
The existence and the S-asymptotic ω-periodic of the solution in fractional-order Cohen-Grossberg neural networks with inertia are studied in this paper. Based on the properties of the Riemann-Liouville (R-L) fractional-order derivative and integral, the contraction mapping principle, and the Arzela-Ascoli theorem, sufficient conditions for the existence and the S-asymptotic ω-period of the system are achieved. In addition, an example is simulated to testify the theorem.
Citation: Zhiying Li, Wangdong Jiang, Yuehong Zhang. Dynamic analysis of fractional-order neural networks with inertia[J]. AIMS Mathematics, 2022, 7(9): 16889-16906. doi: 10.3934/math.2022927
[1] | Huizhen Qu, Jianwen Zhou . S-asymptotically ω-periodic dynamics in a fractional-order dual inertial neural networks with time-varying lags. AIMS Mathematics, 2022, 7(2): 2782-2809. doi: 10.3934/math.2022154 |
[2] | Zhiying Li, Wei Liu . The anti-periodic solutions of incommensurate fractional-order Cohen-Grossberg neural network with inertia. AIMS Mathematics, 2025, 10(2): 3180-3196. doi: 10.3934/math.2025147 |
[3] | Yang Xu, Zhouping Yin, Yuanzhi Wang, Qi Liu, Anwarud Din . Mittag-Leffler projective synchronization of uncertain fractional-order fuzzy complex valued neural networks with distributed and time-varying delays. AIMS Mathematics, 2024, 9(9): 25577-25602. doi: 10.3934/math.20241249 |
[4] | Wei Liu, Qinghua Zuo, Chen Xu . Finite-time and global Mittag-Leffler stability of fractional-order neural networks with S-type distributed delays. AIMS Mathematics, 2024, 9(4): 8339-8352. doi: 10.3934/math.2024405 |
[5] | Yuehong Zhang, Zhiying Li, Wangdong Jiang, Wei Liu . The stability of anti-periodic solutions for fractional-order inertial BAM neural networks with time-delays. AIMS Mathematics, 2023, 8(3): 6176-6190. doi: 10.3934/math.2023312 |
[6] | Aziz Khan, Thabet Abdeljawad, Manar A. Alqudah . Neural networking study of worms in a wireless sensor model in the sense of fractal fractional. AIMS Mathematics, 2023, 8(11): 26406-26424. doi: 10.3934/math.20231348 |
[7] | M. Syed Ali, M. Hymavathi, Bandana Priya, Syeda Asma Kauser, Ganesh Kumar Thakur . Stability analysis of stochastic fractional-order competitive neural networks with leakage delay. AIMS Mathematics, 2021, 6(4): 3205-3241. doi: 10.3934/math.2021193 |
[8] | Pratap Anbalagan, Evren Hincal, Raja Ramachandran, Dumitru Baleanu, Jinde Cao, Chuangxia Huang, Michal Niezabitowski . Delay-coupled fractional order complex Cohen-Grossberg neural networks under parameter uncertainty: Synchronization stability criteria. AIMS Mathematics, 2021, 6(3): 2844-2873. doi: 10.3934/math.2021172 |
[9] | Zainab Alsheekhhussain, Ahmed Gamal Ibrahim, Rabie A. Ramadan . Existence of S-asymptotically ω-periodic solutions for non-instantaneous impulsive semilinear differential equations and inclusions of fractional order 1<α<2. AIMS Mathematics, 2023, 8(1): 76-101. doi: 10.3934/math.2023004 |
[10] | Yucai Ding, Hui Liu . A new fixed-time stability criterion for fractional-order systems. AIMS Mathematics, 2022, 7(4): 6173-6181. doi: 10.3934/math.2022343 |
The existence and the S-asymptotic ω-periodic of the solution in fractional-order Cohen-Grossberg neural networks with inertia are studied in this paper. Based on the properties of the Riemann-Liouville (R-L) fractional-order derivative and integral, the contraction mapping principle, and the Arzela-Ascoli theorem, sufficient conditions for the existence and the S-asymptotic ω-period of the system are achieved. In addition, an example is simulated to testify the theorem.
As a generalization of integer-order calculus, one of the main advantages of fractional-order calculus is that it can be considered as a superset of the integer one. It has the ability to accomplish what the integer-order calculus can not. In the last few decades, it has received a lot of attention because of its increasing applications in various fields, including bioengineering, economics, signal and image processing, rheology, biology and electrochemistry[1,2,3,4]. Compared with the traditional integer-order model, the fractional-order differential model provides an effective tool to describe the inherent memory and heredity effects of real materials and processes, it provides a powerful tool for describing the memory and heredity of different things. Nowadays, fractional-order dynamical systems have been widely used in electromagnetic wave, electrolyte polarization, viscoelastic, economic, biology systems [5,6,7,8,9]. Moreover, there are strong links between fractional-order derivative operators and fractional-order Brownian motion, continuous time random walk method and generalized central limit theorem. In addition, fractional-order derivative operators allow the long-term memory and nonlocal dependence of many anomalous processes. Some significant results regarding fractional-order neural networks (FNNs) have been published, especially on the dynamics of such systems, such as finite time stability, asymptotic stability, Mittag-Leffler stability, application and control of neural networks[10,11,12,13,14,15].
However, the FNNs they studied only contain one fractional-order derivative of the state in the system. Mathematics and physics consider them as super damping. But the neuron's state will change its dynamic properties when the inertia exceeds a critical point. The inertia in neural networks will increase the instability of the system. Furthermore, it has a profound meaning in biology, such as the cuttlefish's axon [16,17,18]. According to the biological background of the inertial neural system, the fractional-order inertial neural network simulates the dynamic behavior of the neurons more closely. Therefore, it is meaningful to consider damping in the system.
Integer-order neural networks with inertia have been studied a lot, and some exciting results have been obtained. Such as exponential synchronization, the fixed-time synchronization are studied in [19,20,21,22] and synchronization of coupled memristive neural networks (NNs) [23]. These inertial NNs have two fractional orders in the differential equations. Compared with the traditional NNs with a single order, the NNs with two different orders are more suitable to simulate the complex biological neural networks, and they can be used in many other fields, such as biology and cybernetics. Therefore, it is valuable to discuss the dynamics of FNNs with inertia. However, according to the enquires, most of the researches studied the stability of the FNNs without inertia, like Ke and Henriquez et al. who studied the asymptotic ω-periodicity of FNNs [24,25], etc. Researchers have studied the FNNs with inertia, except that Gu et al. and Zhang et al. who considered the stability and synchronicity of Riemann-Liouville FNNs with inertia [26,27].
In 1983, Cohen and Grossberg proposed a generalized neural network and ecological model, namely the Cohen-Grossberg neural network[28]. It contains many ecosystems and neural networks. At present, parallel processing, associative memory, and especially optimization computation have attracted wide interest. From the view of mathematics and physics, the Cohen-Grossberg neural network is a model of super-damping (damping tending to infinity), and the system with damping (weak damping) should also be considered in practical problems, that is, inertia should be considered in a Cohen-Grossberg neural network.
In this paper, we extend the inertial Cohen-Grossberg neural network model from integral-order to fractional-order. Inspired by the studies mentioned above, the existence and the S-asymptotic ω-periodic of the solution for fractional-order Cohen-Grossberg NNs with inertia (FCGNNIs) are discussed in this paper.
The main features of this paper are as follows:
1) A fractional-order Cohen-Grossberg neural network model with an inertia term is proposed for the first time. This model has practical application value in simulating complex biological neural networks.
2) By proper variable substitution, the model with two different fractional-order derivatives is simplified to a model with only one fractional-order derivative, which simplifies the study of the problem.
3) The sufficient conditions for the existence and S-asymptotic ω-periodic of the solution of the system are given.
The results are new. It is of theoretical significance and practical value to study the dynamic characteristics of the FCGNNI further.
Consider the following FCGNNI:
D2αt(σi(t))=−γiDαt(σi(t))−αi(σi(t))[hi(σi(t))−n∑j=1aijfj(σj(t))−Ii(t)], | (1.1) |
for all i=1,2,...,n, t≥0, where Dαt is the R-L fractional-order derivative with order α(0<α<1); σi(t), α(.)>0, hi(.), Ii(t) and γi>0 are the state variable, the abstract amplification function, the behavior function, the external input and the damping coefficient of the ith neuron at time t; aij represents the connection weight; fj(.) is the activation function of the jth neuron.
The system's initial conditions are
{σi(s)=φσi(s),Dαt(σi(s))=ψσi(s),s≤0,i=1,2,...,n, | (1.2) |
where φσi(s) and ψσi(s) are bounded and continuous in (−∞,0].
Remark. If α=1, (1.1) is the integer-order Cohen-Grossberg NNs with inertia:
d2σi(t)dt2=−γidσi(t)dt−αi(σi(t))[hi(σi(t))−n∑j=1aijfj(σj(t))−Ii(t)], |
i=1,2,...,n.
The discussion in this paper has been established on the following assumptions (i,j=1,2,...,n):
H1:αi(.) and α′i(.) are bounded. That is, there exist α_i≥0,¯αi>0 and Ai>0 such that
0≤α_i≤αi(σi(t))≤¯αi,|α′i(.)|≤Ai. |
H2: fj(.) is bounded and Lipschitz-continuous. That is, there exist lj>0 and ¯fj>0 such that
|fj(s)−fj(v)|≤lj|s−v|,|fj(.)|≤¯fj. |
H3: Let pi(ζi)=αi(ζi)hi(ζi), its derivative is bounded with p_i>0 and ¯pi>0. That is,
0≤p_i≤p′i(ζi)≤¯pi. |
H4:Ii(t) is bounded with the boundary Ii>0. That is, |Ii(t)|≤Ii.
Let χi(t)=Dαtσi(t)+ηiσi(t),ηi>0, from (1.1) one gets
{Dαtσi(t)=−ηiσi(t)+χi(t),Dαtχi(t)=−ηi(ηi−γi)σi(t)−(γi−ηi)χi(t)−αi(σi(t))[hi(σi(t))−n∑j=1aijfj(σj(t))−Ii(t)]. | (2.1) |
In this section, three theorems about the existence and the S−asymptotic ω−periodic of the solution to the system given by (1.1) are proposed.
Theorem 3.1. Under the assumptions H1−H4, if
Γ(α+1)>[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi+n∑j=1|aij|(¯αjli+Aj¯fi)+AiIi} +max1≤i≤n{1+|γi−ηi|}]Tα, | (3.1) |
then, the system given by (1.1) has a unique solution on [0,T](T>0).
Proof. Derive from Lemma 2 and (2.1):
{σi(t)=−ηiD−αtσi(t)+D−αtχi(t),χi(t)=−ηi(ηi−γi)D−αtσi(t)−(γi−ηi)D−αtχi(t)−D−αtαi(σi(t))[hi(σi(t))−n∑j=1aijfj(σj(t)−Ii(t)]. | (3.2) |
Let
U(t)={u(t)|u(t)= (ζ1(t),ζ2(t),...,ζn(t),χ1(t),χ2(t),...,χn(t))T,σi(t), χi(t)∈C[0,t],i=1,2,...,n}, it is clear that it is a Banach space, where the norm is ‖U‖=sup0≤t≤Tn∑i=1(|σi(t)|+|χi(t)|).
Define P:U(t)→U(t) as
(Pu)(t)={(Pσ1)(t),(Pσ2)(t),...,(Pσn)(t),(Pχ1)(t),(Pχ2)(t),...,(Pχn)(t)}T, |
where
{(Pζi)(t)=−ηiD−αtσi(t)+D−αtχi(t),(Pχi)(t)=−ηi(ηi−γi)D−αtσi(t)−(γi−ηi)D−αtχi(t)−D−αtαi(σi(t))[hi(σi(t))−n∑j=1aijfj(σj(t))−Ii(t)]. | (3.3) |
The first thing is to prove that PUδ⊂Uδ, where Uδ={u(t)∈U(t):‖u‖≤δ},
δ≥n∑i=1¯αi(|hi(0)|+n∑j=1|aij|¯fj+Ii)TαΓ(α+1)−[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}+max1≤i≤n{1+|γi−ηi|}]Tα. |
Derive from H1 and H3 that
{αi(σi(t))−αi(¯σi(t))=α′i(ξi)(σi(t)−¯σi(t)),αi(σi(t))hi(σi(t))−αi(¯σi(t))hi(¯σi(t))=p′i(¯ξi)(σi(t)−¯σi(t)), | (3.4) |
where ξi and ¯ξi are between σi and ¯σi.
From (3.3), one has
‖Pu‖=sup0≤t≤Tn∑i=1(|(Pσi)(t)|+|(Pχi)(t)|)≤sup0≤t≤Tn∑i=1{ηiΓ(α)∫t0(t−z)α−1|σi(z)|dz+1Γ(α)∫t0(t−z)α−1|χi(z)|dz+ηi|ηi−γi|Γ(α)∫t0(t−z)α−1|σi(z)|dz|γi−ηi|Γ(α)∫t0(t−z)α−1|χi(z)|dz+1Γ(α)∫t0(t−z)α−1|αi(σi(z))hi(σi(z))−σi(0)hi(0)|dz+1Γ(α)∫t0(t−z)α−1[¯αi|hi(0)|+¯αin∑j=1|aij|¯fj+¯αiIi]dz}≤sup0≤t≤Tn∑i=1{ηi+ηi|ηi−γi|+¯piΓ(α)∫t0(t−z)α−1|σi(z)|dz+1+|γi−ηi|Γ(α)∫t0(t−z)α−1|χi(z)|dz+¯αi|hi(0)|+¯αin∑j=1|aij|¯fj+¯αiIiΓ(α)∫t0(t−z)α−1dz}≤max1≤i≤n{ηi+ηi|ηi−γi|+¯piΓ(α)}∫t0(t−z)α−1‖U‖dz+max1≤i≤n{1+|γi−ηi|Γ(α)}∫t0(t−z)α−1‖U‖dz+¯αi|hi(0)|+¯αin∑j=1|aij|¯fj+¯αiIiΓ(α)∫t0(t−z)α−1dz≤[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}+max1≤i≤n{1+|γi−ηi|}]‖u‖TαΓ(α+1)+n∑i=1¯αi(|hi(0)|+n∑j=1|aij|¯fj+Ii)Γ(α+1)Tα≤[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}+max1≤i≤n{1+|γi−ηi|}]δTαΓ(α+1)+Γ(α+1)−[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}+max1≤i≤n{1+|γi−ηi|}]TαΓ(α+1)δ=Γ(α+1)Γ(α+1)δ=δ. | (3.5) |
Therefore, PUδ⊂Uδ.
Let
u(t)=(σ1(t),σ2(t),...,σn(t),χ1(t),χ2(t),...,χn(t))T∈U(t), |
¯u(t)=(¯σ1(t),¯σ2(t),...,¯σn(t),¯χ1(t),¯χ2(t),...,¯χn(t))T∈U(t). |
Then,
‖Pu−P¯u‖=sup0≤t≤tn∑i=1{|Pσi)(t)−(P¯σi)(t)|+|Pχi)(t)−(P¯χi)(t)|}≤sup0≤t≤Tn∑i=1{ηiD−αt|σi(t)−¯σi(t)|+D−αt|χi(t)−¯χi(t)|+ηi|ηi−γi|D−αt|σi(t)−¯σi(t)|+|γi−ηi|D−αt|χi(t)−¯χi(t)|+D−αt|αi(σi(t))hi(σi(t))−αi(¯σi(t))hi(¯σi(t))|+D−αt|αi(σi(t))|n∑j=1|aij||fj(σj(t))−fj(¯σj(t))|+D−αt|αi(σi(t))−αi(¯σi(t))|n∑j=1|aij||fj(¯σj(t))|+D−αt|αi(σi(t))−αi(¯σi(t))||Ii(t)|}≤sup0≤t≤tn∑i=1{[ηi+ηi|ηi−γi|+¯pi+n∑j=1¯αj|aji|li+n∑j=1Aj|aji|¯fi+AiIi]1Γ(α)∫t0(t−z)α−1|σi(z)−¯σi(z)|dz+(1+|γi−ηi|)1Γ(α)∫t0(t−z)α−1|χi(z)−¯χi(z)|dz}≤max1≤i≤n{ηi+ηi|ηi−γi|+¯pi+n∑j=1¯αj|aji|li+n∑j=1Aj|aji|¯fi+AiIi}1Γ(α)∫t0(t−z)α−1‖u−¯u‖dz+max1≤i≤n{1+|γi−ηi|}1Γ(α)∫t0(t−z)α−1‖u−¯u‖dz≤[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi+n∑j=1¯αj|aji|li+n∑j=1Aj|aji|¯fi+AiIi}+max1≤i≤n{1+|γi−ηi|}]TαΓ(α+1)‖u−¯u‖≤‖u−¯u‖. | (3.6) |
One can see from (3.6) that P is a contraction mapping. Hence, there exists a unique fixed point u∗(t), which satisfies P(u∗(t))=u∗(t). That is,
{σ∗i(t)=−ηiD−αtσ∗i(t)+D−αtχ∗i(t),χ∗i(t)=−ηi(ηi−γi)D−αtσ∗i(t)−(γi−ηi)D−αtχ∗i(t)−D−αtαi(σ∗i(t))[hi(σ∗i(t))−n∑j=1aijfj(σ∗j(t)−Ii(t)]. |
Therefore, the system described by (1.1) has a unique solution.
Theorem 3.2. If H1−H4 are satisfied and
Γ(α+1)>[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi+AiIi}+max1≤i≤n{1+|γi−ηi|}]Tα, |
then system (1.1) has at least one solution in [0,T].
Proof. Let
U(t)={u(t)|u(t)=(σ1(t),...,σn(t),χ1(t),...,χn(t))T,σi(t),χi(t)∈C[0,t],i=1,2,...,n}, |
V(t)={v(t)|v(t)=(0,0,...,0,v1(t),v2(t),...,vn(t))T}, |
one can see that U(t) and V(t) are Banach spaces with the norms
‖u‖=sup0≤t≤Tn∑i=1(|σi(t)|+|χi(t)|),‖v‖=sup0≤t≤Tn∑i=1|vi(t)| |
separately.
Respectively define operators Q:U(t)→U(t) and R:V(t)→V(t) as follows:
(Qu)(t)=((Qσ1)(t),(Qσ2)(t),...,(Qσn)(t),(Qχ1)(t),(Qχ2)(t),...,(Qχn)(t))T, |
(Rv)(t)=(0,0,...,0,(Rv1)(t),(Rv2)(t),...,(Rvn)(t))T, where
{(Qσi)(t)=−ηiD−αtσi(t)+D−αtχi(t),(Qχi)(t)=−ηi(ηi−γi)D−αtσi(t)−(γi−ηi)D−αtχi(t)−D−αtαi(σi(t))(hi(σi(t))−Ii(t)),(Rvi)(t)=D−αtαi(σi(t))n∑j=1aijfj(σj(t),i=1,2,...,n. | (3.7) |
Let
δ≥n∑i=1¯αi[|hi(0)|+Ii+n∑j=1aij¯fj]TαΓ(α+1)−[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}+max1≤i≤n{1+|γi−ηi|}]Tα, |
Bδ={u(t)∈U(t),v(t)∈V(t):‖u‖≤δ,‖v‖≤δ}. |
It will begin with the proof that Qu+Rv∈Bδ, where u,v∈Bδ.
Based on (3.2),
‖Qu+Rv‖≤‖Qu‖+‖Rv‖≤sup0≤t≤Tn∑i=1{ηiD−αt|σi(t)|+D−αt|χi(t)|+ηi|ηi−γi|D−αt|σi(t)|+|γi−ηi|D−αt|χi(t)|+D−αt|αi(σi(t))(hi(σi(t))−Ii(t))|}+sup0≤t≤Tn∑i=1{D−αt|αi(σi(t))|n∑j=1aij|fj(σj(t)|}≤sup0≤t≤Tn∑i=1{(ηi++ηi|ηi−γi|+¯pi)D−αt|σi(t)|+(1+|γi−ηi|)D−αt|χi(t)|+D−αt¯αi(|hi(0)|+Ii)}+sup0≤t≤Tn∑i=1{n∑j=1D−αt¯αi|aij|¯fj}≤max1≤i≤n{ηi++ηi|ηi−γi|+¯pi}1Γ(α)∫t0(t−z)α−1‖u‖dz+max1≤i≤n{1+|γi−ηi|}1Γ(α)∫t0(t−z)α−1‖u‖dz+n∑i=1¯αi(|hi(0)|+Ii)1Γ(α)∫t0(t−z)α−1dz+n∑i=1¯αin∑j=1|aij|¯fj1Γ(α)∫t0(t−z)α−1dz≤max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}TαΓ(α+1)‖u‖+max1≤i≤n{1+|γi−ηi|}TαΓ(α+1)‖u‖+n∑i=1¯αi(|hi(0)|+Ii+n∑j=1|aij|¯fj)TαΓ(α+1)=[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi}+max1≤i≤n{1+|γi−ηi|}]TαδΓ(α+1)+n∑i=1¯αi(|hi(0)|+Ii+n∑j=1|aij|¯fj)TαΓ(α+1)≤δ, |
then Qu+Rv∈Bδ.
Next, let u(t)∈Bδ,
¯U(t)={¯u(t)|¯u(t)=(¯σ1(t),¯σ2(t),...,¯σn(t),¯χ1(t),¯χ2(t),...,¯χn(t))T∈Bδ},‖Qu−Q¯u‖≤sup0≤t≤Tn∑i=1{ηiD−αt|σi(t)−¯σi(t)|+D−αt|χi(t)−¯χi(t)|+ηi|ηi−γi|D−αt|σi(t)−¯σi(t)|+|γi−ηi|D−αt|χi(t)−¯χi(t)|+D−αt|αi(σi(t))hi(σi(t))−αi(¯σi(t))hi(¯σi(t))|+D−αt|αi(σi(t))−αi(¯σi(t))|Ii(t)}≤sup0≤t≤Tn∑i=1{[ηi+ηi|ηi−γi|+¯pi+AiIi]D−αt|σi(t)−¯σi(t)|+(1+|γi−ηi|)D−αt|χi(t)−¯χi(t)|}≤max1≤i≤n{ηi+ηi|ηi−γi|+¯pi+AiIi}1Γ(α)∫t0(t−z)α−1‖u−¯u‖dz+max1≤i≤n{1+|γi−ηi|}1Γ(α)∫t0(t−z)α−1‖u−¯u‖dz=[max1≤i≤n{ηi+ηi|ηi−γi|+¯pi+AiIi}+max1≤i≤n{1+|γi−ηi|}]TαΓ(α+1)‖u−¯u‖≤‖u−¯u‖. | (3.8) |
Therefore, Q is a contraction mapping.
Then, it comes to prove that R is continuous and compact.
Since αi(⋅) and fj(⋅) are continuous, one can see that R is continuous.
Let v(t)∈Bδ, one has
‖(Rv)(t)‖≤sup0≤t≤Tn∑i=1{D−αt|αi(σi(t))|n∑j=1|aij|fj(σj(t))|}≤n∑i=1D−αt¯αin∑j=1|aij|¯fj=n∑i=1¯αin∑j=1|aij|¯fj1Γ(α)∫t0(t−z)α−1dz≤n∑i=1n∑j=1¯αi|aij|¯fjTαΓ(α+1), |
which means that R is uniformly bounded in Bδ.
Next, it will be proved that (Qv)(t) is uniformly continuous.
In fact, for v∈Bδ, when 0<t1<t2, one has
|(Rvi)(t2)−(Rvi)(t1)|=|1Γ(α)∫t20(t2−z)α−1αi(σi(z))n∑j=1aijfj(σj(z))dz−1Γ(α)∫t10(t1−z)α−1αi(σi(z))n∑j=1aijfj(σj(z))dz|=|1Γ(α)∫t10(t2−z)α−1αi(σi(z))n∑j=1aijfj(σj(z))dz+1Γ(α)∫t2t1(t2−z)α−1αi(σi(z))n∑j=1aijfj(σj(z))dz−1Γ(α)∫t10(t1−z)α−1αi(σi(z))n∑j=1aijfj(σj(z))dz|=|1Γ(α)∫t10[(t2−z)α−1−(t1−z)α−1]αi(σi(z))n∑j=1aijfj(σj(z))dz+1Γ(α)∫t2t1(t2−z)α−1αi(σi(z))n∑j=1aijfj(σj(z))dz|≤1Γ(α)∫t10[(t2−z)α−1−(t1−z)α−1]|αi(σi(z))|n∑j=1|aij||fj(σj(z))|dz+1Γ(α)∫t2t1(t2−z)α−1|αi(σi(z))|n∑j=1|aij||fj(σj(z))|dz=¯αin∑j=1|aij|¯fj{1Γ(α)∫t10[(t1−z)α−1−(t2−z)α−1]dz+1Γ(α)∫t2t1(t2−z)α−1dz}=¯αin∑j=1|aij|¯fj1Γ(α+1)[tα1+(t2−t1)α−tα2+(t2−t1)α]=¯αin∑j=1|aij|¯fj1Γ(α+1)[tα1+2(t2−t1)α−tα2]. | (3.9) |
The right side of (3.9) tends to 0 when t1→t2. Therefore, R(Bδ) is a compact set. It comes out from the Arzela-Ascoli theorem that R is compact. Therefore, Q is a contraction mapping and R is continuous and compact. According to Lemma 3, there exists at least one u∗(t)∈Bδ such that Qu∗(t)+Ru∗(t)=u∗(t). That is,
{σ∗i(t)=−ηiD−αtσ∗i(t)+D−αtχ∗i(t),χ∗i(t)=−ηi(ηi−γi)D−αtσ∗i(t)−(βi−ηi)D−αtχ∗i(t)−D−αtαi(σ∗i(t))[hi(σ∗i(t))−n∑j=1aijfj(σ∗j(t))−Ii(t)], | (3.10) |
which means that the system described by (1.1) has at least one solution.
Theorem 3.3. If the assumptions H1−H4 are satisfied, Ii(t+ω)=Ii(t) for a certain constant ω>0, and
1+ηi(γi−ηi)−¯pi>0, |
δ=min1≤i≤n{2ηi+p_i−[1+ηi(γi−ηi)]−Ain∑j=1|aij|¯fj−n∑j=1|aji|li¯αj, |
2(γi−ηi)+p_i−[1+ηi(γi−ηi)]−Ain∑j=1|aij|¯fj−n∑j=1|aij|lj¯αi−2¯αiIi}>0, |
i=1,2,...,n, then the solution of the system described by (1.1) is S-asymptotic ω-periodic, where 0<T<∞.
Proof. Let σT(t)=(σ1(t),σ2(t),⋯,σn(t)) be a solution of the system given by (1.1) with the initial values σi(s)=φi(s),(i=1,2,⋯,n, s≤0).
From the assumptions H1 and H3,
αi(σi(t+ω))−αi(σi(t))=α′i(ξ∗i)[σi(t+ω)−σi(t)], |
αi(σi(t+ω))hi(σi(t+ω))−αi(σi(t))hi(σi(t))=p′i(ξi)[σi(t+ω)−σi(t)], |
where ξi is between 0 and ζi.
From the conditions in Theorem 3.3,
1+ηi(γi−ηi)−p_i≥1+ηi(γi−ηi)−p′i(ξi)≥1+ηi(γi−ηi)−¯pi>0, |
then it follows from (3.2) and Lemma 5 that
Dαt[σi(t+ω)−σi(t)]2≤2[σi(t+ω)−σi(t)]Dαt[σi(t+ω)−σi(t)]=2[σi(t+ω)−σi(t)][−ηiσi(t+ω)+χi(t+ω)+ηiσi(t)−χi(t)]=−2ηi[σi(t+ω)−σi(t)]2+2[σi(t+ω)−σi(t)][χi(t+ω)−χi(t)]Dαt[χi(t+ω)−χi(t)]2≤2[χi(t+ω)−χi(t)]Dαt[χi(t+ω)−χi(t)]=2[χi(t+ω)−χi(t)]{−ηi(ηi−γi)[σi(t+ω)−σi(t)]−(γi−ηi)[χi(t+ω)−χi(t)]−[αi(σi(t+ω))hi(σi(t+ω))−αi(σi(t))hi(σi(t))]+[αi(σi(t+ω))n∑j=1aijfj(σj(t+ω))−αi(σi(t))n∑j=1aijfj(σj(t))]+[αi(σi(t+ω))Ii(t+ω)−αi(σi(t))Ii(t)]}=2[χi(t+ω)−χi(t)]{−ηi(ηi−γi)[σi(t+ω)−σi(t)]−(γi−ηi)[χi(t+ω)−χi(t)]−p′i(ξi)[σi(t+ω)−σi(t)]+[αi(σi(t+ω))n∑j=1aij[fj(σj(t+ω))−fj(σj(t))]+[αi(σi(t+ω))−αi(σi(t))]n∑j=1aijfj(σj(t))+[αi(σi(t+ω))−αi(σi(t))]Ii(t)}. |
Therefore,
Dαt{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}=−2ηi[σi(t+ω)−σi(t)]2−2(γi−ηi)[χi(t+ω)−χi(t)]2+[2+2ηi(γi−ηi)−2p′i(ξi)][σi(t+ω)−σi(t)][χi(t+ω)−χi(t)]+2αi(σi(t+ω))[χi(t+ω)−χi(t)]n∑j=1aij[fj(σj(t+ω))−fj(σj(t))]+2[αi(σi(t+ω))−αi(σi(t))][χi(t+ω)−χi(t)][n∑j=1aijfj(σj(t))+Ii(t)]≤−2ηi[σi(t+ω)−σi(t)]2−2(γi−ηi)[χi(t+ω)−χi(t)]2+[2+2ηi(γi−ηi)−2p_i]|σi(t+ω)−σi(t)||χi(t+ω)−χi(t)|+2¯αi|χi(t+ω)−χi(t)|n∑j=1|aij|lj|σj(t+ω)−σj(t)|+4¯αiIi|χi(t+ω)−χi(t)|+2Ai|σi(t+ω)−σi(t)||χi(t+ω)−χi(t)|n∑j=1|aij|¯fj≤−2ηi[σi(t+ω)−σi(t)]2+[1+ηi(γi−ηi)−p_i]{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}−2(γi−ηi)[χi(t+ω)−χi(t)]2+¯αin∑j=1|aij|lj[σj(t+ω)−σj(t)]2+¯αin∑j=1|aij|lj[χi(t+ω)−χi(t)]2+2¯αiIi+2¯αiIi[χi(t+ω)−χi(t)]2}+Ain∑j=1|aij|¯fj{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}={−2ηi+1+ηi(γi−ηi)−p_i+Ain∑j=1|aij|¯fj]}[σi(t+ω)−σi(t)]2+[1+ηi(γi−ηi)−p_i−2(γi−ηi)+Ain∑j=1|aij|¯fj+n∑j=1|aij|lj¯αi+2¯αiIi]][χi(t+ω)−χi(t)]2+n∑j=1|aij|lj¯αi[σj(t+ω)−σj(t)]2+2¯αiIi. | (3.11) |
From (3.11),
Dαtn∑i=1{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}≤n∑i=1[−2ηi+1+ηi(γi−ηi)−p_i+Ain∑j=1|aij|¯fj+n∑j=1|aji|li¯αj][σi(t+ω)−σi(t)]2+n∑i=1{[1+ηi(γi−ηi)]−2(γi−ηi)−p_i+Ain∑j=1|aij|¯fj+n∑j=1|aij|lj¯αi+2¯αiIi}[χi(t+ω)−χi(t)]2+n∑i=12¯αiIi≤−δn∑i=1{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}+A, | (3.12) |
where δ=min1≤i≤n{2ηi+p_i−[1+ηi(γi−ηi)]−Ain∑j=1|aij|¯fj−n∑j=1|aji|li¯αj,2(γi−ηi)+p_i−[1+ηi(γi−ηi)]−Ain∑j=1|aij|¯fj−n∑j=1|aij|lj¯αi−2¯αiIi}, and A=n∑i=12¯αiIi.
Therefore,
n∑i=1{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}≤−δD−αtn∑i=1{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}+D−αt(A)≤−δD−αtn∑i=1{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}+ATαΓ(α+1), |
for 0<t<T<+∞.
Let P(t)=n∑i=1{[σi(t+ω)−σi(t)]2+[χi(t+ω)−χi(t)]2}.
Obviously, P(0)>0, and
P(t)≤ −δD−αtP(t)+ATαΓ(α+1). |
From Lemma 4,
P(t)≤ ATαΓ(α+1)Eα(−δtα), |
then
n∑i=1{[σi(t+ω)−σi(t)]2}≤ATαΓ(α+1)Eα(−δtα). |
Therefore,
limt→∞[σi(t+ω)−σi(t)]2=0,i=1,2,...,n. |
from Definition 4, one sees that the solution of the system given by (1.1) is S−asymptotic ω−periodic.
Remark. Among Theorems 3.1–3.3, under the condition of Theorem 3.1, we reveal that the system given by (1.1) has a unique solution; under the condition of Theorem 3.2, we reveal that the system given by (1.1) has at least one solution; if the solution of the system given by (1.1) exists and the conditions in Theorem 3.3 are satisfied, we reveal that the solution of the system given by (1.1) is S-asymptotic ω-periodic.
In this section, a numerical example is simulated to verify the results.
Example. Consider the following FCGNNI:
D2αt(σi(t))=−γiDαt(σi(t))−αi(σi(t))[hi(σi(t))−2∑j=1aijfj(σj(t))−Ii(t)], | (4.1) |
where i=1,2, 0<α<1.
Let
α=0.75,γ1=1.5,γ2=1.3,η1=0.4,η2=0.2, |
α1(σ1)=320(2+11+σ21),α2(σ2)=53500(2−11+σ22), |
h1(σ1)=2.6σ1,h2(σ2)=6σ2, |
fj(σj)=0.1sin(σj(t)),(j=1,2), |
Ii=0.1cos2t3,i=1,2.ω=3π, |
a11=0.15,a12=0.1,a21=0.15,a22=0.15. |
Therefore, the parameters in Theorem 3.3 are
α_1=0.3,¯α1=0.45,α_2=0.106,¯α2=0.212,A1=0.15,A2=0.106, |
¯fj=lj=1,Ii=0.1,i,j=1,2,p_1=0.7,¯p1=1.17,p_2=1.1667,¯p2=1.1925. |
After calculating, one has
1+η1(γ1−η1)−¯p1}=0.27>0,1+η2(γ2−η2)−¯p2}=0.0275>0,
δ=min1≤i≤2{2ηi+p_i−[1+ηi(γi−ηi)]−Ai2∑j=1|aij|¯fj−2∑j=1|aji|li¯αj,
2(γi−ηi)+p_i−[1+ηi(γi−ηi)]−Ai2∑j=1|aij|¯fj−2∑j=1|aij|lj¯αi−2¯αiIi}=1.3827>0.
Due to Theorem 3.3, the solution of the system given by (4.1) is S-asymptotic 3π−periodic.
The states of σ1(t) and σ2(t) are shown in Figure 1. It is clear that the numerical simulation is consistent with Theorem 3.3.
To date, scholars have done a lot of research on the dynamic behavior of inertial neural networks, and most of them focused on the chaos, bifurcation, stability and synchronization dynamics of NNs with inertia. But the existence of the solution and the S-asymptotic ω-periodic have not been studied. In this article, the existence and the S-asymptotic ω-period of the solution in for FCGNNIs are discussed. By employing the contraction mapping principle and the differential mean-value theorem, the existence and the S-asymptotic ω-periodic of the solution in the system, namely, Theorems 3.1–3.3. Finally, an example is simulated to demonstrate the correctness and validity of the results, which has a particular significance in both theory and application. The idea is innovative and meaningful. Similarly, the stability of other types of FNNs can be studied by using the methods employed in the theorems, including fractional-order inertial BAM neural networks, fractional-order inertial BAM Cohen-Grossberg neural networks, etc.
Funding: Scientific Research Project of Zhejiang Education Department (No.Y202145903), Shaoxing University Yuanpei College (No.KY2020C01, KY2021C04) and Shaoxing University (No.2020LG1009).
We declare that we have no financial or personal relationships with other people or organizations that can inappropriately influence our work, and that there is no professional or other personal interest of any nature or kind in any product, service or company that could be construed as influencing the position presented in the manuscript.
Definition 1. [4] The R-L fractional-order integral of the function u(t) is defined as
t0D−qtu(t)=1Γ(q)∫tt0(t−r)q−1u(r)dr, | (5.1) |
where Γ(τ)=∫+∞0tτ−1e−tdt, where 0<q<1.
Definition 2. [4] The R-L fractional-order derivative is defined as
RLaDαxv(x)=Dn(aDα−nxv(x))=1Γ(n−α)dndxn∫xav(t)(x−t)α−n+1dt, | (5.2) |
where the order α is a positive real number, and n−1≤α≤n.
Definition 3. [4] The Mittag-Leffler function with one parameter is
Eρ(s)=∞∑k=0skΓ(kρ+1), | (5.3) |
where ρ>0 and s∈C.
Definition 4. [19] If p(t)∈Cb([0,+∞),R), there exists ω>0 such that limt→+∞[p(t+ω)−p(t)]=0, then p(t) is S−asymptotic ω−periodic and ω is an asymptotic period of p(t).
Lemma 1. [2] If s(t)∈Cr[0,+∞), n−1≤q<n,m−1≤p<m,n,m∈Z+,r=max{n,m}, then
(1)D−qtA=AtqΓ(q+1), where A is a constant.
(2)Dqt(ks1(t)+hs2(t))=kDqts1(t)+hDqts2(t), where k and h are constants.
(3)DptD−qts(t)=Dp−qts(t),D−qtD−pts(t)=D−p−qts(t).
Lemma 2. [26] If s(t)∈R is derivable in [0,δ](δ>0) and 0<q<1,n−1<p<n, then
(1)DptDqts(t)=Dp+qts(t),
(2)D−ptDqts(t)=D−p+qts(t).
Lemma 3. [29] Z is a Banach space, C is a closed convex and nonempty subset of Z, and G1,G2 are operators which satisfy
(1) G1x+G2y∈C whenever x,y∈C;
(2) G1 is compact and continuous;
(3) G2 is a contraction mapping;
then there exists x∈C which satisfies G1x+G2x=x.
Lemma 4. [24] If s(t)∈C[0,+∞), and there exist d1>0 and d2>0 which satisfy s(t)≤−d1D−q0,ts(t)+d2, then
s(t)≤d2Eq(−d1tq). | (5.4) |
Lemma 5. [26] If v(t) has a continuous derivative, then
12Dqtr2(t)≤v(t)Dqtv(t),0<q≤1. | (5.5) |
[1] | V. Lakshmikantham, S. Leela, J. V. Devi, Theory of fractional dynamic systems, Cambridge Scientific Publishers Ltd., 2009. |
[2] |
C. P. Li, W. H. Deng, Remarks on fractional derivatives, Appl. Math. Comput, 187 (2007), 777–784. https://doi.org/10.1016/j.amc.2006.08.163 doi: 10.1016/j.amc.2006.08.163
![]() |
[3] | A. Kilbas, H. M. SrivastavaH, J. J. Trujillo, Theory and applications of fractional differential equations, Amsterdan: Elsevier, 2006. |
[4] | I. Podlubny, Franctional differential equations, New York: Academic Press, 1999. |
[5] | O. Heaviside, Electromagnetic theory, New York: Chelsea, 1971. |
[6] |
H. Sun, A. Abdelwahab, B. Onaral, Linear approximation of transfer function with a pole of fractional power, IEEE T. Automat. Contr., 29 (1984), 441–444. https://doi.org/10.1109/TAC.1984.1103551 doi: 10.1109/TAC.1984.1103551
![]() |
[7] |
R. C. Koeller, Application of fractional calculus to the theory of viscoelasticity, J. Appl. Mech., 51 (1984), 299–307. https://doi.org/10.1115/1.3167616 doi: 10.1115/1.3167616
![]() |
[8] |
W. C. Chen, Nonlinear dynamics and chaos in a fractional-order finacial system, Chaos Solition. Fract., 36 (2008), 1305–1314. https://doi.org/10.1016/j.chaos.2006.07.051 doi: 10.1016/j.chaos.2006.07.051
![]() |
[9] |
T. J. Anastasio, The fractional-order dynamics of brainstem vestibulo-oculomotor neurons, Biolo. Cybern., 72 (1994), 69–79. https://doi.org/10.1007/BF00206239 doi: 10.1007/BF00206239
![]() |
[10] |
C. J. Xu, P. L. Li, On finite-time Stability for fractional-order neural networks with proportional delays, Neural Process. Lett., 50 (2019), 1241–1256. https://doi.org/10.1007/s11063-018-9917-2 doi: 10.1007/s11063-018-9917-2
![]() |
[11] |
J. D. Li, Z. B. Wu, N. J. Huang, Asymptotical stability of Riemann-Liouville fractional neutral-type delayed projective neural networks, Neural Process. Lett., 50 (2019), 565–579. https://doi.org/10.1007/s11063-019-10050-8 doi: 10.1007/s11063-019-10050-8
![]() |
[12] |
F. Y. Zhou, C. Y. Ma, Mittag-Leffler stability and global asymptotically ω−periodicity of fractional BAM neural networks with time-varying delays, Neural Process. Lett., 47 (2018), 71–98. https://doi.org/10.1007/s11063-017-9634-2 doi: 10.1007/s11063-017-9634-2
![]() |
[13] |
C. J. Xu, W. Zhang, C. Aouiti, Z. X. Liu, M. X. Liao, P. L. Li, Further investigation on bifurcation and their control of fractional-order BAM neural networks involving four neurons and multiple delays, Math. Method. Appl. Sci., 2021. https://doi.org/10.1002/mma.7581 doi: 10.1002/mma.7581
![]() |
[14] |
S. C. Xu, X. Y. Wang, X. L. Ye, A new fractional-order chaos system of Hopfield neural network and its application in image encryption, Chaos Soliton. Fract., 157 (2022), 111889. https://doi.org/10.1016/j.chaos.2022.111889 doi: 10.1016/j.chaos.2022.111889
![]() |
[15] |
Y. R. Liu, L. Wang, K. X. Gu, M. Li, Artificial Neural Network (ANN)–Bayesian probability framework (BPF) based method of dynamic force reconstruction under multi-source uncertainties, Knowl.-Based Syst., 237 (2022), 107796. https://doi.org/10.1016/j.knosys.2021.107796 doi: 10.1016/j.knosys.2021.107796
![]() |
[16] |
A. Mauro, F. Conti, F. Dodge, R. Schor, Subthreshold behavior and phenomenological impedance of the squid giant axon, J. Gen. Physiol., 55 (1970), 497–523. https://doi.org/10.1085/jgp.55.4.497 doi: 10.1085/jgp.55.4.497
![]() |
[17] |
D. E. Angelaki, M. J. Correia, Models of membrane resonance in pigeon semicircular canal type Ⅱ hair cells, Biol. Cybern., 65 (1991), 1–10. https://doi.org/10.1007/BF00197284 doi: 10.1007/BF00197284
![]() |
[18] | J. F. Ashmore, D. Attwell, Models for electrical tuning in hair cells, Proc. Royal Soc. B: Biol. Sci., 226 (1985), 325–344. |
[19] |
L. Ke, W. L. Li, Exponential synchronization in inertial Cohen-Grossberg neural networks with time delays, J. Franklin I., 356 (2019), 11285–11304. https://doi.org/10.1016/j.jfranklin.2019.07.027 doi: 10.1016/j.jfranklin.2019.07.027
![]() |
[20] |
L. Ke, W. L. Li, Exponential synchronization in inertial neural networks with time delays, Electronics, 8 (2019), 356. https://doi.org/10.3390/electronics8030356 doi: 10.3390/electronics8030356
![]() |
[21] |
S. Li, X. M. Wang, H. Y. Qin, X. M. Zhong, Synchronization criteria for neutral-type quaternion-valued neural networks with mixed delays, AIMS Mathematics, 6 (2021), 8044–8063. https://doi.org/10.3934/math.2021467 doi: 10.3934/math.2021467
![]() |
[22] |
F. C. Kong, Q. X. Zhu, R. Sakthivel, A. Mohammadzadeh, Fixed-time synchronization analysis for discontinuous fuzzy inertial neural networks with parameter uncertainties, Neurocomputing, 422 (2021), 295–313. https://doi.org/10.1016/j.neucom.2020.09.014 doi: 10.1016/j.neucom.2020.09.014
![]() |
[23] |
W. Zhang, J. T. Qi, Synchronization of coupled memristive inertial delayed neural networks with impulse and intermittent control, Neural Comput. Appl., 33 (2020), 7953–7964. https://doi.org/10.1007/s00521-020-05540-z doi: 10.1007/s00521-020-05540-z
![]() |
[24] |
L. Ke, Mittag-Leffler stability and asymptotic ω-periodicity of fractional inertial neural networks with time-delays, Neurocomputing, 465 (2021), 53–62. https://doi.org/10.1016/j.neucom.2021.08.121 doi: 10.1016/j.neucom.2021.08.121
![]() |
[25] |
H. R. Henriquez, M. Pjerri, P. Tboas, On S-asymptotically ω-periodicityic functions on Banach spaces and applications, J. Math. Anal. Appl., 343 (2008), 1119–1130. https://doi.org/10.1016/j.jmaa.2008.02.023 doi: 10.1016/j.jmaa.2008.02.023
![]() |
[26] |
Y. J. Gu, H. Wang, Y. G. Yu, Stability and synchronization for Riemann-Liouville fractional time-delayed inertial neural networks, Neurocomputing, 340 (2019), 270–280. https://doi.org/10.1016/j.neucom.2019.03.005 doi: 10.1016/j.neucom.2019.03.005
![]() |
[27] |
S. L. Zhang, M. L. Tang, X. G. Liu, Synchronization of a Riemann-Liouville fractional time-delayed neural network with two inertial terms, Circuits Syst. Signal Process., 40 (2021), 5280–5308. https://doi.org/10.1007/s00034-021-01717-6 doi: 10.1007/s00034-021-01717-6
![]() |
[28] |
M. A. Cohen, S. Grossberg, Absolute stability of global pattern formation and parallel memory storage by competitive neural networks, IEEE T. Syst. Man. Cy., SMC-13 (1983), 815–826. https://doi.org/10.1109/TSMC.1983.6313075 doi: 10.1109/TSMC.1983.6313075
![]() |
[29] | M. A. Krasnoselskii, Topological methods in the theory of nonlinear integral equations, 1964. |
1. | Zhiying Li, Wei Liu, The anti-periodic solutions of incommensurate fractional-order Cohen-Grossberg neural network with inertia, 2025, 10, 2473-6988, 3180, 10.3934/math.2025147 |