Abstract
This article is concerned with shunting inhibitory cellular neural networks involving neutral type proportional delays and D operators. Applying Lyapunov functional method and differential inequality techniques, we employ a novel argument to establish some new criteria on the existence and generalized exponential stability of almost periodic solutions of the addressed systems. In addition, numerical simulations are carried out to verify our theoretical findings.
Similar content being viewed by others
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.Avoid common mistakes on your manuscript.
1 Introduction
In implementation of neural networks (NNs), the finite switching speed of neurons and amplifiers unavoidably produce time delays, which may lead to instability and oscillation in neural networks. As such, it is of great significance to investigate the mathematical properties of NNs with neutral type delays. Usually, neutral type neural networks can be described as non-operator-based neutral functional differential equations (NFDEs) ([1,2,3,4,5,6,7,8,9]) or D-operator-based NFDEs ([10,11,12,13,14]). In particular, based on the complexity of the dynamical behaviors for functional differential equations under the effect of time delays, neutral type cellular neural networks with D operator have more realistic significance than non-operator-based ones in many practical applications of neural networks dynamics. From above results, the readers can observe that most of the considered neutral type shunting inhibitory cellular neural networks (SICNNs) only touched non-operator-based NFDEs, and the dynamical behaviors of neutral type SICNNs with D operator have not been deeply studied (see [15, 16]).
In the past two decades, proportional delays occurring in nonlinear dynamic have gotten more and more attention because of its wide application in a variety of areas including probability theory on algebraic structures, electrodynamics and electric locomotive [17,18,19]. Moreover, many biological and cognitive activities (e.g., heartbeat, mastication, memorization) require repetition. For example, in many electronic circuits, oscillators are essential components and usually generate almost periodical signals (sinusoids, square waves, etc.). Consequently, almost periodic oscillations play important roles in characterizing the behaviours of nonlinear dynamiccal systems [20,21,22,23,24]. Consequently, some interesting results have been published on the existence and stability of almost periodic solution for cellular neural networks (CNNs) with proportional delays [25, 26]. However, to the best of our knowledge, no such work has been carried out on SICNNs with neutral type proportional delays and D operators.
Motivated by the above analysis, the main goal of this paper is to study the existence and generalized exponential stability of almost periodic solutions for a class of SICNNs with neutral type proportional delays and D operators, which can be described as the following NFDEs:
where \(ij\in J=\{11, 12, \ldots , 1n, \ldots , m1, m2, \ldots , mn \},\) mn corresponds to the number of units in a neural network, \(C_{ij}\) is the cell at the (i, j) position of the lattice, \( N_r (i, j) = \Big \{C_{kl}: \max (| k - i |, | l - j |)\le r, 1 \le k \le m, 1 \le l \le n\Big \} \) is the r neighborhood of \( C_{ij}\), \( x_{ij}(t )\) designates the ijth neuron state, \(a_{ij}(t) \) represents the decay rate, f is the activation function, proportional delay factors \(q _{ij} \) and \( r_{ij} \) satisfy \(0< q _{kl}, r_{ij} < 1\), \(r _{ij } t=t -(1-r _{ij })t\) and \(q _{kl} t=t -(1-q _{kl})t\), in which \((1-r _{ij })t\) and \( (1-q _{kl})t\) are the transmission delay functions. The detailed biological description on the input \(L_{ij}(t)\) and the coefficient \( C_{ij}^{kl}(t)\) can be found in [15, 16].
The initial value conditions of SICNNs (1.1) are of the form
where \(C([ \rho _{ij}, \ 1], {\mathbb {R}} )\) denotes the set of continuous functions on \([ \rho _{ij}, \ 1]\), and \(\rho _{ij} =\min \{r_{ij}, \ q_{ ij} \}\).
Contributions of the paper are summarized as follows. Firstly, basic definitions and hypotheses are imposed. By the method-of-steps, we derive that every solution of (1.1) with initial condition (1.2) exists and is unique on \([t_{0}, +\infty )\). Secondly, based on some novel differential inequality techniques, the existence and generalized exponential stability of almost periodic solutions for (1.1) will be subsequently established. Finally, the effectiveness of the obtained result is illustrated by a numerical example.
2 Preliminary Results
Throughout this paper, the following concepts and notations will be adopted:
Definition 2.1
(see [20, 21]). \(u(t)\in C({\mathbb {R}},{\mathbb {R}}^{mn})\) is said to be almost periodic on \({\mathbb {R}}\) if, for any \( \varepsilon >0\), it is possible to find a real number \(l=l(\varepsilon )>0 \) with the property that, for any interval with length \(l(\varepsilon )\), there exists a number \(\delta =\delta (\varepsilon )\) in this interval such that \(\Vert u(t+\delta )-u(t)\Vert <\varepsilon \ \text{ for } \text{ all } \ t\in {\mathbb {R}}.\)
We designate by \(AP({\mathbb {R}},\mathbb {R}^{mn})\) the set of the almost periodic functions from \({\mathbb {R}}\) to \(\mathbb {R}^{mn}\). Then \((AP({\mathbb {R}},\mathbb {R}^{mn}), \Vert \cdot \Vert _{\infty })\) is a Banach space, where \(\Vert \cdot \Vert _{\infty }\) denotes the supremum norm \(\Vert f\Vert _{\infty } := \sup \limits _{ t\in {\mathbb {R}}} \Vert f (t)\Vert \) (see [20, 21]). For all \( ij, kl \in J\), we suppose that the parameters of (1.1) and activation functions \( p_{ij}, \ a_{ij}, \ C^{kl}_{ij}, L_{ij } \in AP({\mathbb {R}},\mathbb {R})\) satisfy the following assumptions:
\((T_0)\) \(M[a_{ij}]=\lim \limits _{T\rightarrow +\infty }\frac{1}{T}\int _{t}^{t+T}a_{ij}(s)ds>0 \ \text{ for } \text{ all } t\in {\mathbb {R}} ,\) and there exist a bounded and continuous function \({\tilde{a}}_{ij} :{\mathbb {R}}\rightarrow (0, \ +\infty )\) and a positive constant \(K_{ij} \) such that
\((T_{1})\) \( \text{ for } \text{ all } \ u , v \in {\mathbb {R}}\), there exist constants \(M ^{f} \) and \(L ^{f} \) such that
\((T_{2})\) there exist positive constants L and \(\kappa \) such that
and
Lemma 2.1
(see [26, Lemma 2.2]). Let \(x(t)\in AP({\mathbb {R}}, \mathbb {R} ) \) and \(q\in {\mathbb {R}}\) be a constant. Then, \(x(q t)\in AP({\mathbb {R}}, \mathbb {R} )\).
Lemma 2.2
Suppose that \((T_0)\), \((T_1)\) and \((T_2)\) are satisfied. Then, every solution of (1.1) with initial condition (1.2) exists and is unique on \([t_{0}, +\infty )\).
Proof
Firstly, we show that x(t) exists and is unique on \([t_{0},\ t_{0}r]\). In fact, for \( ij\in J \), let
and
Then
From \((T_{1})\), we can easily show that the solution y(t) of ordinary differential equations (2.1) with initial value \(y(t_{0})=\{x _{ij} (t_{0})-\beta _{ij}(t_{0})\}\) exists and is unique on \([t_{0},\ t_{0}r]\). Hence, \(x (t )=y(t)+\{\beta _{ij}(t)\}\) exists and is unique on \([t_{0},\ t_{0}r]\). Similarly, one can prove that \(x (t )=y(t)+\{\beta _{ij}(t)\}\) exists and is unique on \([t_{0}r,\ t_{0}r^{2}]\). Consequently, the Lemma 2.2 follows from the induction. \(\square \)
3 Main Results
Theorem 3.1
Under the assumptions of Lemma 2.2, system (1.1) has a unique almost periodic solution \(x^{*}(t)\). Moreover, \(x^{*}(t)\) is generalized exponential stability, i.e., there exists a positive constant \(\sigma \), which is independent of solutions of (1.1), such that for arbitrary solution x(t) of (1.1),
Proof
Let
Then
Let \( \varphi \in AP({\mathbb {R}},\mathbb {R}^{mn})\). In view of Lemma 2.1, we can show
It follows from the uniform continuity of f that
Now, we consider the following auxiliary equations:
According to Lemma 1.1 in [22] and the fact that \( M[a_{ij}]>0 (ij\in J) \), one can see that the system (3.2) has exactly one almost periodic solution:
Obviously,
Likewise, we can prove \(Q(t)=\{Q_{ij}(t)\}=\{ \int _{-\infty }^{t}e^{-\int _{s}^{t}a_{ij}(u)du}L_{ij}(s)ds\}\) is the unique almost periodic solution of the following differential equations:
and \(\{Q_{ij}(t)\} \in AP({\mathbb {R}},\mathbb {R}^{mn})\), \(L= \Vert Q\Vert _{\infty }.\)
Set
If \(\varphi \in \Omega \) , then
Moreover, we define a mapping \(T:\Omega \rightarrow \Omega \) by setting
We next prove that for any \(\varphi \in \Omega , \ T \varphi \in \Omega \). Clearly, it follows from (3.3), (3.4), (\(T_{0}\)), (\(T_{1}\)) and (\(T_{2}\)) that
which entails that \( T \varphi \in \Omega \).
Furthermore, we show that T is a contract operator. In fact, again from (3.3), (3.4), (\(T_{0}\)), (\(T_{1}\)) and (\(T_{2}\)), we obtain
which yields
Therefore, Theorem 0.3.1 of [27] and the fact that \( \max \limits _{ ij\in J}( p_{ij} ^{+} + F_{ij})<1\) entail that T possesses a unique fixed point \(x^{* }=\{x_{ij}^{* }(t)\}\in \Omega \) such that
and
which, together with (3.3) leads to
and \(x^{* }(t)\) is an almost periodic solution of system (1.1).
Finally, we demonstrate the global generalized exponential stability of \(x^{*}(t)\). Let \( x(t)=\{x_{ij}(t)\} \) be an arbitrary solution of (1.1) with initial value \( \phi (t)=\{\phi _{ij}(t)\} \) satisfying (1.2), and
Then
From \((T_{2})\), we can choose constant \(\sigma \in (0, \ \min \{ \beta , \ \min \limits _{ij\in J}{\tilde{a}}_{ij} ^{-}\}) \) such that \( e^{ \sigma \ln \left( \frac{1 }{ r_{ij} }\right) } p_{ij} ^{+}<1\) and
In view of the facts that
(3.6) implies that
We trivially extend x(t) to \([r_{ij}\rho _{ij}t_{0}, \ +\infty )\) by letting \(x_{ij}(t) =\varphi _{ij}(t) = \varphi _{ij}(\rho _{ij}t_{0}) \) for \(t\in [r_{ij}\rho _{ij}t_{0}, \ \rho _{ij}t_{0}], \ ij\in J.\) Let
For any \(\varepsilon >0\), consider the function V(t) defined as follows:
Therefore,
and
Hereafter, we prove that
Otherwise, there must exist \(ij\in J\) and \(\theta \in (t_{0}, \ +\infty ) \) such that
and
Consequently,
and
where \(\nu \in [\rho _{kl}t_{0} , \ t], \ t \in [t_{0} , \ \theta ), \ kl\in J\).
We multiply both sides of (3.5) through by \(e ^{ \int _{t_{0}}^{t}a_{ij}(u)du} \) and integrating it on \( [t_{0}, \ \theta ]\). It follows that
which, with the help of (3.7), (3.10), (3.13) and (3.15), entails that
This is a clear contradiction of (3.12). Thus, (3.11) is true. Letting \(\varepsilon \longrightarrow 0^{+}\) leads to
Then, using a similar theoretical derivation in the proof of (3.14) and (3.15), according to (3.16), we get
and
This ends the proof. \(\square \)
4 An Example and Its Numerical Simulations
Example 4.1
Consider the following SICNNs with neutral type proportional delays and D operators:
where \(p_{ij}(t)=\frac{1}{5}e^{-t^2}\sin (i+j)t , \ r _{ij} = \ q _{ij} = \frac{1}{2}, \ i,j=1,2, \)
Obviously, we can choose
such that system (4.1) satisfies \((T_{0})\), \((T_{1})\) and \((T_{2})\) and consequently it follows from Theorem 3.1 that the almost periodic solution of system (4.1) is globally generalized exponentially stable.The numerical simulations in Fig. 1 strongly support the conclusion and there are three groups of different initial values.
Remark 4.1
In this paper, it is the first time to deal with the existence and generalized exponential stability of almost periodic solutions for SICNNs involving neutral type proportional delays and D operators. Here, the criteria established in this paper are easily testable and independent of delays, that is, the considered neural network models for which stability is insensitive to the presence of the delays. On the other hand, one can see that all results obtained in [15,16,17,18,19] are invalid for system (4.1). Thus, the main results obtained in this paper are essentially new and complement some corresponding ones of the existing literature. Furthermore, the method used in this paper affords a possible approach to analysis the global generalized convergence of other cellular neural networks with neutral type proportional delays and D operators.
References
Xiao B (2009) Existence and uniqueness of almost periodic solutions for a class of Hopfield neural networks with neutral delays. Appl Math Lett 22:528–533
Mandal S, Majee NC (2011) Existence of periodic solutions for a class of Cohen–Grossberg type neural networks with neutral delays. Neurocomputing 74(6):1000–1007
Liu B (2015) Pseudo almost periodic solutions for neutral type CNNs with continuously distributed leakage delays. Neurocomputing 148:445–454
Yu Y (2017) Global exponential convergence for a class of neutral functional differential equations with proportional delays. Math Methods Appl Sci 39:4520–4525
Yu Y (2016) Global exponential convergence for a class of HCNNs with neutral time-proportional delays. Appl Mathods Comput 285:1–7
Chen Z (2013) A shunting inhibitory cellular neural network with leakage delays and continuously distributed delays of neutral type. Neural Comput Appl 23:2429–2434
Liu X (2015) Exponential convergence of SICNNs with delays and oscillating coefficients in leakage terms. Neurocomputing 168:500–504
Zhao C, Wang Z (2015) Exponential convergence of a SICNN with leakage delays and continuously distributed delays of neutral type. Neural Process Lett 41:239–247
Wang W, Liu B (2014) Global exponential stability of pseudo almost periodic solutions for SICNNs with time-varying leakage delays. Abst Appl Anal 967328:1–17
Komanovskii VB, Nosov VR (1986) Stability of functional differential equations. Academic, London
Kuang Y (1993) Delay differential equations with applications in population dynamical system. Academic, New York
Liu B, Huang L (2006) Existence and uniqueness of periodic solutions for a kind of first order neutral functional differential equations. J Mathods Anal Appl 322(1):121–132
Xu Y (2017) Exponential stability of pseudo almost periodic solutions for neutral type cellular neural networks with D operator. Neural Process Lett. doi:10.1007/s11063-017-9584-8
Candan T (2016) Existence of positive periodic solutions of first order neutral differential equations with variable coefficients. Appl Mathods Lett 52:142–148
Yao L (2017) Global exponential convergence of neutral type shunting inhibitory cellular neural networks with D operator. Neural Process Lett 45:401–409
Zhang A (2016) Pseudo almost periodic solutions for neutral type SICNNs with \(D\) operator. J Exp Theor Artif Intell. doi:10.1080/0952813X.2016.1259268
Ockendon JR, Tayler AB (1971) The dynamics of a current collection system for an electric locomotive. Proc R Soc Lond Ser A Math Phys Eng Sci 322(1551):447–468
Liu B (2016) Global exponential convergence of non-autonomous cellular neural networks with multi-proportional delays. Neurocomputing 191:352–355
Liu B (2017) Finite-time stability of CNNs with neutral proportional delays and time-varying leakage delays. Math Methods Appl Sci 40:167–174
Fink AM (1974) Almost periodic differential equations. Lecture notes in mathematics. Springer, Berlin
Zhang C (2003) Almost periodic type functions and ergodicity. Kluwer, Beijing
Liu B, Huang L (2007) Existence and exponential stability of almost periodic solutions for cellular neural networks with mixed delays. Chaos Solitons Fractals 32:95–103
Diagana T (2006) Weighted pseudo almost periodic functions and applications. C R Acad Sci Paris Ser I 343(10):643–646
Diagana T (2008) Weighted pseudo-almost periodic solutions to some differential equations. Nonlinear Anal 68:2250–2260
Yu Y (2017) Exponential stability of pseudo almost periodic solutions for cellular neural networks with multi-proportional delays. Neural Process Lett 45:141–151
Huang Z (2016) Almost periodic solutions for fuzzy cellular neural networks with multi-proportional delays. Int J Mach Learn Cyber. doi:10.1007/s13042-016-0507-1
Hale JK (1980) Ordinary differential equations. Krieger, Malabar
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
The author declare that no conflict of interest.
Rights and permissions
About this article
Cite this article
Zhang, A. Almost Periodic Solutions for SICNNs with Neutral Type Proportional Delays and D Operators. Neural Process Lett 47, 57–70 (2018). https://doi.org/10.1007/s11063-017-9631-5
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11063-017-9631-5
Keywords
- Generalized exponential stability
- Almost periodic solution
- Shunting inhibitory cellular neural network
- Neutral type proportional delay
- D operator