Exponential Stability of Hopfield Neural Network Model with Non-Instantaneous Impulsive Effects

Rui Ma

Michal Fečkan

^2,3,*

and

Jinrong Wang

Department of Mathematics, Guizhou University, Guiyang 550025, China

Department of Mathematical Analysis and Numerical Mathematics, Faculty of Mathematics, Physics and Informatics, Comenius University in Bratislava, Mlynská Dolina, 842 48 Bratislava, Slovakia

Mathematical Institute, Slovak Academy of Sciences, Štefánikova 49, 814 73 Bratislava, Slovakia

Author to whom correspondence should be addressed.

Axioms2023,12(2), 115;https://doi.org/10.3390/axioms12020115

Submission received: 29 December 2022 /Revised: 18 January 2023 /Accepted: 18 January 2023 /Published: 22 January 2023

(This article belongs to the Special IssueDifferential Equations and Dynamical Systems—Theory and Applications)

Downloadkeyboard_arrow_downVersions Notes

Abstract

We introduce a non-instantaneous impulsive Hopfield neural network model in this paper. Firstly, we prove the existence and uniqueness of an almost periodic solution of this model. Secondly, we prove that the solution of this model is exponentially stable. Finally, we give an example of this model.

Keywords:

non-instantaneous impulsive;Hopfield neural network;almost periodic;exponentially stable

MSC:

26A33

1. Introduction

It is well known that neural network models have many applications in the area of parallel computing, associative memory, pattern recognition, computer vision etc. [1,2,3,4,5]. Therefore, more and more experts and scholars pay attention to neural network models.

The studies on neurocomputing have been improved very fast after the work of McCulloch et al. [6]. One of the neural networks model was given by Hopfield [7,8]. In the actual situation, system can be affected by short-term fluctuations in the environment. Impulses are commonly used to describe this phenomenon. For instance, according to Arbib [9] and Haykin [10], when a stimulus from the body or the external environment is received by receptors, the electrical impulses will be conveyed to the neural net and impulsive effects arise naturally in the net. Stamova and Stamov [11] proposed a Hopfield neural network with impulsive effects at fixed moments as follows

\begin{matrix} \{\begin{matrix} {\dot{w}}_{i} (ι) = \sum_{j = 1}^{n} a_{i j} (ι) w_{j} (ι) + \sum_{j = 1}^{n} b_{i j} (ι) f_{j} (w_{j} (ι)) + g_{i} (ι), ι \neq ι_{k}, \\ Δ w (ι_{k}) = B_{k} w (ι_{k}) + C_{k} (w (ι_{k})) + h_{k}, k \in N_{+}, \end{matrix} \end{matrix}

(1)

where

ι \in J : = {0} ⋃ R_{+}

R_{+} : = {b | b

is a positive real number},

ι_{i}

(0 \leq ι_{1} < ι_{2} < \dots)

stand for the times which are impulses,

a_{i j}, b_{i j}, g_{i} \in C (J, R)

f_{j} \in C (J, R)

i = 1, 2, \dots, n, j = 1, 2, \dots, n

C (J, R)

is the space of all the continuous functions from

J

R

w (ι) = c o l (w_{1} (ι), w_{2} (ι), \dots, w_{n} (ι))

Δ w (ι_{k}) = w (ι_{k}^{+}) - w (ι_{k}^{-})

w (ι_{k}^{+})

is the right limits of

w (ι_{k})

and

w (ι_{k}^{-})

is the left limits of

w (ι_{k})

B_{k} \in R^{n \times n}

C_{k} \in C (R_{+}^{n}, R^{n})

R_{+}^{n} = {x = (x_{1}, x_{2}, \dots, x_{n}) \in R^{n} | x_{i} > 0, i = 1, 2, \dots, n}

, where

R^{n}, n \in N

isn-dimensional Euclidean space,

h_{k} \in R^{n}, k \in N_{+}

N : = {0, 1, 2, 3, \dots}

and

N_{+} : = N / {0}

However, most systems do not return to normal immediately after the impulse [12]. The system stays active for a limited period of time. Therefore, Hernández et al. [13] firstly introduced the theory of non-instantaneous impulses and established the existence of solutions for a class of impulsive differential equations. After that, Wang et al. [14,15,16] generalized this model and carried out more in-depth research on non-instantaneous impulsive differential equations. In general, there are no impulses that happen instantaneously, that is to say, it is non-instantaneous, even if the event occurs over a short period of time. Non-instantaneous impulsive effects exist in Hopfield neural network. For instance, in implementation of electronic networks, the state of the network is often subject to non-instantaneous perturbations, which may be caused by noise instances. Moreover, many evolutionary processes, particularly some biological systems, such as biological neural networks and bursting rhythm models in pathology, might exhibit non-instantaneous impulsive effects as well. Therefore, it is beneficial to study a class of differential equations with non-instantaneous impulses.

Then, we consider the case of the model (1) with non-instantaneous impulses as follows

\begin{matrix} \{\begin{matrix} {\dot{w}}_{i} (ι) = \sum_{j = 1}^{n} a_{i j} (ι) w_{j} (ι) + \sum_{j = 1}^{n} b_{i j} (ι) f_{j} (w_{j} (ι)) + g_{i} (ι), ι \in (l_{k}, m_{k + 1}], k \in N, \\ w (m_{k}^{+}) = B_{k} w (m_{k}^{-}) + C_{k} (w (m_{k}^{-})) + h_{k}, k \in N_{+}, \\ w (ι) = B_{k} w (m_{k}^{-}) + C_{k} (w (m_{k}^{-})) + h_{k}, ι \in (m_{k}, l_{k}], k \in N_{+}, \\ w (l_{k}^{+}) = w (l_{k}^{-}), k \in N_{+}, \end{matrix} \end{matrix}

(2)

where

0 = l_{0} < m_{1} < l_{1} < m_{2} < l_{2} < \dots < m_{k} < l_{k} < m_{k + 1} < \dots

. The solution

w (ι) = w (ι; ι_{0}, w_{0})

of model (2) with the initial condition

w (ι_{0}^{+}) = w_{0} \in R_{+}^{n}, ι_{0} \in J

is a piecewise continuous function with points of discontinuity of the first kind at the moments

m_{k}, k \in N_{+}

, at which it is continuous from the left.

Periodic phenomenon is one of the phenomena widely existing in nature [17]. But many motion processes in the present world are approximate to periodic instead of strictly periodicity. Therefore, Danish mathematicians Bohr [18] first proposed the concept of almost periodic (AP), which is a significant generalization for practical application. Many scholars have demonstrated that it is more realistic to adopt an AP hypothesis in the process of AP study, when taking into account the impact of environmental factors, and this has certain ergodicity [19,20,21,22,23,24,25].

The rest of this paper is arranged as follows. InSection 2, we provide some of the necessary preliminaries for this paper. InSection 3, we prove the existence, uniqueness and exponential stability of the AP solution to (2). InSection 4, we present an Example to support our theoretical results.

2. Preliminaries

For the sequences

{m_{k}}

and

{l_{k}}

k \in N_{+}

, assume that

lim_{k \to + \infty} m_{k} = + \infty

lim_{k \to + \infty} l_{k} = + \infty

. Let the norm

∥ j (ι) ∥ = max {| j_{1} (ι) |, | j_{2} (ι) |, \dots, | j_{n} (ι) |}

for

j (ι) = {(j_{1} (ι), j_{2} (ι), \dots, j_{n} (ι))}^{⊤}

. The space

P C ([0, \infty), R^{n}) : = {w : [0, \infty) \to R^{n} :

w \in C ((m_{i}, m_{i + 1}]

R^{n})

w (m_{i}^{-}) = w (m_{i})

w (m_{i}^{+})

exist for any

i \in N}

endowed with norm

{∥ w ∥}_{P C} = sup_{ι \in [0, \infty)} ∥ w (ι) ∥

, where

C ((m_{i}, m_{i + 1}], R^{n})

represents the space which is made up of all the continuous functions from

(m_{i}, m_{i + 1}]

R^{n}

. It is obvious that

(P C ([0, \infty), R^{n}), ∥ \cdot ∥_{P C})

is a Banach space.

Definition 1

(see [26]).For the sequences

{M_{i}}_{i \in N_{+}}

M_{i} \in R^{n}

, if for any

i \in N_{+}

there exist

ε > 0

and integer p such that the following inequality hold

∥ M_{i + p} - M_{i} ∥ < ε,

(3)

then p is called to be ε-AP of the

{M_{i}}_{i \in N_{+}}

M_{i} \in R^{n}

Definition 2

(see [27]).

{M_{i}}_{i \in N_{+}}

M_{i} \in R^{n}

are called to be AP sequences if for any

ε > 0

, there exists a relatively dense set of its ε-AP.

Definition 3

(see [26]).The

w \in P C ([0, \infty), R^{n})

is called an AP function if all of the conditions are satisfied as follows

(i): ${m_{i}^{j}}$ , $i, j \in N_{+}$ are uniformly AP sequences, where $m_{i}^{j} = m_{i + j} - m_{i}$ .
(ii): For any $ε > 0$ , there exists a number $δ = δ (ε)$ which is positive, such that if $ι_{1}$ and $ι_{2}$ are the points in the same continuous interval and $| ι_{1} - ι_{2} | < δ$ , then $∥ w (ι_{1}) - w (ι_{2}) ∥ < ε$ .
(iii): For any $ε > 0$ , there exists a relatively dense set Γof $ε -$ AP, such that if $ϑ \in Γ$ , then $∥ w (ι + ϑ) - w (ι) ∥ < ε$ for all $ι \in [0, \infty)$ satisfying the condition $| ι - m_{i} | > ε, i \in N_{+}$ .

Together with model (2), we shall consider the linear model

\begin{matrix} \{\begin{matrix} \dot{w} (ι) = A (ι) w (ι), ι \in (l_{k}, m_{k + 1}], k \in N, \\ w (m_{k}^{+}) = B_{k} w (m_{k}^{-}), k \in N_{+}, \\ w (ι) = B_{k} w (m_{k}^{-}), ι \in (m_{k}, l_{k}], k \in N_{+}, \\ w (l_{k}^{+}) = w (l_{k}^{-}), k \in N_{+}, \end{matrix} \end{matrix}

(4)

where

A (ι) = (a_{i j}), i = 1, 2, 3, \dots, n, j = 1, 2, \dots, n

Let

w (ι) = W (ι, ι_{0}) w_{ι_{0}}

0 \leq ι_{0} \leq ι

represents the solution of (4) with

w (ι_{0}) = w_{ι_{0}}

, where

W (ι, ι_{0})

is the Cauchy matrix of model (4) which can be looked up on [28].

We propose some assumptions as follows.

(H₁): The sequences ${l_{k}^{τ}}$ , $l_{k}^{τ} = l_{k + τ} - l_{k}$ and ${m_{k}^{τ}}$ , $m_{k}^{τ} = m_{k + τ} - m_{k}$ , $k, τ \in N_{+}$ are uniformly AP and $0 < l_{k} - m_{k} \leq θ < + \infty$ , $0 < ς \leq m_{k + 1} - l_{k} \leq \bar{θ} < + \infty$ , $k \in N_{+}$ .
(H₂): The matrix function $A \in C (J, R^{n \times n})$ is AP in the sense of Bohr.
(H₃): The sequence ${B_{k}}, k \in N_{+}$ is AP.
(H₄): The functions $f_{j} (ι)$ are AP in the sense of Bohr, and
$0 < sup_{ι \in J} | f_{j} (ι) | < \infty, f_{j} (0) = 0,$
and there exists an $L_{1} > 0$ such that for $ι, s \in R$ ,
$max_{j = 1, 2, 3, \dots, n} | f_{j} (ι) - f_{j} (s) | < L_{1} | ι - s | .$
(H₅): The functions $b_{i j} (ι)$ are AP in the sense of Bohr, and
$0 < sup_{ι \in J} | b_{i j} (ι) | = {\bar{b}}_{i j} < \infty .$
(H₆): The functions $g_{i} (ι), i = 1, 2, 3, \dots, n$ , are AP in the sense of Bohr, the sequences ${h_{k}}, k \in N_{+}$ are AP and there exists a $C > 0$ such that
$max \{{∥ g ∥}_{P C}, sup_{k \in N_{+}} ∥ h_{k} ∥\} \leq C,$
where $g (ι) = (g_{1} (ι), g_{2} (ι), \dots, g_{n} (ι))$ .
(H₇): The sequence of functions ${C_{k} (x)}, k \in N_{+}$ is AP uniformly with respect to $x \in R_{+}^{n}$ , and there exists an $L_{2} > 0$ such that
$∥ C_{k} (x) - C_{k} (y) ∥ \leq L_{2} ∥ x - y ∥,$
for $k \in N_{+}, x, y \in R^{n}$ . $C_{k} (x) = x$ if and only if $x = (0, 0, \dots, 0)$ .

Now, we need the following Lemmas.

Lemma 4

(see [28]).Assume that

(H_{1})

–

(H_{3})

hold. Then, for the Cauchy matrix

W (ι, ι_{0})

of model (4)there exist positive constants

K \geq 0

and

Υ > 0

such that

∥ W (ι, ι_{0}) ∥ \leq K e^{- Υ (ι - ι_{0})}, 0 \leq ι_{0} \leq ι .

Lemma 5

(see [28]).For any

ε > 0

0 \leq ι_{0} < ι

| ι - m_{i} | > ε

| ι - l_{i} | > ε

| ι_{0} - m_{i} | > ε

and

| ι_{0} - l_{i} | > ε

i \in N_{+}

, there exist a constant

K > 0

and a relatively dense set ofΓ of ε-AP such that

∥ W (ι + r, ι_{0} + r) - W (ι, ι_{0}) ∥ \leq ε K e^{- \frac{1}{2} Υ (ι - ι_{0})}, r \in Γ .

Lemma 6

(see [11]).Let conditions

(H_{1})

–

(H_{6})

hold. Then for each

ε > 0

, there exist

ε_{1}

0 < ε_{1} < ε

, a relatively dense setΓ of real numbers and a set Q of integers such that the following relations are fulfilled.

(a): $∥ A (ι + r) - A (ι) ∥ < ε, ι \in J, r \in Γ$ ;
(b): $| b_{i j} (ι + r) - b_{i j} (ι) | < ε, ι \in J, r \in Γ, i, j = 1, 2, 3, \dots, n$ ;
(c): $| f_{j} (ι + r) - f_{j} (ι) | < ε, ι \in J, r \in Γ, j = 1, 2, 3, \dots, n$ ;
(d): $| g_{j} (ι + r) - g_{j} (ι) | < ε, ι \in J, r \in Γ, j = 1, 2, 3, \dots, n$ ;
(e): $∥ B_{k + q} - B_{k} ∥ < ε, q \in Q, k \in R_{+}$ ;
(f): $| h_{k + q} - h_{k} | < ε, q \in Q, k \in R_{+}$ ;
(g): $| l_{k}^{q} - r | < ε_{1}$ , $| m_{k}^{q} - r | < ε_{1}$ , $q \in Q$ , $r \in Γ$ , $k \in N_{+}$ .

Lemma 7

(see [26]).If the sequences

{m_{i}^{j}}

i, j \in N

are uniformly AP, then we can get

(i): There exists a constant $ρ > 0$ such that $sup_{t \to + \infty} \frac{μ (ι + t, ι)}{t} = ρ$ which is uniformly with respect to $ι > 0$ .
(ii): For any $p > 0$ , there exists N which is a positive integer such that the number of elements in the sequences ${m_{i}}$ on each interval of length p does not exceed N. We can choose $N \geq ρ$ .

3. Main Results

Theorem 8.

Assume that conditions

(H_{1})

–

(H_{7})

are satisfied, model(2) has a unique positive AP solution if

K \{\frac{L_{1}}{Υ} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + L_{2} N_{1}\} < 1 .

Proof.

Let

N_{1} = sup_{ι \in J} \sum_{k = 1}^{μ (ι, 0)} e^{- Υ (ι - m_{k}^{+})}

N_{2} = sup_{ι \in J} \sum_{k = 1}^{μ (ι, 0)} e^{- \frac{1}{2} Υ (ι - m_{k}^{+})}

Ω : = {w \in P C (J, R_{+}^{n}), w

is AP

(∥ w (\cdot + r) - w (\cdot) ∥ < ε, r \in Γ)

and

{∥ w ∥}_{P C} \leq ℵ}

, where

Γ

is mentioned in Lemma 5.

For

l_{k} < ι < m_{k + 1}, k \in N

, let

\begin{matrix} φ & = & \sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} W (ι, u) g (u) d u + \int_{l_{μ (ι, 0)}}^{ι} W (ι, u) g (u) d u \\ + \sum_{k = 1}^{μ (ι, 0)} W (ι, m_{k}^{+}) h_{k} . \end{matrix}

Then,

\begin{matrix} {∥ φ ∥}_{P C} & \leq & sup_{ι \in J} {\sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι, u) ∥ ∥ g (u) ∥ d u + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι, u) ∥ ∥ g (u) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ h_{k} ∥} \\ \leq & sup_{ι \in J} \{\int_{0}^{ι} ∥ W (ι, u) ∥ ∥ g (u) ∥ d u + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ h_{k} ∥\} \\ \leq & sup_{ι \in J} \{max_{i = 1, 2, \dots, n} \int_{0}^{ι} ∥ W (ι, u) ∥ | g_{i} (u) | d u + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ h_{k} ∥\} \\ \leq & sup_{ι \in J} {\int_{0}^{ι} K e^{- Υ (ι - u)} C d u + \sum_{k = 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{+})} C} \\ \leq & \frac{K C}{Υ} + K N_{1} C \\ \leq & K C (\frac{1}{Υ} + N_{1}) = ℵ . \end{matrix}

(5)

Let

r \in Γ

q \in Q

, where the sets

Γ

andQ are determined in Lemma 6. Then,

\begin{matrix} sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ & \leq & sup_{ι \in J} {\sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι + r, u + r) - W (ι, u) ∥ ∥ g (u + r) ∥ d u \\ + \sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι, u) ∥ ∥ g (u + r) - g (u) ∥ d u \\ + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι + r, u + r) - W (ι, u) ∥ ∥ g (u + r) ∥ d u \\ + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι, u) ∥ ∥ g (u + r) - g (u) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι + r, m_{k + q}^{+}) - W (ι, m_{k}^{+}) ∥ ∥ h_{k + q} ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ h_{k + q} - h_{k} ∥} \\ \leq & sup_{ι \in J} {\int_{0}^{ι} ∥ W (ι + r, u + r) - W (ι, u) ∥ ∥ g (u + r) ∥ d u \\ + \int_{0}^{ι} ∥ W (ι, u) ∥ ∥ g (u + r) - g (u) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι + r, m_{k + q}^{+}) - W (ι, m_{k}^{+}) ∥ ∥ h_{k + q} ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ h_{k + q} - h_{k} ∥} \\ \leq & sup_{ι \in J} {\int_{0}^{ι} ε K e^{- \frac{1}{2} Υ (ι - u)} C d u + \int_{0}^{ι} K e^{- Υ (ι - u)} ε d u \\ + \sum_{k = 1}^{μ (ι, 0)} ε K e^{- \frac{1}{2} Υ (ι - m_{k}^{+})} C + \sum_{k = 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{+})} ε} \\ \leq & ε K C \frac{2}{Υ} + K ε \frac{1}{Υ} + ε K N_{2} C + K ε N_{1} \\ \leq & ε (K C \frac{2}{Υ} + K \frac{1}{Υ} + K N_{2} C + K N_{1}) . \end{matrix}

Set

F (ι, w) = c o l {F_{1} (ι, w), F_{2} (ι, w), \dots, F_{n} (ι, w)},

where

F_{i} (ι, w) = \sum_{j = 1}^{n} b_{i j} (ι) f_{j} (w_{j} (ι)), i = 1, 2, 3, \dots, n .

We define in

Ω

an operatorT,

\begin{matrix} T_{w} & = & \sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} W (ι, u) (F (u, w (u)) + g (u)) d u \\ + \int_{l_{μ (ι, 0)}}^{ι} W (ι, u) (F (u, w (u)) + g (u)) d u \\ + \sum_{k = 1}^{μ (ι, 0)} W (ι, m_{k}^{+}) (C_{k} (w (m_{k}^{-})) + h_{k}) \end{matrix}

(6)

and consider a subset

\hat{Ω} \subset Ω

, where

\hat{Ω} = \{{w \in Ω : ∥ w - φ ∥}_{P C} \leq \frac{R ℵ}{1 - R}\} .

Consequently, for an arbitrary

w \in \hat{Ω}

from (5) and (6) it follows that

{∥ w ∥}_{P C} \leq {∥ w - φ ∥}_{P C} + {∥ φ ∥}_{P C} \leq \frac{R ℵ}{1 - R} + ℵ = \frac{ℵ}{1 - R} .

Now, we prove thatT is a self-mapping from

\hat{Ω}

\hat{Ω}

For

w \in \hat{Ω}

we have

\begin{matrix} ∥ T_{w} {- φ ∥}_{P C} \\ \leq & sup_{ι \in J} {\sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι, u) ∥ ∥ F (u, w (u)) ∥ d u \\ + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι, u) ∥ ∥ F (u, w (u)) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k} (w (m_{k}^{-})) ∥} \\ \leq & sup_{ι \in J} {\int_{0}^{ι} ∥ W (ι, u) ∥ ∥ F (u, w (u)) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k} (w (m_{k}^{-})) ∥} \\ \leq & sup_{ι \in J} {max_{i = 1, 2, \dots, n} \int_{0}^{ι} ∥ W (ι, u) ∥ \sum_{j = 1}^{n} ∥ b_{i j} (ι) ∥ ∥ f_{j} (w_{j} (u) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k} (w (m_{k}^{-})) ∥} \\ \leq & sup_{ι \in J} {max_{i = 1, 2, \dots, n} \int_{0}^{ι} ∥ W (ι, u) ∥ \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} ∥ w (u) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ L_{2} ∥ w (m_{k}^{-}) ∥} \\ \leq & {max_{i = 1, 2, \dots, n} \int_{0}^{ι} K e^{- Υ (ι - u)} L_{1} \sum_{j = 1}^{n} {\bar{b}}_{i j} d u \\ + \sum_{k = 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{+})} L_{2}} {∥ w ∥}_{P C} \\ \leq & K \{max_{i = 1, 2, \dots, n} \frac{L_{1}}{Υ} \sum_{j = 1}^{n} {\bar{b}}_{i j} + N_{1} L_{2}\} {∥ w ∥}_{P C} \\ = & {R ∥ w ∥}_{P C} \leq \frac{R ℵ}{1 - R} . \end{matrix}

(7)

Let

r \in Γ

q \in Q

, where the sets

Γ

andQ are determined in Lemma 6. Then

\begin{matrix} ∥ T_{w} (ι + r) - T_{w} (ι) ∥ \\ \leq & sup_{ι \in J} ∥ (T_{w} (ι + r) - T_{w} (ι)) - (φ (ι + r) - φ (ι)) ∥ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & sup_{ι \in J} ∥ (T_{w} (ι + r) - φ (ι + r)) - (T_{w} (ι) - φ (ι)) ∥ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & sup_{ι \in J} {\sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι + r, u + r) - W (ι, u) ∥ ∥ F (u + r, w (u + r)) ∥ d u \\ + \sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι, u) ∥ ∥ F (u + r, w (u + r)) - F (u, w (u)) ∥ d u \\ + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι + r, u + r) - W (ι, u) ∥ ∥ F (u + r, w (u + r)) ∥ d u \\ + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι, u) ∥ ∥ F (u + r, w (u + r)) - F (u, w (u)) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι + r, m_{k + q}^{+}) - W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) - C_{k} (w (m_{k}^{-})) ∥} \\ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & sup_{ι \in J} {\int_{0}^{ι} ∥ W (ι + r, u + r) - W (ι, u) ∥ ∥ F (u + r, w (u + r)) ∥ d u \\ + \int_{0}^{ι} ∥ W (ι, u) ∥ ∥ F (u + r, w (u + r)) - F (u, w (u)) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι + r, m_{k + q}^{+}) - W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) - C_{k} (w (m_{k}^{-})) ∥} \\ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & sup_{ι \in J} {max_{i = 1, 2, \dots, n} (\int_{0}^{ι} ∥ W (ι + r, u + r) - W (ι, u) ∥ | \sum_{j = 1}^{n} b_{i j} (u + r) f_{j} (w_{j} (u + r)) | d u \\ + \int_{0}^{ι} ∥ W (ι, u) ∥ | \sum_{j = 1}^{n} b_{i j} (u + r) f_{j} (w_{j} (u + r)) - \sum_{j = 1}^{n} b_{i j} (u) f_{j} (w_{j} (u)) | d u) \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι + r, m_{k + q}^{+}) - W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) - C_{k} (w (m_{k}^{-})) ∥} \\ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \end{matrix}

\begin{matrix} \leq & sup_{ι \in J} {max_{i = 1, 2, \dots, n} (\int_{0}^{ι} ∥ W (ι + r, u + r) - W (ι, u) ∥ | \sum_{j = 1}^{n} b_{i j} (u + r) f_{j} (w_{j} (u + r)) | d u \\ + \int_{0}^{ι} ∥ W (ι, u) ∥ | \sum_{j = 1}^{n} (b_{i j} (u + r) - b_{i j} (u)) f_{j} (w_{j} (u + r)) \\ + \sum_{j = 1}^{n} b_{i j} (u) (f_{j} (w_{j} (u + r)) - f_{j} (w_{j} (u))) | d u) \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι + r, m_{k + q}^{+}) - W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k + q} (w (m_{k + q}^{-})) - C_{k} (w (m_{k}^{-})) ∥} \\ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & sup_{ι \in J} {max_{i = 1, 2, \dots, n} (\int_{0}^{ι} ε K e^{- \frac{1}{2} Υ (ι - u)} \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} | w_{j} (u + r)) | d u \\ + \int_{0}^{ι} K e^{- Υ (ι - u)} (\sum_{j = 1}^{n} ε L_{1} | w_{j} (u + r) | + \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} | w_{j} (u + r) - w_{j} (u) |) d u) \\ + \sum_{k = 1}^{μ (ι, 0)} ε K e^{- \frac{1}{2} Υ (ι - m_{k}^{+})} L_{2} ∥ w (m_{k + q}^{-}) ∥ \\ + \sum_{k = 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{-})} L_{2} ∥ w (m_{k + q}^{-}) - w (m_{k}^{-}) ∥} \\ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & sup_{ι \in J} {max_{i = 1, 2, \dots, n} (\frac{ε 2 K}{Υ} \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} \frac{ℵ}{1 - R} + \frac{K}{Υ} (ε L_{1} \frac{ℵ}{1 - R} + \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} ε)) \\ + \sum_{k = 1}^{μ (ι, 0)} ε K e^{- \frac{1}{2} Υ (ι - m_{k}^{+})} L_{2} \frac{ℵ}{1 - R} + \sum_{k = 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{-})} L_{2} ε} \\ + sup_{ι \in J} ∥ φ (ι + r) - φ (ι) ∥ \\ \leq & ε {\frac{L_{1}}{Υ} (max_{i = 1, 2, \dots, n} (\sum_{j = 1}^{n} 2 K {\bar{b}}_{i j} \frac{ℵ}{1 - R} + K \frac{ℵ}{1 - R} + K \sum_{j = 1}^{n} {\bar{b}}_{i j}) + K L_{2} N_{2} \frac{ℵ}{1 - R} + K L_{2} N_{1}\} \\ + ε (K C \frac{2}{Υ} + K \frac{1}{Υ} + K N_{2} C + K N_{1}) \\ \leq & ε {\frac{L_{1}}{Υ} (max_{i = 1, 2, \dots, n} (\sum_{j = 1}^{n} 2 K {\bar{b}}_{i j} \frac{ℵ}{1 - R} + K \frac{ℵ}{1 - R} + K \sum_{j = 1}^{n} {\bar{b}}_{i j}) + K L_{2} N_{2} \frac{ℵ}{1 - R} + K L_{2} N_{1} \\ + K C \frac{2}{Υ} + K \frac{1}{Υ} + K N_{2} C + K N_{1}} . \end{matrix}

(8)

Consequently, after (7) and (8), we obtain that

T_{w} \in \hat{Ω}

Let

ϕ \in \hat{Ω}

ξ \in \hat{Ω}

. Then,

\begin{matrix} ∥ T_{ϕ} - T_{ξ} ∥_{P C} \\ \leq & sup_{ι \in J} {\sum_{k = 0}^{μ (ι, 0) - 1} \int_{l_{k}}^{m_{k + 1}} ∥ W (ι, u) ∥ ∥ F (u, ϕ (u)) - F (u, ξ (u)) ∥ d u \\ + \int_{l_{μ (ι, 0)}}^{ι} ∥ W (ι, u) ∥ ∥ F (u, ϕ (u)) - F (u, ξ (u)) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k} (ϕ (m_{k}^{-})) - C_{k} (ξ (m_{k}^{-})) ∥ \\ \leq & sup_{ι \in J} {\int_{0}^{ι} ∥ W (ι, u) ∥ ∥ F (u, ϕ (u)) - F (u, ξ (u)) ∥ d u \\ + \sum_{k = 1}^{μ (ι, 0)} ∥ W (ι, m_{k}^{+}) ∥ ∥ C_{k} (ϕ (m_{k}^{-})) - C_{k} (ξ (m_{k}^{-})) ∥} \\ \leq & sup_{ι \in J} {\int_{0}^{ι} max_{i = 1, 2, \dots, n} K e^{- Υ (ι - u)} \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} d u \\ + \sum_{k = 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{+})} L_{2}} {∥ ϕ - ξ ∥}_{P C} \\ \leq & K \{\frac{L_{1}}{Υ} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + L_{2} N_{1}\} {∥ ϕ - ξ ∥}_{P C} . \end{matrix}

(9)

Then from (9) it follows thatT is a contracting operator in

\hat{Ω}

, and there exists a unique AP solution of (2). □

Theorem 9.

Assume that all conditions in Theorem 8 and

K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + N ln (1 + K L_{2}) < Υ

hold. Then, the solution of(2)is globally exponentially stable.

Proof.

Let now

x (ι)

be an arbitrary solution of (2). Then, we obtain

\begin{matrix} ∥ w (ι) - x (ι) ∥ & \leq & K e^{- Υ (ι - ι_{0})} ∥ w (ι_{0}) - x (ι_{0}) ∥ \\ + \int_{ι_{0}}^{ι} max_{i = 1, 2, \dots, n} K e^{- Υ (ι - u)} \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1} ∥ w (u) - x (u) ∥ d u \\ + \sum_{k = μ (ι_{0}, 0) + 1}^{μ (ι, 0)} K e^{- Υ (ι - m_{k}^{+})} L_{2} ∥ w (m_{k}^{-}) - x (m_{k}^{-}) ∥ . \end{matrix}

Set

v (ι) = ∥ w (ι) - x (ι) ∥ e^{Υ ι}

, then by means of Gronwall-Bellman’s inequality, it follows that

\begin{matrix} ∥ w (ι) - x (ι) ∥ & \leq & K e^{- Υ (ι - ι_{0})} ∥ w (ι_{0}) - x (ι_{0}) ∥ \prod_{k = μ (ι_{0}, 0) + 1}^{μ (ι, 0)} (1 + K L_{2} e^{- Υ (ι - m_{k}^{+})}) \\ e^{\int_{ι_{0}}^{ι} max_{i = 1, 2, \dots, n} K e^{- Υ (ι - u)} \sum_{j = 1}^{n} {\bar{b}}_{i j} L_{1}} \\ \leq & K ∥ w (ι_{0}) - x (ι_{0}) ∥ {(1 + K L_{2})}^{μ (ι, ι_{0})} e^{- Υ (ι - ι_{0})} e^{K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} (ι - ι_{0})} \\ \leq & K ∥ w (ι_{0}) - x (ι_{0}) ∥ {(1 + K L_{2})}^{μ (ι, ι_{0})} e^{(- Υ + K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j}) (ι - ι_{0})} \\ \leq & K ∥ w (ι_{0}) - x (ι_{0}) ∥ e^{N (ι - ι_{0}) ln (1 + K L_{2})} e^{(- Υ + K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j}) (ι - ι_{0})} \\ \leq & K ∥ w (ι_{0}) - x (ι_{0}) ∥ e^{(- Υ + K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + N ln (1 + K L_{2})) (ι - ι_{0})} . \end{matrix}

Obviously, if there exists

K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + N ln (1 + K L_{2}) < Υ

, then the solution of (2) is exponentially. □

4. Example

Example 10.

We shall consider the classical model of Hopfield neural networks

\begin{matrix} \{\begin{matrix} {\dot{w}}_{i} (ι) = - \frac{1}{R_{i}} w_{i} (ι) + \sum_{j = 1}^{n} b_{i j} f_{j} (w_{j} (ι)) + g_{i} (ι), ι \in (l_{k}, m_{k + 1}], k \in N, \\ w (m_{k}^{+}) = B w (m_{k}^{-}) + C_{k} (w (m_{k}^{-})) + h_{k}, k \in N_{+}, \\ w (ι) = B w (m_{k}^{-}) + C_{k} (w (m_{k}^{-})) + h_{k}, ι \in (m_{k}, l_{k}], k \in N_{+}, \\ w (l_{k}^{+}) = w (l_{k}^{-}), k \in N_{+}, \end{matrix} \end{matrix}

(10)

where

ι \in J

R_{i} > 0

b_{i j} \in R, γ_{i} \in C (J, R), f_{j} \in C (R_{+}, R), i = 1, 2, \dots, n, j = 1, 2, \dots, n

x (ι) = c o l (x_{1} (ι), x_{2} (ι), \dots, x_{n} (ι))

B = d i a g [b_{i}], b_{i} \in R, i = 1, 2, \dots, n, C_{k} \in C (R_{+}^{n}, R), h_{k} \in R^{n}

Let

\begin{matrix} \{\begin{matrix} {\dot{w}}_{i} (ι) = - \frac{1}{R_{i}} w_{i} (ι), ι \in (l_{k}, m_{k + 1}], k \in N, \\ w (m_{k}^{+}) = B w (m_{k}^{-}), k \in N_{+}, \\ w (ι) = B w (m_{k}^{-}), ι \in (m_{k}, l_{k}], k \in N_{+}, \\ w (l_{k}^{+}) = w (l_{k}^{-}), k \in N_{+}, \end{matrix} \end{matrix}

be the linear part of (10).

The Cauchy matrix

W (ι, ι_{0})

of(10)is in the form

\begin{matrix} W (ι, ι_{0}) = \{\begin{matrix} B^{μ (ι, ι_{0}) + 1} e^{A (ι - l_{μ (ι, 0)})} \prod_{k = μ (ι, 0)}^{μ (ι_{0}, 0) + 2} e^{A (m_{k} - l_{k - 1})} e^{A (m_{μ (ι_{0}, 0) + 1} - ι_{0})}, \\ ι_{0} < m_{μ (ι_{0}, 0) + 1} < \dots < l_{μ (ι, 0)} < ι, \\ B^{μ (ι, ι_{0}) + 1} \prod_{k = μ (ι, 0)}^{μ (ι_{0}, 0) + 2} e^{A (m_{k} - l_{k - 1})} e^{A (m_{μ (ι_{0}, 0) + 1} - ι_{0})}, \\ ι_{0} < m_{μ (ι_{0}, 0) + 1} < \dots < m_{μ (ι, 0)} < ι, \\ B^{μ (ι, ι_{0}) + 1} e^{A (ι - l_{μ (ι, 0)})} \prod_{k = μ (ι, 0)}^{μ (ι_{0}, 0) + 1} e^{A (m_{k} - l_{k - 1})}, \\ ι_{0} < l_{μ (ι_{0}, 0)} < \dots < l_{μ (ι, 0)} < ι, \\ B^{μ (ι, ι_{0}) + 1} \prod_{k = μ (ι, 0)}^{μ (ι_{0}, 0) + 1} e^{A (m_{k} - l_{k - 1})}, \\ ι_{0} < l_{μ (ι_{0}, 0)} < \dots < m_{μ (ι, 0)} < ι . \end{matrix} \end{matrix}

Then,

\begin{matrix} ∥ W (ι, ι_{0}) ∥ & \leq & ∥B^{μ (ι, ι_{0}) + 1} e^{A (ι - ι_{0})}∥ \\ \leq & ∥e^{ln B^{μ (ι, ι_{0}) + 1}} e^{A (ι - ι_{0})}∥ \\ \leq & ∥e^{(μ (ι, ι_{0}) + 1) ln B} e^{A (ι - ι_{0})}∥ \\ \leq & ∥e^{ln B} e^{N (ι - ι_{0}) ln B} e^{A (ι - ι_{0})}∥ \\ \leq & ∥e^{ln B} e^{(A + N ln B) (ι - ι_{0})}∥ \\ \leq & e^{max_{i = 1, 2, \dots, n} ln b_{i}} e^{(max_{i = 1, 2, \dots, n} - \frac{1}{R_{i}} + max_{i = 1, 2, \dots, n} N ln b_{i}) (ι - ι_{0})} \\ \leq & e^{max_{i = 1, 2, \dots, n} ln b_{i}} e^{- (min_{i = 1, 2, \dots, n} \frac{1}{R_{i}} - max_{i = 1, 2, \dots, n} N ln b_{i}) (ι - ι_{0})} . \end{matrix}

Let

K = exp (max_{i = 1, 2, \dots, n} ln b_{i})

Υ = min_{i = 1, 2, \dots, n} \frac{1}{R_{i}} - max_{i = 1, 2, \dots, n} N ln b_{i}

, then we can obtain

∥ W (ι, ι_{0}) ∥ \leq K e^{- Υ (ι - ι_{0})}

According to the Theorems 8 and 9, assume that

(H_{1})

–

(H_{7})

are met and the following inequalities hold

\begin{matrix} K = exp (max_{i = 1, 2, \dots, n} ln b_{i}), \\ Υ = min_{i = 1, 2, \dots, n} \frac{1}{R_{i}} - max_{i = 1, 2, \dots, n} N ln b_{i}, \\ K \{\frac{L_{1}}{Υ} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + L_{2} N_{1}\} < 1 . \end{matrix}

Then, there exists a unique AP solution

w (ι)

of(10).

In addition, if the following inequalities hold

K L_{1} max_{i = 1, 2, \dots, n} \sum_{j = 1}^{n} {\bar{b}}_{i j} + N ln (1 + K L_{2}) < Υ,

then the solution

w (ι)

is globally exponentially stable.

5. Conclusions

Neural network models with impulses can study many phenomena in life. We note that Stamova and Stamov [11] proposed a Hopfield neural network with impulsive effects at fixed moments. We are very interested in this work. After careful reading, we introduced the non-instantaneous impulse factor into this model and proposed a Hopfield neural network non-instantaneous impulsive model. Then, we provided conditions for the existence of a unique AP solution and the exponential stability of the solution for this model.

There are many limitations to our work. It is known that asymptotic stability of solutions to impulsive systems can be treated in both weak (convergence towards the solution depends only on the elapsed time) and strong (convergence depends on the elapsed on the elapsed time and the number of impulses) flavors [29,30]. We deal with the classical weak stability in this paper. Then, we will gradually consider the case of strong stability for the model (2) with non-instantaneous impulses in future.

Author Contributions

The contributions of all authors (R.M., M.F. and J.W.) are equal. All the main results were developed together. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by Guizhou Data Driven Modeling Learning and Optimization Innovation Team ([2020]5016), the Slovak Research and Development Agency under the contract No. APVV-18-0308, and the Slovak Grant Agency VEGA No. 2/0127/20 and No. 1/0084/23.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

References

Viviani, P.; Drocco, M.; Baccega, D.; Colonnelli, I.; Aldinucci, M. Deep learning at scale. In Proceedings of the 2019 IEEE 27th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), Pavia, Italy, 13–15 February 2019; pp. 124–131. [Google Scholar]
Wouafo, H.; Chavet, C.; Coussy, P. Clone-Based encoded neural networks to design efficient associative memories.IEEE Trans. Neural Netw. Learn. Syst.2019,30, 3186–3199. [Google Scholar] [CrossRef] [PubMed] [Green Version]
Allen, F.T.; Kinser, J.M.; Caulfield, H.J. A neural bridge from syntactic to statistical pattern recognition.Neural Netw.1999,12, 519–526. [Google Scholar] [CrossRef] [PubMed]
Kriegeskorte, N. Deep neural networks: A new framework for modeling biological vision and brain information processing.Annu. Rev. Vis. Sci.2015,1, 417–446. [Google Scholar] [CrossRef] [PubMed] [Green Version]
Calise, A.J.; Rysdyk, R.T. Nonlinear adaptive flight control using neural networks.IEEE Control. Syst. Mag.1998,18, 14–25. [Google Scholar]
McCulloch, W.S.; Pits, W. A logical calculus of the ideas immanent in nervous activity.Bull. Math. Biophys.1943,5, 115–133. [Google Scholar] [CrossRef]
Hopfield, J.J. Neural networks and physical systems with emergent collective computational abilities.Proc. Natl. Acad. Sci. USA1982,79, 2554–2558. [Google Scholar] [CrossRef] [Green Version]
Hopfield, J.J. Neurons with graded response have collective computational properties like those of two-state neurons.Proc. Natl. Acad. Sci. USA1984,81, 3088–3092. [Google Scholar] [CrossRef] [Green Version]
Arbib, M.A.Brains, Machines and Mathematics; Springer: New York, NY, USA, 1987. [Google Scholar]
Haykin, S.Neural Networks: A Comprehensive Foundation; Prentice-Hall: Ehglewood Cliffs, NJ, USA, 1998. [Google Scholar]
Stamova, I.; Stamov, G.Applied Impulsive Mathematical Models; Springer International Publishing: Berlin/Heidelberg, Germany, 2016. [Google Scholar]
Li, M.M.; Wang, J.R.; O’Regan, D. Positive almost periodic solution for a noninstantaneous impulsive Lasota-Wazewska model.Bull. Iran. Math. Soc.2019,46, 851–864. [Google Scholar] [CrossRef]
Hernández, E.; O’Regan, D.; Benxax, M.A. On a new class of abstract integral equations and applications.Appl. Math. Comput.2012,219, 2271–2277. [Google Scholar] [CrossRef]
Wang, J.R.; Fečkan, M. A general class of impulsive evolution equations.Topol. Methods Nonlinear Anal.2015,46, 915–933. [Google Scholar] [CrossRef]
Wang, J.R.; Zhou, Y.; Lin, Z. On a new class of impulsive fractional differential equations.Appl. Math. Comput.2014,242, 649–657. [Google Scholar] [CrossRef]
Wang, J.R.; Fečkan, M.Non-Instantaneous Impulsive Differential Equations; IOP: London, UK, 2018. [Google Scholar]
Guan, Y.; Fečkan, M.; Wang, J.R. Periodic solutions and Hyers-Ulam stability of atmospheric Ekman flows.Discret. Contin. Dyn. Syst.2021,41, 1157–1176. [Google Scholar] [CrossRef]
Bohr, H. Zur theorie der fast periodischen funktionen: I. eine verallgemeinerung der theorie der fourierreihen.Acta Math.1925,45, 29–127. [Google Scholar] [CrossRef]
Chen, X.X. Almost periodic solutions of nonlinear delay population equation with feedback control.Nonlinear Anal. Real World Appl.2007,8, 62–72. [Google Scholar]
Chen, X.X.; Chen, F.D. Almost-periodic solutions of a delay population equation with feedback control.Nonlinear Anal. Real World Appl.2006,7, 559–571. [Google Scholar]
Zhang, R.; Wang, L. Almost periodic solutions for cellular neural networks with distributed delays.Acta Math. Sci.2011,31, 422–429. [Google Scholar]
Menouer, M.A.; Moussaoui, A.; Dads, E.A. Existence and global asymptotic stability of positive almost periodic solution for a predator-prey system in an artificial lake.Chaos Solitons Fractals2017,103, 271–278. [Google Scholar] [CrossRef]
Zhang, T.; Gan, X. Almost periodic solutions for a discrete fishing model with feedback control and time delays.Commun. Nonlinear Sci. Numer. Simul.2014,19, 150–163. [Google Scholar] [CrossRef]
Huang, P.; Li, X.; Liu, B. Almost periodic solutions for an asymmetric oscillation.J. Differ. Equ.2017,263, 8916–8946. [Google Scholar] [CrossRef] [Green Version]
Zhou, H.; Wang, W.; Yang, L. Stage-structured hematopoiesis model with delays in an almost periodic environment.Appl. Math. Lett.2021,120, 107336. [Google Scholar] [CrossRef]
Samoilenko, A.M.; Perestyuk, N.A.Impulsive Differential Equations; World Scientific: Singapore, 1995. [Google Scholar]
Stamova, I.Stability Analysis of Impulsive Functional Differential Equations; Walter de Gruyter: Berlin, Germany, 2009. [Google Scholar]
Ma, R.; Wang, J.R.; Li, M.M. Almost periodic solutions for two non-instantaneous impulsive biological models.Qual. Theory Dyn. Syst.2022,21, 84. [Google Scholar] [CrossRef]
Mancilla-Aguilar, J.L.; Haimovich, H.; Feketa, P. Uniform stability of nonlinear time-varying impulsive systems with eventually uniformly bounded impulse frequency.Nonlinear Anal. Hybrid Syst.2020,38, 100933. [Google Scholar] [CrossRef]
Feketa, P.; Klinshov, V.; Lücken, L. A survey on the modeling of hybrid behaviors: How to account for impulsive jumps properly.Commun. Nonlinear Sci. Numer. Simul.2021,103, 105955. [Google Scholar] [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Ma, R.; Fečkan, M.; Wang, J. Exponential Stability of Hopfield Neural Network Model with Non-Instantaneous Impulsive Effects.Axioms2023,12, 115. https://doi.org/10.3390/axioms12020115

AMA Style

Ma R, Fečkan M, Wang J. Exponential Stability of Hopfield Neural Network Model with Non-Instantaneous Impulsive Effects.Axioms. 2023; 12(2):115. https://doi.org/10.3390/axioms12020115

Chicago/Turabian Style

Ma, Rui, Michal Fečkan, and Jinrong Wang. 2023. "Exponential Stability of Hopfield Neural Network Model with Non-Instantaneous Impulsive Effects"Axioms 12, no. 2: 115. https://doi.org/10.3390/axioms12020115

APA Style

Ma, R., Fečkan, M., & Wang, J. (2023). Exponential Stability of Hopfield Neural Network Model with Non-Instantaneous Impulsive Effects.Axioms,12(2), 115. https://doi.org/10.3390/axioms12020115

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further detailshere.

Article Metrics

Article Access Statistics

For more information on the journal statistics, clickhere.

Multiple requests from the same IP address are counted as one view.

Abstract

1. Introduction

2. Preliminaries

3. Main Results

4. Example

5. Conclusions

Author Contributions

Funding

Institutional Review Board Statement

Informed Consent Statement

Data Availability Statement

Conflicts of Interest

References

Share and Cite

Article Metrics

Article Access Statistics

Movatterモバイル変換

Journals

Topics

Information

Author Services

Initiatives

About

Notice

Notice

Article Menu

Need Help?

Support

Feedback

Information

JSmol Viewer

Exponential Stability of Hopfield Neural Network Model with Non-Instantaneous Impulsive Effects

Cite

Further Information

Guidelines

MDPI Initiatives

Follow MDPI

Share Link

Share