj(s)|ds+eλ(ξ+τ)ni=1nj=1pi|bij|mj0τij(0)eλs|yj(s)|dseλξmax1in{piγ+nj=1pj|eji|}y(0)1+eλξni=1nj=1pi|eij|0ξ|˙yj(s)|ds+eλ(ξ+τ)ni=1nj=1pi|bij|mj0τ|yj(s)|ds. (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

    ˙V(t)=λeλ(t+ξ)ni=1[piγnj=1pj|eji|sgn(yi(t))sgn(˙yi(t))]|yi(t)|+eλtni=1eλξ[piγnj=1pj|eji|sgn(yi(t))sgn(˙yi(t))]sgn(yi(t))˙yi(t)+eλ(t+ξ)ni=1nj=1pi|eij||˙yj(t)|ni=1nj=1(1˙ξij(t))eλ(tξij(t)+ξ)pi|eij||˙yj(tξij(t))|+eλξni=1nj=1pi|bij|mj(eλ(t+τ)|yj(t)|(1˙τij(t))eλ(tτij(t)+τ)|yj(tτij(t))|)λeλ(t+ξ)ni=1(piγ+nj=1pj|eji|)|yi(t)|+eλtni=1{eλξpiγsgn(yi(t))˙yi(t)eλξnj=1pj|eji|(sgn(yi(t)))2|˙yi(t)|+eλξnj=1pj|eji||˙yi(t)|}+ni=1nj=1(˙ξij(t)eλ(tξij(t)+ξ)eλ(tξij(t)+ξ))pi|eij||˙yj(tξij(t))|+eλξni=1nj=1pi|bij|mj(eλ(t+τ)|yj(t)|+[˙τij(t)eλ(tτij(t)+τ)eλ(tτij(t)+τ)]|yj(tτij(t))|)λeλ(t+ξ)ni=1(piγ+nj=1pj|eji|)|yi(t)|+eλtni=1{eλξpiγsgn(yi(t))˙yi(t)eλξnj=1pj|eji|(sgn(yi(t)))2|˙yi(t)|+eλξnj=1pj|eji||˙yi(t)|}+ni=1nj=1(ˉξeλ(t+ξ)eλt)pi|eij||˙yj(tξij(t))|+eλξni=1nj=1pi|bij|mj(eλ(t+τ)|yj(t)|+(ˉτeλ(t+τ)eλt)|yj(tτij(t))|)=λeλ(t+ξ)ni=1(piγ+nj=1pj|eji|)|yi(t)|+eλtni=1{eλξpiγsgn(yi(t))˙yi(t)eλξnj=1pj|eji|(sgn(yi(t)))2|˙yi(t)|+eλξnj=1pj|eji||˙yi(t)|}+eλt(¯ξeλξ1)ni=1nj=1pi|eij||˙yj(tξij(t))|+eλξni=1nj=1pi|bij|mj(eλ(t+τ)|yj(t)|+(¯τeλτ1)eλt|yj(tτij(t))|). (2.8)

    It is noted that for yi(t)0,

    eλξpiγsgn(yi(t))˙yi(t)eλξnj=1pj|eji|(sgn(yi(t)))2|˙yi(t)|+eλξnj=1pj|eji||˙yi(t)|=eλξpiγsgn(yi(t))˙yi(t)=eλξ{γpisgn(yi(t))ciyi(t)+γpisgn(yi(t))nj=1aij˜fj(yj(t))+γpisgn(yi(t))nj=1bij˜gj(yj(tτij(t)))+γpisgn(yi(t))nj=1eij˙yj(tξij(t))}eλξ{γpisgn(yi(t))ciyi(t)+|γpisgn(yi(t))nj=1aij˜fj(yj(t))|+|γpisgn(yi(t))nj=1bij˜gj(yj(tτij(t)))|+|γpisgn(yi(t))nj=1eij˙yj(tξij(t))|}eλξ{γpici|yi(t)|+γpinj=1|aij|lj|yj(t)|+γpinj=1|bij|mj|yj(tτij(t))|}+eλξγpinj=1|eij||˙yj(tξij(t))|,

    and for yi(t)=0,

    eλξpiγsgn(yi(t))˙yi(t)eλξnj=1pj|eji|(sgn(yi(t)))2|˙yi(t)|+eλξnj=1pj|eji||˙yi(t)|=eλξnj=1pj|eji||˙yi(t)|=eλξ{nj=1pj|eji|sgn(˙yi(t))ciyi(t)+nj=1pj|eji|sgn(˙yi(t))nj=1aij˜fj(yj(t))+nj=1pj|eji|sgn(˙yi(t))nj=1bij˜gj(yj(tτij(t)))+nj=1pj|eji|sgn(˙yi(t))nj=1eij˙yj(tξij(t))}eλξ{γpici|yi(t)|+|nj=1pj|eji|sgn(˙yi(t))nj=1aij˜fj(yj(t))|+|nj=1pj|eji|sgn(˙yi(t))nj=1bij˜gj(yj(tτij(t)))|+|nj=1pj|eji|sgn(˙yi(t))nj=1eij˙yj(tξij(t))|}eλξ{γpici|yi(t)|+nj=1pj|eji|nj=1|aij|lj|yj(t)|+nj=1pj|eji|nj=1|bij|mj|yj(tτij(t))|}+eλξnj=1pj|eji|nj=1|eij||˙yj(tξij(t))|eλξ{γpici|yi(t)|+γpinj=1|aij|lj|yj(t)|+γpinj=1|bij|mj|yj(tτij(t))|}+eλξγpinj=1|eij||˙yj(tξij(t))|,

    where we use nj=1pj|eji|<piγ, and nj=1pj|eji|sgn(˙yi(t))ciyi(t)=γpici|yi(t)|=0 when yi(t)=0.

    Therefore, for every yi(t)R, we have

    eλξpiγsgn(yi(t))˙yi(t)eλξnj=1pj|eji|(sgn(yi(t)))2|˙yi(t)|+eλξnj=1pj|eji||˙yi(t)|eλξ{γpici|yi(t)|+γpinj=1|aij|lj|yj(t)|+γpinj=1|bij|mj|yj(tτij(t))|}+eλξγpinj=1|eij||˙yj(tξij(t))|. (2.9)

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

    ˙V(t)λeλ(t+ξ)ni=1(piγ+nj=1pj|eji|)|yi(t)|+eλtni=1{eλξ(γpici|yi(t)|+γpinj=1|aij|lj|yj(t)|+γpinj=1|bij|mj|yj(tτij(t))|)+(eλξγ+eλξ¯ξ1)pinj=1|eij||˙yj(tξij(t))|}+eλξni=1nj=1pi|bij|mj(eλ(t+τ)|yj(t)|+(¯τeλτ1)eλt|yj(tτij(t))|)=eλ(t+ξ)ni=1(γpiciλpiγλnj=1pj|eji|nj=1pj(li|aji|γ+mi|bji|eλτ))|yi(t)|eλt(1eλξ¯ξeλξγ)ni=1nj=1pi|eij||˙yj(tξij(t))|(1eλτ¯τγ)eλ(t+ξ)ni=1nj=1pi|bij|mj|yj(tτij(t))|0. (2.10)

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

    y(t)1δeλt(φ1+ϕ1),t0,

    where

    φ1=supt[max{τ,ξ},0]φ(t)1,ϕ1=supt[max{τ,ξ},0]ϕ(t)1.

    Generally speaking, it is not easy to find the values of the positive constants p1,,pn. Therefore, it is necessary to give a result without involving the constants p1,,pn. The following result is a special case of Theorem 1 for p1==pn, which is easier to validate.

    Theorem 2. Suppose that there exists a positive number γ such that

    max{ˉξ,ˉτ}<1γ,nj=1|eji|<γ<1,γcinj=1(li|aji|γ+mi|bji|)>0,i=1,2,,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 γ,p1,p2,,pn such that γ<1,

    γpicinj=1pjli(|aji|γ+|bji|)>0,piγnj=1pj|eji|>0,i=1,2,,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 γ such that γ<1,

    γcinj=1li(|aji|γ+|bji|)>0,γnj=1|eji|>0,i=1,2,,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 γpicinj=1pjli(|aji|γ+|bji|)>0 and 0<γ<1 can deduce that picinj=1pjli(|aji|+|bji|)>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:

    αi=c2il2inj=1|nk=1akiakj|l2inj=1nk=1(|aji||bjk|+|bji||ajk|+|bji||bjk|)l2inj=1nk=1cj|ejk|(|aji|+|bji|)nj=1nk=1cj|eji|(|ajk|+|bjk|)nj=1c2i|eij|nj=1c2j|eji|>0,

    and 1nj=1|eji|>0,i=1,,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]:

    σi=2cinj=1(lj|aij|+li|aji|)nj=1(lj|bij|+li|bji|)nj=1(|eij|+|eji|)nj=1nk=1(li|aki||ekj|+li|bki||ekj|)nj=1nk=1(lk|ajk||eji|+lk|bjk||eji|)>0,

    and 1nj=1|eji|>0,i=1,,n.

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

    ϵi=c2inj=1|nk=1akiakj|nj=1nk=1(|aji||bjk|+|aji||ejk|+|bji||ajk|+|bji||ejk|+|bji||bjk|)>0,
    εij=1nnk=1(|eji||ejk|+|ajk||eji|+|bjk||eji|)>0,i,j=1,,n.

    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=(1111111111111111),B=(1111111111111111),E=(0.10.10.10.10.10.10.10.10.10.10.10.10.10.10.10.1),

    c1=c2=c3=9,c4=8,fi(x)=tanh(x),gi(x)=0.5tanh(x),ξii(t)=0.1sint+0.1,τii(t)=0.1cost+0.1;ξij(t)=0.1cost+0.1,τij(t)=0.1sint+0.1,ij;i,j=1,2,3,4.

    We calculate ˉξ=ˉτ=0.1,li=1,mi=0.5,4j=1|eji|=0.4 and

    ciγ4j=1(li|aji|γ+mi|bji|)={5γ2,i=1,2,3;4γ2,i=4.

    It is clear that when γ(0.5,0.9), Theorem 2 holds.

    Now we choose γ=0.424 and require that

    γpici4j=1pj(li|aji|γ+mi|bji|)=0.01,i=1,2,3,4.

    Then, we calculate p1=p2=p3=2.2222,p4=2.5000, and

    piγ4j=1pj|eji|=0.424pi0.14j=1pj={0.025552,i=1,2,3;0.14334,i=4.

    Thus, all conditions of Theorem 1 are satisfied.

    If we choose γ=0.423 and still require that

    γpici4j=1pj(li|aji|γ+mi|bji|)=0.01,i=1,2,3,4.

    Then pi<0,i=1,2,3,4. Therefore, when γ[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: c1=c2=c3=9,c4=8.5,fi(x)=tanh(x),ξii=0.2,τii=0.1;ξij=0.3,τij=0.4,ij;i,j=1,2,3,4, the matrices A,B,E are the same as in Example 1.

    Then, we calculate li=1,4j=1|eji|=0.4 and

    ciγ4j=1li(|aji|γ+|bji|)={5γ4,i=1,2,3;4.5γ4,i=4.

    It is clear that when γ(8/9,1), Corollary 2 holds.

    Now we choose γ=0.85 and require that

    γpici4j=1pjli(|aji|γ+|bji|)=0.01,i=1,2,3,4.

    Then, we calculate p1=p2=p3=0.0708,p4=0.0750, and

    piγ4j=1pj|eji|=0.85pi0.14j=1pj={0.03144,i=1,2,3;0.03501,i=4.

    Thus, all conditions of Corollary 1 are satisfied.

    On the other hand, we calculate

    α1=814j=1|4k=1ak1akj|3×4×435.5×4×0.235.5×4×0.281×0.4315.25×0.1<0,

    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 li=1,4j=1|eji|=0.4,

    εij=140.21×4=0.59<0,
    σ1=2×92×42×40.2×40.2×420.2×42=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).

    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.

    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).

    All authors declare no conflicts of interest in this paper.



    [1] J. J. Hopfield, Neural networks and physical systems with emergent collect computational abilities, PNAS, 79 (1982), 2254–2558.
    [2] J. J. Hopfield, Neurons with graded response have collective computational properties like those of two-state neurons, PNAS, 81 (1984), 3088–3092. doi: 10.1073/pnas.81.10.3088
    [3] P. N. Suganthan, E. K. Teoh, D. P. Mital, Pattern recognition by homomorphic graph matching using Hopfield neural networks, Image Vision Comput., 13 (1995), 45–60. doi: 10.1016/0262-8856(95)91467-R
    [4] T. Deb, A. K. Ghosh, A. Mukherjee, Singular value decomposition applied to associative memory of Hopfield neural network, Materialstoday: Proc., 5 (2018), 2222–2228.
    [5] M. P. Kennedy, L. O. Chua, Neural networks for nonlinear programming, IEEE T. Circuits Syst., 35 (1988), 554–562. doi: 10.1109/31.1783
    [6] V. Donskoy, BOMD: building optimization models from data (neural networks based approach), Quant. Finance Econ., 3 (2019), 608–623. doi: 10.3934/QFE.2019.4.608
    [7] B. Y. Zhang, J. Lam, S. Y. Xu, Stability analysis of distributed delay neural networks based on relaxed Lyapunov-Krasovskii functionals, IEEE Trans. Neural Netw. Learn. Syst., 26 (2015), 1480–1492. doi: 10.1109/TNNLS.2014.2347290
    [8] X. D. Li, D. O'Regan, H. Akca, Global exponential stabilization of impulsive neural networks with unbounded continuously distributed delays, IMA J. Appl. Math., 80 (2015), 85–99. doi: 10.1093/imamat/hxt027
    [9] Q. K. Song, H. Yan, Z. J. Zhao, Y. R. Liu, Global exponential stability of complex-valued neural networks with both time-varying delays and impulsive effects, Neural Networks, 79 (2016), 108–116. doi: 10.1016/j.neunet.2016.03.007
    [10] B. Song, Y. Zhang, Z. Shu, F. N. Hu, Stability analysis of Hopfield neural networks perturbed by Poisson noises, Neurocomputing, 196 (2016), 53–58. doi: 10.1016/j.neucom.2016.02.034
    [11] N. Cui, H. J. Jiang, C. Hu, A. Abdurahman, Global asymptotic and robust stability of inertial neural networks with proportional delays, Neurocomputing, 272 (2018), 326–333. doi: 10.1016/j.neucom.2017.07.001
    [12] X. Y. Yang, X. D. Li, Q. Xi, P. Y. Duan, Review of stability and stabilization for impulsive delayed systems, Math. Biosci. Eng., 15 (2018), 1495–1515. doi: 10.3934/mbe.2018069
    [13] Q. Yao, L. S. Wang, Y. F. Wang, Existence-uniqueness and stability of reaction-diffusion stochastic Hopfield neural networks with S-type distributed time delays, Neurocomputing, 275 (2018), 470–477. doi: 10.1016/j.neucom.2017.08.060
    [14] J. T. Hu, G. X. Sui, X. X. Lu, X. D. Li, Fixed-time control of delayed neural networks with impulsive perturbations, Nonlinear Anal.-Model, 23 (2018), 904–920. doi: 10.15388/NA.2018.6.6
    [15] F. X. Wang, X. G. Liu, M. L. Tang, L. F. Chen, Further results on stability and synchronization of fractional-order Hopfield neural networks, Neurocomputing, 346 (2019), 12–19. doi: 10.1016/j.neucom.2018.08.089
    [16] W. Q. Shen, X. Zhang, Y. T. Wang, Stability analysis of high order neural networks with proportional delays, Neurocomputing, 372 (2020), 33–39. doi: 10.1016/j.neucom.2019.09.019
    [17] H. M. Wang, G. L. Wei, S. P. Wen, T. W. Huang, Generalized norm for existence, uniqueness and stability of Hopfield neural networks with discrete and distributed delays, Neural Networks, 128 (2020), 288–293. doi: 10.1016/j.neunet.2020.05.014
    [18] S. M. Lee, O. M. Kwon, J. H. Park, A novel delay-dependent criterion for delayed neural networks of neutral type, Phys. Lett. A, 374 (2010), 1843–1848. doi: 10.1016/j.physleta.2010.02.043
    [19] P. L. Liu, Improved delay-dependent stability of neutral type neural networks with distributed delays, ISA Trans., 52 (2013), 717–724. doi: 10.1016/j.isatra.2013.06.012
    [20] S. Arik, An analysis of stability of neutral-type neural systems with constant time delays, J. Franklin I., 351 (2014), 4949–4959. doi: 10.1016/j.jfranklin.2014.08.013
    [21] S. Dharani, R. Rakkiyappan, J. D. Cao, New delay-dependent stability criteria for switched Hopfield neural networks of neutral type with additive time-varying delay components, Neurocomputing, 151 (2015), 827–834. doi: 10.1016/j.neucom.2014.10.014
    [22] K. B. Shi, S. M. Zhong, H. Zhu, X. Z. Liu, Y. Zeng, New delay-dependent stability criteria for neutral-type neural networks with mixed random time-varying delays, Neurocomputing, 168 (2015), 896–907. doi: 10.1016/j.neucom.2015.05.035
    [23] K. B. Shi, H. Zhu, S. M. Zhong, Y. Zeng, Y. P. Zhang, New stability analysis for neutral type neural networks with discrete and distributed delays using a multiple integral approach, J. Franklin. I., 352 (2015), 155–176. doi: 10.1016/j.jfranklin.2014.10.005
    [24] R. Samidurai, S. Rajavel, R. Sriraman, J. D. Cao, A. Alsaedi, F. E. Alsaadi, Novel results on stability analysis of neutral-type neural networks with additive time-varying delay components and leakage delay, Int. J. Control Autom. Syst., 15 (2017), 1888–1900. doi: 10.1007/s12555-016-9483-1
    [25] M. W. Zheng, L. X. Li, H. P. Peng, J. H. Xiao, Y. X. Yang, H. Zhao, Finite-time stability analysis for neutral-type neural networks with hybrid time-varying delays without using Lyapunov method, Neurocomputing, 238 (2017), 67–75. doi: 10.1016/j.neucom.2017.01.037
    [26] G. B. Zhang, T. Wang, T. Li, S. M. Fei, Multiple integral Lyapunov approach to mixed-delay-dependent stability of neutral neural networks, Neurocomputing, 275 (2018), 1782–1792. doi: 10.1016/j.neucom.2017.10.021
    [27] G. M. Zhuang, S. Y. Xu, J. W. Xia, Q. Ma, Z. Q. Zhang, Non-fragile delay feedback control for neutral stochastic Markovian jump systems with time-varying delays, Appl. Math. Comput., 355 (2019), 21–32.
    [28] G. M. Zhuang, J. W. Xia, J. E. Feng, B. Y. Zhang, J. W. Lu, Z. Wang, Admissibility analysis and stabilization for neutral descriptor hybrid systems with time-varying delays, Nonlinear Anal.-Hybrid Syst., 33 (2019), 311–321. doi: 10.1016/j.nahs.2019.03.009
    [29] S. Arik, A modified Lyapunov functional with application to stability of neutral-type neural networks with time delays, J. Franklin I., 356 (2019), 276–291. doi: 10.1016/j.jfranklin.2018.11.002
    [30] O. Faydasicok, A new Lyapunov functional for stability analysis of neutral-type Hopfield neural networks with multiple delays, Neural Netw., 129 (2020), 288–297. doi: 10.1016/j.neunet.2020.06.013
    [31] N. Ozcan, Stability analysis of Cohen-Grossberg neural networks of neutral-type: Multiple delays case, Neural Netw., 113 (2019), 20–27. doi: 10.1016/j.neunet.2019.01.017
    [32] O. Faydasicok, New criteria for global stability of neutral-type Cohen- Grossberg neural networks with multiple delays, Neural Netw., 125 (2020), 330–337. doi: 10.1016/j.neunet.2020.02.020
    [33] Y. K. Deng, C. X. Huang, J. D. Cao, New results on dynamics of neutral type HCNNs with proportional delays, Math. Comput. Simulat, 187 (2021), 51–59. doi: 10.1016/j.matcom.2021.02.001
    [34] S. I. Niculescu, Delay effects on stability: A robust control approach, Berlin: Springer, 2001.
    [35] V. B. Kolmanovskii, V. R. Nosov, Stability of functional differential equations, London: Academic Press, 1986.
    [36] Y. Kuang, Delay differential equations with applications in population dynamics, Boston: Academic Press, 1993.
    [37] J. K. Hale, S. M. Verduyn Lunel, Introduction to functional differential equations, New York: Springer, 1993.
    [38] G. Michael, Stability of neutral functional differential equations, Paris: Atlantis Press, 2014.
    [39] Y. He, M. Wu, J. H. She, Delay-dependent exponential stability of delayed neural networks with time-varying delay, IEEE T. Circuits Syst-II, 53 (2006), 553–557. doi: 10.1109/TCSII.2006.876385
    [40] N. N. Krasovskii, Stability of Motion, Stanford: Stanford University Press, 1963.
  • 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
  • Reader Comments
  • © 2021 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

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2965) PDF downloads(166) Cited by(10)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog