Professional Documents
Culture Documents
Dynamics of Fuzzy in Doubly-Measure Pseudo-Almost Periodic 2022
Dynamics of Fuzzy in Doubly-Measure Pseudo-Almost Periodic 2022
Dynamics of Fuzzy in Doubly-Measure Pseudo-Almost Periodic 2022
Frontiers
a r t i c l e i n f o a b s t r a c t
Article history: The study of dynamics behaviors of genetic regulatory networks (GRNs) is essential for the understanding
Received 12 March 2021 of living organisms at both molecular and cellular levels. This paper deals with a class of fuzzy genetic
Revised 6 August 2021
regulatory networks (FGRNs) with time varying-delays in leakage terms, time-varying discrete delays,
Accepted 23 November 2021
unbounded distributed delays, and doubly-measure pseudo-almost periodic parameters. Based on the
doubly-measure pseudo-almost periodic theory, exponential dichotomy, differential inequality, and the
Keywords: Banach fixed point theorem, we establish some sufficient conditions to support the existence and global
Fuzzy genetic regulatory networks exponential stability of doubly-measure pseudo-almost periodic solutions for the considered model. A
Doubly-measure pseudo-almost periodic numerical example along with a graphical illustration are presented to support our main results. The re-
solutions
sults of this paper are new and extend existing GRNs models using almost periodic and weighted pseudo-
Mixed delays
almost periodic functions to support a wider range of regulatory processes.
Leakage delays
Global exponential stability © 2021 Elsevier Ltd. All rights reserved.
1. Introduction such as how cells react with DNA damage [55], what is the dy-
namics of tumor suppressor protein [55]. It is also expected that
The genomics era is attracting more and more researchers and GRNs models will make a significant contribution for cancer bi-
changing the old paradigm of medicine and healthcare. The study ology, for instance [22] investigated the genome scale breast can-
of genome sequencing became an emerging research field focus- cer network and analyzed and validate the model and statistical
ing on the study of gene development and reactions to external hypothesis computationally. Another concrete application of GRNs
signals within their highly connected and complex networks. As inference was established to ovarian cancer in [42] where they
the genes in DNA (deoxyribonucleic acid) is about encodes a set demonstrate the complete process for computational systems in bi-
of protein molecules, which are the ”workhorses” of the cell, mod- ology research, and identified potential drug targets.
eling gene behavior means analyzing and manufacturing its cor- Mathematically speaking, there is no unified, exact description
responding protein. A gene development process consists of two of the natural process in GRNs modeling. All existing models are
phases; first, the DNA informations are transferred to a messenger about approximations aimed to capture all aspects of interest that
RNA (mRNA) molecule via an activity called transcription. Then, a have been availed from the performed analysis. In the literature,
translation phase is applied, during which, the pre-mRNA is pro- several types of mathematical models have been used to describe
cessed to create a mature mRNA molecule that can be transformed the GRNs, such as the Boolean model [28], the Bayesian networks
to build the protein molecule (polypeptide) encoded by the orig- model [24], the differential equation model [11,50], and the func-
inal gene. A DNA chromosome controls the process and regulates tional differential equation (FDE) model [58]. Among these mod-
the gene expression, it is known as genetic regulatory networks els, FDE model looks more appropriate to describe the gene regu-
GRNs. lation process due to its ability to consider time delays required
From a biological and medical perspectives, severals phe- to model the slow process of translation, transcription, translo-
nomenon related to gene pathways could be modeled and studied cation, and diffusion, especially in a eukaryotic cell and one can
throughout the gene regulatory networks. GRNs allowed to model see some experimental results in [30,37,46]. The time delay is in-
several processes of body feedback such as heart beat, metabolic evitable in modeling the gene regulation process, because GRNs
pathways, temperature breathing etc [40,53,55]. Further in sever- models without consideration of time delay can lead to wrong pre-
als cases GRNs models were helpful to answer biological questions diction. Moreover, time delays make the dynamics behaviors of
GRNs more complex and may lead to instability, oscillations, bi-
furcation, and chaos. Therefore, considerable attention has been
E-mail address: Moez.Ayachi@fsg.rnu.tn
https://doi.org/10.1016/j.chaos.2021.111659
0960-0779/© 2021 Elsevier Ltd. All rights reserved.
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
given to the study of dynamical behaviors of GRNs including vari- and global exponential stability of almost periodic and weighted
ous types of delays [3,7,10,13,18,29,32,39,41,43,47–49,52,59]. pseudo-almost periodic solutions of the following non-autonomous
Assuming that each macromolecule takes the time to translo- GRNs with discrete time-varying delays
cate from its place of synthesis to the location where it exerts an $ n
2
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
and feedback regulation delays, respectively; αi pt q and α˜ i pt q are the We denote the collection of all such functions by OpRn , ν1 , ν2 q.
leakage delays; Q j is the feedback kernel; Ji ptq is the basal rate;
Lemma 3 ([17]). If ν1 , ν2 P M satisfy (E0). Then the space
di j pt q, d˜i j pt q and ei j pt q, e˜i j pt q, are the elements of fuzzy feedback
pOpRn , ν1 , ν2 q, }.}8 q is a Banach space.
MIN template and fuzzy feedback MAX template, respectively; ^
and _ `denotes ˘ the fuzzy AND and fuzzy OR operations, respec- Definition 4 ([9,23]). A function u P BC pRn q is called almost
tively; bi j ptq 1ďi, jďn is the coupling matrix of the genetic net- periodic if, for any ε ą 0, the set T “ tω : }upt ` ωq ´ uptq} ă
works, which is defined as follows: for all t P Ru is relatively dense in R, i.e., for any ε ą 0, there is a
constant l “ lε ą 0 such that every interval of length lpε q contains
& >0, if transcription j is an activator of gene i,
$
a number η with the property that }upt ` ηq ´ uptq} ă ε , for each
bij pt q : “ 0, if there is no link from node j to i,
t P R. We denote by AP pRn q the collection of all almost periodic
ă 0, if transcription j is a repressor of gene i,
%
functions from R to Rn . Then pAP pRn q, } ¨ }8 q is a Banach space.
ˆ ˙h
j
Definition 5 ([17]). Let ν1 , ν2 P M. A function u P BC pRn q is called
x
θj
and g j pxq “ ˆ ˙h is the feedback regulation of the protein in doubly-measure pseudo-almost periodic, or pν1 , ν2 q-p.a.p. (for
j
1` θx
j short), if it can be expressed as u “ ua ` ue with ua P AP pRn q is
the transcription process, where h j is the Hill coefficient and θ j is the almost periodic component and ue P OpRn , ν1 , ν2 q is the er-
a positive constant. godic perturbation. We denote by PAP pRn , ν1 , ν2 q the collection of
The initial conditions of (1) are associated to be of the form all pν1 , ν2 q-p.a.p. functions from R to Rn .
mi psq “ φi psq, pi psq “ ψi psq, s P p´8, 0s , Remark 2. The concept of pν1 , ν2 q-pseudo almost periodicity gen-
where φi and ψi are continuously differentiable on p´8, 0s. eralizes; (1) the concept of almost periodicity introduced by Bohr
in [9], (2) the concept of the classical pseudo-almost periodicity
Remark 1. g j p.q is a monotonically increasing function with satu- introduced by Zhang in [56,57], (3) the concept of -weighted
ration, and satisfies g j p0q “ 0 and pseudo-almost periodicity introduced by Diagana in [15,16], as well
g j pxq ´ g j pyq as the concept of ν -pseudo almost periodicity introduced by Blot
0ď ď L j, @x ‰ y P R, j P vnw, et al. in [8], as what follows:
x´y
where 1. The pν1 , ν2 q-pseudo-almost periodicity coincides with the al-
ph j ´1q{h j ph j `1q{h j
most periodicity, when the ergodic component is equal to zero.
ph j ´ 1q ph j ` 1q 2. The pν1 , ν2 q-pseudo-almost periodicity coincides with the
L j fi max g1j pxq “ ą 0.
xě0 4θ j h j pseudo-almost periodicity, when ν1 „ ν2 and ν1 , ν2 are the
Lebesgue measures.
3. Notations and definitions 3. The pν1 , ν2 q-pseudo-almost periodicity coincides with the -
´ ¯ weighted pseudo-almost periodicity, when ptq ą 0 almost ev-
For convenience, Rn R1 “ R denotes the set of all n- erywhere on R for the Lebesgue measure, and ν1 „ ν2 defined
dimensional real vectors (real numbers). BC pRn q denotes the col- by
lection of all bounded continuous functions from R to Rn , and ż
BC 1 pRn q denotes the collection of all functions u : R ÝÑ Rn , such ν1 pBq “ pt qdt , for B P B,
B
that u, u1 P BC¯pRn q. It is well known that pBC pRn q, }.}8 q and
where dt denotes the Lebesgue measure on R. In particular, ac-
´
BC 1 pRn q, }.}1 are Banach spaces, where
cording to [8, Remark 3.4], condition (E1) is equivalent to
}u}8 fi sup }upt q} fi sup max |ui pt q|, pt ` sq
t PR t PR iPvnw (F1) For all s P R, lim sup ă `8.
|t|Ñ`8 ptq
}u}1 fi sup }upt q} “ max }u}8 , }u1 }8 .
(
t PR 4. The pν1 , ν2 q-pseudo almost periodicity coincides with the ν1 -
Let B denotes the Lebesgue σ -field of R, M denotes the col- pseudo almost periodicity, when ν1 „ ν2 .
lection of all positive measures ν on B satisfying ν pRq “ `8 and Example 1. Let the measure ν1 where its Radon-Nikodym deriva-
ν prt1 , t2 sq ă `8 for all t1 , t2 P R pt1 ď t2 q. Let us consider the class tive is 1 ptq “ ecosptq , t P R, and the measure ν2 ptq where its
of measures ν1 , ν2 P M which satisfies: Radon-Nikodym derivative is
(E0) Let ν1 , ν2 P M satisfy if t P r0, `8q,
" t
e
2 pt q “
ν1 pr´γ , γ sq 1 if t P p´8, 0q.
lim sup ă `8.
γ Ñ`8 ν2 pr´γ , γ sq For more details about the Radon-Nikodym derivative, we refer
the reader to [21].
(E1) For all s P R, there exists λ ą 0 and a bounded interval J of
First, it is easy to see that ν1 , ν2 P M.
R such that ν1 ptq ` s : q P Buq ď λν1 pBq, when B P B satisfies
Note that, for γ ą 0, we have
B X J “ H.
żγ
Definition 1 ([8]). Let ν1 , ν2 P M. We say that ν1 and ν2 are equiv- 2γ e´1 ď ν1 pr´γ , γ sq “ ecospt q dt ď 2γ e1 , (2)
´γ
alent (i.e. ν1 „ ν2 ) if there exists nonnegative constants c1 , c2 , and
a bounded interval J Ă R (eventually H) such that and
żγ
c1 ν1 pBq ď ν2 pBq ď c2 μ1 pBq, when B P B, B X J “ H. ν2 pr´γ , γ sq “ 2 pt qdt “ γ ` 1 ´ e´γ , (3)
´γ
Definition 2 ([17]). Let ν1 , ν2 P M. A function u P BC pRn q is said to
which implies that
be pν1 , ν2 q-ergodic if
şγ
1
żγ ν1 pr´γ , γ sq ´γ 1 pt qdt
lim }upt q}dν1 pt q “ 0. lim sup “ lim sup şγ ď 2e1 ă `8.
γ Ñ`8 ν2 pr´γ , γ sq ´γ γ Ñ`8 ν2 pr´γ , γ sq γ Ñ`8 ´γ 2 pt qdt
3
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
ψ p s q “ φ1 p s q , . . . , φ n p s q , ϕ 1 p s q , . . . , ϕ n p s q T ,
` ˘
1. u P OpRn , ν1 , ν2 q.
ż t ´ϑ a pt q
2 pt q “ ξ dξ ,
` ˘
1
ż u
2. lim }upt q}dν1 pt q “ 0. t ´ϑ pt q
γ Ñ`8 ν2 pr´γ , γ szJq r´γ ,γ szJ żt
3. For any nonnegative real number ω, 3 pt q “ ue ξ dξ .
` ˘
t ´ϑ a pt q
1
ż
lim dν1 pt q “ 0, Firstly, according to [51, Lemma 2.2.], since ϑ a and ua are almost
γ Ñ`8 ν2 pr´γ , γ szJ q γ ,ω puq
periodic, it follows that 1 belongs to AP pRn q.
where γ ,ω puq “ tt P r´γ , γ szJ : }uptq} ą ωu. Furthermore, because ϑ e P OpRn , ν1 , ν2 q, it follows that
żγ
Lemma 9 ([45]). Let ν1 , ν2 P M. If u1 , u2 P PAP pR, ν1 , ν2 q, then lim
1
}2 pt q}dν1 pt q
u1 ˆ u2 belongs to PAP pR, ν1 , ν2 q. γ Ñ`8 ν2 pr´γ , γ sq ´γ
ż γ › ż t ´ϑ a pt q
1
›
Lemma 10 ([54]). Suppose that si j , s˜i j , x j , y j P R and f j : R ÝÑ R be ξ ξ ›dν1 pt q
› ›
“ lim u p q d
γ Ñ`8 ν2 pr´γ , γ sq ´γ › t ´ϑ pt q
›
continuous functions, i, j P vnw. Then the following inequality holds:
›
żγ
1
|ϑ e pt q|dν1 pt q
ˇ n n n
ď }u}8 lim
ˇ
ˇsi j ˇˇ f j px j q ´ f j py j qˇ, γ Ñ`8 ν2 pr´γ , γ sq ´γ
ˇľ ľ ˇ ÿ ˇ ˇˇ ˇ
ˇ
ˇ si j f j px j q ´ si j f j py j qˇˇ ď
j“1 j“1 j“1 “ 0.
ˇ n n ˇ n
ˇs˜i j ˇˇ f j px j q ´ f j py j qˇ. Which implies that 2 P OpRn , ν1 , ν2 q.
ˇł ł ˇ ÿ ˇ ˇˇ ˇ
ˇ ˜
s f
ij j p x j q ´ ˜
s f
ij j p y q
j ˇ ď
ˇ
Finally, it remains to prove that 3 P OpRn , ν1 , ν2 q. In fact, by
ˇ
j“1 j“1 j“1
using Fubini’s theorem, it follows that
Definition 11. Let żγ
1
pm˚ pt q, p˚ pt qqT “ pm˚ T } pt q}dν1 pt q
1 pt q, . . . , mn pt q, p1 pt q, . . . , pn pt qq
˚ ˚ ˚
ν2 pr´γ , γ sq ´γ 3
4
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
ż γ ›ż t
1 1
›
e
ξ ξ ›dν1 pt q “ lim pvpt ` ηq ´ vpt qq
› ›
“ u p q d
ν2 pr´γ , γ sq ´γ › t ´ϑ a pt q η
›
› ηÑ0
1
ż γ ż t `ϑ a` “ v1 pt q,
ď }ue pξ q}dξ dν1 pt q 1
ż t `η
ν2 pr´γ , γ sq ´γ t ´ϑ a` ϕ pt q “ lim ϕ psqds
ż γ ż ϑ a` ηÑ0 η t
1
“ }ue pt ` ξ q}dξ dν1 pt q 1
ν2 pr´γ , γ sq ´γ ´ϑ a` “ lim pwpt ` ηq ´ wpt qq
ηÑ0 η
ż ϑ a` ż γ
1 “ w1 pt q.
“ }ue pt ` ξ q}dν1 pt qdξ
ν2 pr´γ , γ sq ´ϑ a` ´γ
The proof is completed. ˝
ż ϑ a`
“ Tγ pξ qdξ , Lemma 14. Let ν1 , ν2 P M satisfying (E0) and (E1). Let
´ϑ a
`
Lemma 13. Let ν1 , ν2 P M satisfying (E0) and (E1). Suppose that Let u “ v ` w and g “ φ ` ϕ, then u and g belongs to
both functions u and u1 belongs to PAP pRn , ν1 , ν2 q. That is, u “ v ` PAP pRn , ν1 , ν2 q and
w and u1 “ φ ` ϕ , where v, φ P AP pRn q, and w, ϕ P OpRn , ν1 , ν2 q.
Then the functions v and w are differentiable and
lim }uk ´ u}8 “ 0, lim }u1k ´ g}8 “ 0. (6)
kÑ`8 kÑ`8
5
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
(T2) There exists ζ˜ P CpR` q, such that Which implies that u2 P OpR, ν1 , ν2 q.
Step 3. We claim that u3 P OpR, ν1 , ν2 q.
dν1 ζ pt q “ ζ˜ pt qdν1 pt q @t P R,
` ˘
In view of condition (E0), we can find γ0 enough large positive
and real number and a constant m ą 0, such that
Mpγ qν2 pr´κ pγ q, κ pγ qsq ν1 pr´γ , γ sq
limsup ă `8, @γ ą γ0 , ă m.
γ Ñ`8 ν2 pr´γ , γ sq ν2 pr´γ , γ sq
where Further, since ua is almost periodic, then it is uniformly continuous
on R, that is;
M pγ q “ ˇζ pt qˇ and κ pγ q “ sup |t ´ ηpt q|.
ˇ ˇ
ˇ˜ ˇ
sup
t Pr´κ pγ q,κ pγ qs t Pr´γ ,γ s
@ ą 0, Dδ ą 0, s.t. @t, s P R, |t ´ s| ă δ ùñ |ua pt q ´ ua psq| ă .
2m
Then up. ´ ηp.qq P PAP pR, ν1 , ν2 q.
On the other hand, since ηe P OpR, ν1 , ν2 q, then from Lemma 8 we
Proof. According to Definition 5, since u and η are pν1 , ν2 q-p.a.p., know that, for any ω ą 0, we have
then there exists ua , ηa P AP pR, ν1 , ν2 q and ue , ηe P OpR, ν1 , ν2 q, 1
ż
such that lim dν1 pt q “ 0,
γ Ñ`8 ν2 pr´γ , γ sq γ ,ω pηe q
u “ ua ` ue and η “ ηa ` ηe .
where
Then, we can write
γ ,ω ηe “ t P r´γ , γ s :|ηpt q ´ ηa pt q| “ˇηe pt q| ą ω .
` ˘ ˇ (
upt ´ ηpt qq “ u1 pt q ` u2 pt q ` u3 pt q
In particular, we can find γ1 ą 0 such that
“ ua t ´ ηa pt q ` ue pt ´ ηpt qq
` ˘
1
ż
` u pt ´ ηpt qq ´ ua t ´ ηa pt q @γ ą γ1 , dν pt q ď
“ a
.
` ˘‰
ν2 pr´γ , γ sq γ ,ω pηe q 1 4 }u a }8
Our proof consists of three steps:
Step 1. We claim that u1 P AP pRq. Consequently, for all γ ą maxtγ0 , γ1 u, we obtain that
Indeed, since ua and ηa belongs to AP pRq, then by using [12,
żγ
1
Lemma 2.4.], we can conclude that u1 also belongs to AP pRq. |u pt q|dν1 pt q
ν2 pr´γ , γ sq γ 3
Step 2. We claim that u2 P OpR, ν1 , ν2 q. żγ
1
|ua pt ´ ηpt qq ´ ua t ´ ηa pt q |dν1 pt q
` ˘
By using (T1) and (T2), we obtain that “
żγ ν2 pr´γ , γ sq γ
1
|u pt q|dν1 pt q 1
ż
ν2 pr´γ , γ sq ´γ 2 |ua pt ´ ηpt qq ´ ua t ´ ηa pt q |dν1 pt q
` ˘
ď
żγ ν2 pr´γ , γ sq γ , pηe q
1
|ue pt ´ ηpt qq|dν1 pt q 1
ż
“
ν2 pr´γ , γ sq ´γ |ua pt ´ ηpt qq ´ ua t ´ ηa pt q |dν1 pt q
` ˘
`
ν2 pr´γ , γ sq r´γ ,γ szγ , pηe q
1
ż
|ue pt q|ζ˜ pt qdν1 pt q 1
ż
“
ν2 pr´γ , γ sq pI´ηqpr´γ ,γ sq ď 2 }u a }8 dν pt q
ż κ pγ q ν2 pr´γ , γ sq γ , pηe q 1
1
|ue pt q|ζ˜ pt qdν1 pt q 1
ż
ď
ν2 pr´γ , γ sq ´κ pγ q ` dν1 pt q
2m ν2 pr´γ , γ sq r´γ ,γ szγ , pηe q
M pγ q
ż κ pγ q
ď |ue pt q|dν1 pt q. 1
ż
ν1 pr´γ , γ sq
ν2 pr´γ , γ sq ´κ pγ q ď 2 }u a }8 dν pt q `
ν2 pr´γ , γ sq γ , pηe q 1 2m ν2 pr´γ , γ sq
Here, we distinguish two possible cases:
Case 1. Suppose that lim κ pγ q ă `8, then Mpγ q and ď 2 }u a }8 ` m
γ Ñ`8 4 }u a }8 2m
ν1 pr´κ pγ q, κ pγ qsq are finite. Furthermore, since is pν1 , ν2 q- ue “ ,
ergodic then it is bounded on R. Consequently, we obtain
żγ which implies that
1
lim |u2 pt q|dν1 pt q 1
żγ
γ Ñ`8 ν2 pr´γ , γ sq ´γ lim |u3 pt q|dν1 pt q “ 0,
γ Ñ`8 ν2 pr´γ , γ sq ´γ
1
ď lim Mpγ qν1 pr´κ pγ q, κ pγ qsq}u }8 e
that is u3 P OpR, ν1 , ν2 q.
γ Ñ`8 ν2 pr´γ , γ sq
Cst The proof is completed. ˝
ď lim
γ Ñ`8 ν2 pr´γ , γ sq Remark 3. Lemma 15 improve the result obtained in [38].
“ 0.
Lemma 16. Let ν1 , ν2 P M. Suppose that, for all i, j P vnw, κi j and
Case 2. Suppose that lim κ pγ q “ `8. Then considering (T2) and u j belongs to PAP pR, ν1 , ν2 q. Then for all i P vnw, the functions
γ Ñ`8
Fi : t ÞÝÑ nj“1 κi j pt qu j pt q and F˜i : t ÞÝÑ nj“1 κi j pt qu j pt q, belongs to
Ź Ž
the fact that ue P OpR, ν1 , ν2 q, it follows that
żγ PAP pR, ν1 , ν2 q.
1
lim |u2 pt q|dν1 pt q Proof. Let us show that Fi belongs to PAP pR, ν1 , ν2 q. The second
γ Ñ`8 ν2 pr´γ , γ sq ´γ
assertion can be showed similarly.
Mpγ qν2 pr´κ pγ q, κ pγ qsq Suppose that κi j and u j belongs to PAP pR, ν1 , ν2 q for all
ď lim
γ Ñ`8 ν2 pr´γ , γ sq i, j P vnw. By Definition 5, there exists κiaj , uaj P AP pRq, and κiej , uej P
ż κ pγ q
1 OpR, ν1 , ν2 q, such that κi j “ κiaj ` κiej and u j “ uaj ` uej . Conse-
ˆ |ue pt q|dν1 pt q
ν2 pr´κ pγ q, κ pγ qsq ´κ pγ q quently, we can write
“ 0. Fi pt q “ Fi1 pt q ` Fi2 pt q ` Fi3 pt q
6
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
n n n
which implies that Fi2 P OpR, ν1 , ν2 q.
„ľ j
κiaj pt quaj pt q ` κiaj pt qu j pt q ´ κiaj pt quaj pt q
ľ ľ
“
Step 3. We claim that Fi3 P OpR, ν1 , ν2 q.
j“1 j“1 j“1
Let Nu “ sup max |u j ptq| ă 8. Again, from Lemma 10, we obtain
n
„ľ n j tPR iPvnw
κi j pt qu j pt q ´ κiaj pt qu j pt q
ľ
` that
j“1 j“1 żγ ˇ
1
ˇFi pt qˇdν1 pt q
ˇ
ˇ 3 ˇ
Our proof consists of three steps: ν2 pr´γ , γ sq ´γ
Step 1. We claim that Fi1 P AP pRq. ˇ ˇ
ż γ ˇľ
Since, for all i, j P vnw, κiaj and uaj are almost periodic, then 1 ˇ n n ˇ
κ κ a ˇdν1 pt q
ľ ˇ
“ pt q u pt q ´ p t q u pt q
ν2 pr´γ , γ sq ´γ ˇˇ j“1 i j j j
ˇ
ij
Mκ “ sup max |κiaj pt q| ă `8, Mu “ sup max |uaj pt q| ă `8.
ˇ
j“1 ˇ
t PR i, jPvnw t PR iPvnw żγ ÿ n
1 ˇu pt qˇˇˇκi j pt q ´ κ a pt qˇˇdν1 pt q
ˇ ˇˇ ˇ
Further, for any ą 0, there exists T a relatively dense subset of ď
ν2 pr´γ , γ sq ´γ j“1 j ij
R such that, for all t P R, ω P T and i, j P vnw, we have
żγ n
1
|αiaj pt ` ωq ´ κiaj pt q| ă Nu ˇκiej pt qˇdν1 pt q
ˇ ˇ
,
ÿ
(9) ď
ˇ ˇ
2 n pMκ ` Mu q ν2 pr´γ , γ sq ´γ j“1
n żγ ˇ
1
|uaj pt ` ωq ´ uaj pt q| ă ˇκi j pt qˇdν1 pt q.
ˇ
. (10) ˇ e ˇ
ÿ
“ Nu (12)
2 n pMκ ` Mu q
j“1
ν2 pr´γ , γ sq ´γ
By using Lemmas 10, (9) and (10), it follows that, for all ω P T
Since κiej P OpR, ν1 , ν2 q, i, j P vnw, then
ˇFi pt ` ωq ´ Fi1 pt qˇ
ˇ ˇ
ˇ 1 ˇ
żγ ˇ
1
ˇκi j pt qˇdν1 pt q “ 0,
ˇ
ˇ e ˇ
i, j P vnw,
ˇ ˇ
ˇ n n lim
ν2 pr´γ , γ sq
ˇ
γ Ñ`8
κi j pt ` ωqu j pt ` ωq ´ αi j pt qu j pt qˇˇ ´γ
ˇľ a a
ľ
a a
ˇ
“ˇˇ
ˇ j“1 j“1 thus, (12) gives
ˇ
ˇ ˇ
ˇ n n ˇ żγ ˇ
1
κi j pt ` ωqu j pt ` ωq ´ κi j pt ` ωqu j pt qˇˇ ˇFi pt qˇdν1 pt q “ 0,
ˇľ a ˇ
a a a
ľ ˇ ˇ 3 ˇ
ď ˇˇ lim
ˇ j“1 j“1 ˇ γ Ñ`8 ν2 pr´γ , γ sq ´γ
żγ ˇ n żt
1 Q j pt ´ ξ qg j pϕ j pξ qqdξ
ł
ˇu j pt qˇdν1 pt q “ 0,
ei j pt q
ˇ
lim
ˇ e ˇ
j P vnw, `
γ Ñ`8 ν2 pr´γ , γ sq ´γ j“1 ´8
7
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
n żt Lemma 17. Assume that (C3) hold, and for all j P vnw, m j , p j P
d˜i j pt q Q j pt ´ ξ qφ j pξ qdξ
ľ
` PAP pR, ν1 , ν2 q. Then for all j P vnw, the functions S j : t ÞÝÑ
j“1 ´8
´8 Q j pt ´ ξ qg j p p j pξ qqd ξ and S j : t ÞÝÑ ´8 Q j pt ´ ξ qm j pξ qd ξ , be-
şt şt
˜
n
longs to PAP pR, ν1 , ν2 q.
żt
Q j pt ´ ξ qφ j pξ qdξ .
ł
` e˜i j pt q
´8
j“1 Proof. Let us show that S j belongs to PAP pR, ν1 , ν2 q. The second
´ ¯T assertion can be showed similarly.
Let zψ ptq “ mφ1 ptq, . . . , mφn ptq, pϕ1 ptq, . . . , pϕn ptq , where Since, for all j P vnw, p j belongs to PAP pR, ν1 , ν2 q and g j
żt is Lipschitzian, then g j p p j p.qq P PAP pR, ν1 , ν2 q. Thus, there exists
ai pξ qd ξ uaj P AP pRq and uej P OpR, ν1 , ν2 q, such that g j p p j p.qq “ uaj p.q ` uej p.q,
şt
mφi pt q “ e´ s Fi ps, φ , ϕ qds, i P vnw,
´8
żt which implies that
a˜i pξ qdξ
şt
pϕi pt q “ e´ s Gi ps, φ , ϕ qds, i P vnw. S j pt q “ S aj pt q ` S ej pt q
´8
żt żt
Throughout the rest of this paper, we suppose that the measures “ Q j pt ´ ξ quai pξ qdξ ` Q j pt ´ ξ quei pξ qdξ .
ν1 , ν2 P M satisfy pE0q and (E1). We also formulate the following ´8 ´8
assumptions that will be adopted later: By using the same arguments given in [44], we can prove that, for
all j P vnw, S aj belongs to AP pRq.
(C1) For all i P vnw; ai and a˜i belongs to AP pR` q with To complete the proof, it is enough to show that S ej belongs to
a´ “ inf pai pt qq ą 0, a˜´ “ inf pa˜i pt qq ą 0. OpR, ν1 , ν2 q. In fact, Fubini’s Theorem gives
i t PR i t PR żγ
1 ˇS pt qˇdν1 pt q
ˇ e ˇ
(C2) For all i, j P vnw; bi j , di j , ei j , ci , d˜i j , e˜i j , Ji , belongs to ν2 pr´γ , γ sq ´γ j
PAP pR, ν1 , ν2 q, and αi , α˜ i belongs to PAP pR` , ν1 , ν2 q. ż γ ˇˇ ż t ˇ
(C3) For all j P vnw; the delay kernels Q j : r0, `8rÝÑ r0, `8r 1
Q j pt ´ ξ qu j pξ qdξ ˇdν1 pt q
ˇ
e
“
ˇ ˇ
ν2 pr´γ , γ sq ´γ ˇ ´8
ˇ
are bounded-continuous and satisfy ˇ
żγ żt
ż `8 1
Q j pξ qd ξ “ Q ` ă `8. ď |Q pt ´ ξ quej pξ q|dξ dν1 pt q
0
j ν2 pr´γ , γ sq ´γ ´8 j
ż γ ż `8
1
(C4) For all i P vnw; τi , ri P PAP pR` , ν1 , ν2 q, such that t ´ τi ptq, “ Q j pξ q|uej pt ´ ξ q|dξ dν1 pt q
t ´ ri ptq are invertible functions with inverse resp. ζi1 , ζi2 , ν2 pr´γ , γ sq ´γ 0
ż `8 żγ
and there exists continuous functions ζ˜ 1 , ζ˜ 2 : R ÝÑ R` 1
i i “ Q j pξ q |uej pt ´ ξ q|dν1 pt qdξ
such that for all t P R and i P vnw ν2 pr´γ , γ sq 0 ´γ
ż `8
dν1 pζi1 pt qq “ ζ˜i1 pt qdν1 pt q, “ T j,γ pξ qdξ ,
0
dν1 pζi2 pt qq “ ζ˜i2 pt qdν1 pt q,
where
żγ
and 1
T j,γ pξ q “ Q j pξ q |uej pt ´ ξ q|dν1 pt q.
Mil pγ qν2 ´κil pγ q, κil pγ q
´” ı¯
ν2 pr´γ , γ sq ´γ
lim sup ă `8, l “ 1, 2, Since ν1 , ν2 , satisfy (E1), then from Lemma 6, we have uej p. ´ ξ q P
γ Ñ`8 ν2 pr´γ , γ sq
OpR, ν1 , ν2 q for every ξ P R and j P vnw. Consequently
where
lim T j,γ pξ q
γ Ñ`8
Mi1 pγ q “ sup ζ˜i1 pt q,
t Pr´κi1 pγ q,κi1 pγ qs Further, we have
Mi2 p γq “ sup ζ˜i2 pt q, ν1 pr´γ , γ sq
t Pr´κi2 pγ q,κi2 pγ qs 0 ď T j,γ pξ q ď }uej }8 Q pξ q, @ξ ě 0, j P vnw.
ν2 pr´γ , γ sq j
κ γ q “ sup
1
i p |t ´ τi pt q|, Then, using (E0), (C3), and Lebesgue’s dominated convergence The-
t Pr´γ ,γ s
orem, it follows that
κi2 pγ q “ sup |t ´ ri pt q|. ż `8
t Pr´γ ,γ s
lim T j,γ pξ qdξ “ 0, @ j P vnw,
γ Ñ`8 0
(C5) There exists nonnegative constant such that
which implies that
ςi a` ς˜i a˜`
" *
żγ
“ max ςi , ς˜i
´ ¯ ´ ¯
i i 1
, 1` , 1` ă 1, ˇS pt qˇdν1 pt q “ 0, @ j P vnw.
ˇ e ˇ
a´ a´ a˜´ a˜´ lim
iPvnw
i i i i γ Ñ`8 ν2 pr´γ , γ sq ´γ
j
8
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
that Fi ps, φ , ϕ q “ Fia psq ` Fie psq and Gi ps, φ , ϕ q “ Gai psq ` Gei psq, for which implies that
all i P vnw. Therefore, mφi ptq “ m1φ ptq ` m2φ ptq and pϕi ptq “ p1ϕi ptq ` 1
żγ ˇ
ˇmφi pt qˇdν1 pt q “ 0,
ˇ
ˇ 2
i i
lim @i P vnw.
ˇ
p2
ϕi ptq, where γ Ñ`8 ν2 pr´γ , γ sq ´γ
That is m2φ P OpR, ν1 , ν2 q for all i P vnw. Similarly, we can show that
żt
ai pξ qd ξ
şt
m1φi pt q “ e´ s Fia psqds, i
´8 p2ϕi P OpR, ν1 , ν2 q for all i P vnw. Consequently zψ2 P OpR2n , ν , ν q.
1 2
żt
2
şt
ai pξ qd ξ Since zψ 1 P AP pR2n q and z2 P OpR2n , ν , ν q, we can conclude
mφi pt q “ e´ s Fie psqds, ψ 1 2
´8
żt that zψ P PAP pR2n , ν1 , ν2 q.
şt
a˜i pξ qdξ Finally, since for all i P vnw
p1ϕi pt q “ e´ s Gai psqds,
´8
żt m1φi pt q “ ´ai pt qmφi pt q ` Fi pt, φ , ϕ q P PAP pR, ν1 , ν2 q,
a˜i pξ qdξ
şt
p2ϕi pt q “ e´ s Gei psqds. p1ϕi pt q “ ´a˜i pt q pϕi pt q ` Gi pt, φ , ϕ q P PAP pR, ν1 , ν2 q,
´8
Consequently, we can write we can deduce that zψ P PAP 1 pR2n , ν1 , ν2 q, which complete the
1 2 proof. ˝
zψ pt q “ zψ pt q ` zψ pt q
´ ¯ Theorem 19. Assume that assumptions (C1)–(C5) hold. Then Model
“ m1φ1 pt q, . . . , m1φn pt q, p1ϕ1 pt q, . . . , p1ϕn pt q
(1) has a unique pν1 , ν2 q-p.a.p. solution in the region
´ ¯
` m2φ1 pt q, . . . , m2φn pt q, p2ϕ1 pt q, . . . , p2ϕn pt q L
" *
Y “ ψ P PAP 1 pR2n , ν1 , ν2 q : }ψ ´ ψ0 }1 ď ,
1 P AP pR2n q. In view of (C1), the linear sys- 1´
Let us show that zψ
tem where
mi pt q “ ´ai pt qmi pt q, ψ0 pt q “ φ0,1 pt q, . . . , φ0,n pt q, 0, . . . , 0 T ,
" 1 ` ˘
(13)
p1i pt q “ ´a˜i pt qmi pt q, i P vnw żt
ai pξ qd ξ
şt
admits an exponential dichotomy. Since for all i P vnw, Fia and Gai φ0,i pt q “ e´ s Ji psqds, i P vnw,
´8
belongs to AP pRq, then by using the theory of exponential di-
and
chotomy [23], the almost periodic system
Ji` a`
ˆ ˆ ˙ ˙
mi pt q “ ´ai pt qmi pt q ` Fia pt q,
" 1 i
(14) L “ max , 1` Ji` .
p1i pt q “ ´a˜i pt qmi pt q ` Gai pt q, i P vnw, iPvnw a´
i
a´
i
1 . Con-
admits a unique almost periodic solution which is exactly zψ Proof. Rewrite Model (1) in the form
1 2 n żt
sequently zψ P AP pR q.
$
m1i pt q “ ´ai pt qmi pt q ` ai pt q m1i pξ qdξ
’
’
2 P OpR, ν , ν q. In fact, from Fubini’s the-
’
Next, we show that zψ α
’
1 2
’
’ t ´ i pt q
n
’
’
orem, it follows that bi j pt qg j p p j pt ´ τ j pt qqq
’
’ ÿ
’
’ `
żγ ˇ ’
1
’
j“1
’
ˇm pt qˇdν1 pt q
ˇ
ˇ 2 ˇ ’
’
n
ν2 pr´γ , γ sq ´γ φi
’
’ ż `8
Q j pξ qg j p p j pt ´ ξ qqdξ
’ ľ
di j pt q
’
ż γ ˇż t
’
’
’ `
1 0
ˇ ’
´ s ai pξ qd ξ e
şt
ˇdν1 pt q j“1
ˇ ˇ ’
e F s ds
’
“ p q ’
ν2 pr´γ , γ sq ´γ ´8 n
ˇ
i
’
’ ż `8
Q j pξ qg j p p j pt ´ ξ qqdξ ` Ji pt q,
ˇ ˇ & ł
żγ żt ` ei j pt q
1 ´ 0
e´ai pt ´sq ˇFie psqˇdsdν1 pt q j“1
ˇ ˇ ’
ď
’
ν2 pr´γ , γ sq ´γ ´8
’
’ żt
p1i pξ qdξ ` ci pt qmi pt ´ ri pt qq
’
’ p1 pt q “ ´a˜i pt q pi pt q ` a˜i pt q
’
’
ż γ ż `8 i
t ´α˜ i pt q
’
1
’
e´ai ξ ˇFie pt ´ ξ qˇdξ dν1 pt q
´ ˇ ˇ ’
’
“ ’ n ż `8
ν2 pr´γ , γ sq ´γ 0
’
d˜i j pt q Q j pξ qm j pt ´ ξ qdξ
’
’ ľ
’
’ `
ż `8 żγ ’
0
1
ˆ ˙ ’
j“1
e´ai ξ
´
’
ˇF pt ´ ξ qˇdν1 pt q dξ
ˇ e ˇ ’
“
’
n
ν2 pr´γ , γ sq ´γ i
’
’ ż `8
Q j pξ qm j pt ´ ξ qdξ , i P vnw.
’ ł
0 ’
’
’ ` e˜i j pt q
ż `8 ’
% 0
e´ai ξ Tγ pξ qdξ ,
´ j“1
“
0 First of all, it is easy to see that Y is a closed and convex subset of
where the Banach space PAP 1 pR2n , ν1 , ν2 q.
żγ
1 Note that, for all t P R and i P vnw, we have
Tγ pξ q “ ˇF pt ´ ξ qˇdν1 pt q.
ˇ e ˇ
ν2 pr´γ , γ sq ´γ
i
ˇ ˇ t
ˇż
´ s ai pξ qd ξ
şt
ˇ
ˇφ0,i pt qˇ “ ˇ
ˇ ˇ
e J p s q ds
From Lemma 6, we know that OpR, ν1 , ν2 q is invariant bye transla- i
ˇ
ˇ ˇ
´8
tion. This, implies that Fie p. ´ ξ q P OpR, ν1 , ν2 q for every ξ P R. Con- żt
ai pξ qd ξ
şt
sequently ď e´ s |Ji psq|ds
´8
lim Tγ pξ q “ 0, for all ξ P R. Ji`
γ Ñ`8
ď ,
Furthermore, it is easy to see that a´i
9
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
Then n n j
L j }ψ }8 ` Q ` L j }ψ }8 ds
ÿ
di`j Q `
ÿ
` j
e`
ij j
}ψ0 }1 “ max }ψ0 }8 , }ψ01 }8
` ˘
j“1 j“1
˜ ` ˜ `
¸ ¸ żt n
Ji ai
„
´
e´ai pt ´sq
a` αi` `
ÿ
, 1` Ji` b` ` di`j Q ` ` e` Q`
` ˘
ď max ´
ď i ij j ij j
iPvnw ai a´
i
´8 j“1
“ L.
j
L j }ψ }1 ds
Thus, for all ψ P Y,
ςi
“ }ψ }1 , for all i P vnw.
L L a´
}ψ }1 ď }ψ ´ ψ0 }1 ` }ψ0 }1 ď `L“ , i
1´ 1´ Besides,
Next, we define the nonlinear operator as follows; for each ψ P ˇmφi pt q ´ φ01 ,i pt qˇ
ˇ ˇ
ˇ 1 ˇ
PAP 1 pR2n , ν1 , ν2 q,
“ ˇ ´ ai pt qmφi pt q ` Fi pt , φ , ϕ q ` ai pt qφ0,i pt q ´ Ji pt qˇ
ˇ ˇ
pψ qpt q “ zψ pt q.
ď |ai pt q|ˇmφi pt q ´ φ0,i pt qˇ ` ˇFi pt, φ , ϕ q ´ Ji pt qˇ
ˇ ˇ ˇ ˇ
0
ˇ
e´ s ai pξ qdξ Fi ps, φ2 , ϕ2 q ´ Fi ps, φ1 , ϕ1 q dsˇˇ
şt
j“1 ˇ ` ˘ ˇ
“ ˇˇ
n ż `8 j ´8
Q j pξ qg j pϕ j pt ´ ξ qqdξ ˇ ds
ÿ ˇ ˇ
` |ei j psq|ˇ żt
´
e´ai pt ´sq ˇFi ps, φ2 , ϕ2 q ´ Fi ps, φ1 , ϕ1 qˇds
ˇ ˇ
j“1 0 ď
´8
żt „ n
´a´
żt żs
pt ´sq
α ψ ψ }8
ÿ „
ď e i a`
i
`
i
} 1
}8 ` b` L}
ij j ď e´ai
´
pt ´sq
ˇ
ˇai psq
`
φ2,i pξ q ´ φ1,i pξ q 1 dξ ˇ
˘ ˇ
´8 j“1 ´8 s´αi psq
10
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
n n
bi j psq g j pϕ2, j ps ´ τ j psqqq ´ g j pϕ1, j ps ´ τ j psqqq ˇ L }ψ2 ´ ψ1 }8
ˇÿ ÿ
b`
` ˘ˇ
`ˇ ` ij j
j“1 j“1
n ż `8 n
Q j pξ qg j pϕ2, j ps ´ ξ qqdξ L j }ψ2 ´ ψ1 }8
ˇľ ÿ
`ˇ di j psq ` di`j Q `
j
j“1 0 j“1
n ż `8 n
Q j pξ qg j pϕ1, j ps ´ ξ qqdξ ˇ Q ` L j }ψ2 ´ ψ1 }8
ľ ÿ
e`
ˇ
´ di j psq ` ij j
j“1 0 j“1
n ςi
}ψ2 ´ ψ1 }1 ` ςi }ψ2 ´ ψ1 }1
ż `8
ˇł
`ˇ ei j psq Q j pξ qg j pϕ2, j ps ´ ξ qqdξ ď a`
i
0
a´
i
j“1
a` ¯
“ 1 ` i´ ςi }ψ2 ´ ψ1 }1 , for all i P vnw.
´
n ż `8 j
Q j pξ qg j pϕ1, j ps ´ ξ qqdξ ˇ ds
ł ˇ
´ ei j psq ai
j“1 0
żt
Similarly,
„
´
ď e´ai pt ´sq
a`
i
αi` }pψ2 ´ ψ1 q1 }8
´8
n n ˇ pϕ pt q ´ pϕ pt qˇ ď ς˜i }ψ2 ´ ψ1 }1 , for all i P vnw,
ˇ ˇ
L }ψ2 ´ ψ1 }8 ` L j }ψ2 ´ ψ1 }8
ÿ ÿ
` b`
ij j
di`j Q `
j
2,i 1,i ´
a˜i
j“1 j“1
n
and
j
Q ` L j }ψ2 ´ ψ1 }8 ds
ÿ
` e`
ij j
j“1
a˜`
ς˜i }ψ2 ´ ψ1 }1 , for all i P vnw.
´ ¯
i
ˇ 1 1
ˇ
ˇp
n ´ ϕ2,i pt q ´ pϕ1,i pt q ď 1`
żt „´ ˇ
´
αi` ` a˜´
¯
e´ai pt ´sq
ÿ
ď a`
i
b`
ij
` di`j Q `
j
` e` Q`
ij j i
´8 j“1
j Therefore
ˆ L j }ψ2 ´ ψ1 }1 ds
¯
Besides, Theorem 20. Assume that assumptions (C1)–(C6) hold. Then the
unique pν1 , ν2 q-p.a.p. solution of Model (1) is globally exponentially
ˇ 1
ˇm pt q ´ m1 pt qˇ stable.
ˇ
φ2,i φ1,i
ˇ ´ ¯
“ ˇ ´ ai pt q mφ2,i pt q ´ mφ1,i pt q
ˇ
Proof. Rewrite Model (1) in the form
` Fi pt, φ2 , ϕ2 q ´ Fi pt, φ1 , ϕ1 qˇ
ˇ
ˇ
$ żt
1
’mi pt q “ ´ai pt qmi pt q ` ai pt q m1i pξ qdξ
ˇ ˇ ’
ď a` m pt q ´ mφ1,i pt qˇ
’
i ˇ φ2,i
ˇ ˇ ’
’
’ t ´αi pt q
n
’
’
` ˇFi pt, φ2 , ϕ2 q ´ Fi pt, φ1 , ϕ1 qˇ
ˇ ˇ
bi j pt qg j p p j pt ´ τ j pt qqq
ˇ ˇ ’
’ ÿ
’
’
’ `
ςi
’
j“1
’
’
a` }ψ2 ´ ψ1 }1
’
n
’
ď ’ ż `8
Q j pξ qg j p p j pt ´ ξ qqdξ
i
a´
’ ľ
di j pt q
’
i
’
’
’ `
0
’
żt j“1
ˇ ’
’
φ2,i pξ q ´ φ1,i pξ q dξ
˘1 ’
n
ˇ `
` ˇˇai pt q
’
’ ż `8
Q j pξ qg j p p j pt ´ ξ qqds ` Ji pt q,
& ł
t ´αi pt q ` ei j pt q
n
’
’ j“1 0
n ż `8 (15)
Q j pξ qg j pϕ2, j pt ´ ξ qqdξ
ł
` ei j pt q
j“1 0 From Theorem 19, we know that, under (C1)–(C5), Model (1) has
n ż `8 ˇ at least one pν1 , ν2 q-p.a.p. solution pm˚ pt q, p˚ pt qqT P Y, with initial
Q j pξ qg j pϕ1, j pt ´ ξ qqdξ ˇˇ condition ψ ˚ psq. Let pmpt q, ppt qqT be an arbitrary solution of Model
ł ˇ
´ ei j pt q
j“1 0 (1) with initial condition ψ psq. Let
ςi
ď a` }ψ2 ´ ψ1 }1 ` a` αi` }pψ2 ´ ψ1 q1 }8
i
a´ i mi pt q “ mi pt q ´ m˚
i
pt q, pi pt q “ pi pt q ´ p˚
i
pt q, i P vnw, t ą 0.
i
11
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
mi pξ qdξ α˜ i` ´ ci`
1 2
’
ˆ ai psq i p0q “ a˜´ ´ a˜`
’
’
i i
’
s´αi psq
’
’
’
’
’ n n
12
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
1
„ n ´ ż `8 In fact, if (28) is false, then there are four possibles cases:
α˜ i` eλα˜ i ` ci` eλri ` Q j psqeλs ds
` `
a˜` d˜i`j
ÿ
Case 1. There must be some t1 ą 0 and some i1 P vnw, such that
a˜´
i
´λ i
j“1 0
ď Ne´λt }ψ ´ ψ ˚ }0 , @t Ps ´ 8, 0s,
’
’
n
’
’ ż `8 j
Q j pξ qL j | p j ps ´ ξ q|dξ ds
’ ÿ
|ei1 j psq|
’
’
’
’ `
0
“ |ϕi1 pt q ´ pϕi˚ q1 pt q| ď }ψ ´ ψ ˚ }0
’
’ 1
’ j“1
| p pt q|
% i
’
’
ď Ne´λt }ψ ´ ψ ˚ }0 , @t Ps ´ 8, 0s,
’ ż t1
´t a´ ´pt ´sqa´
ď }ψ ´ ψ ˚ }0 e 1 i1 ` e 1 i1
0
which imply that „
›pmpt q, ppt qq ´ pm˚ pt q, p˚ pt qq› ď Ne´λt }ψ ´ ψ ˚ }0 , @t ď 0.
› › ˆ a`i
αi` e´λps´αi1 psqq qN}ψ ´ ψ ˚ }0
1 1
˚
n
We claim that
L qN}ψ ´ ψ ˚ }0 e´λps´τ j psqq
ÿ
` w`
i j j
›pmpt q, ppt qq ´ pm˚ pt q, p˚ pt qq› ď Ne´λt }ψ ´ ψ ˚ }0 , @t ą 0. 1
› ›
˚ (27) j“1
n ż `8
To prove (27), we first show for any q ą 1, the following inequality
di`j L j qN}ψ ´ ψ ˚ }0 Q j pξ qe´λps´ξ q dξ
ÿ
`
hold 1
0
j“1
›pmpt q, ppt qq ´ pm˚ pt q, p˚ pt qq› ď qNe´λt }ψ ´ ψ ˚ }0 , @t ą 0, n
› › ż `8 j
L qN}ψ ´ ψ ˚ }0 Q j pξ qe´λps´ξ q dξ ds
ÿ
˚
` e`
i j j
(28) j“1
1
0
13
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
n
„
´t1 a´
αi` eλαi1
` ż `8
ď }ψ ´ ψ ˚ }0 e i
` qN }ψ ´ ψ ˚ }0 a` Q j pξ qg j p p˚j pt ´ ξ qqdξ
ľ
1
i 1 1 ´ di j pt q
j“1 0
n ´ ¯ ż `8 ¯ j
Q j pξ qeλξ dξ L j
ÿ ´
` w` ` di`j ` e` n ż `8
i j i j
Q j pξ qg j p p j pt ´ ξ qqdξ
ł
j“1
1 1 1
0 ` ei j pt q
j“1 0
ż t1 ´
e´λs ds
´pt1 ´sqai
ˆ e 1 n ż `8
Q j pξ qg j p p˚j pt ´ ξ qqdξ
ł
0 ´ ei j pt q
0
„
´t1 a´ ` λαi1
`
“ }ψ ´ ψ }0 e ` qN }ψ ´ ψ α j“1
˚ i 1
˚
}0 a`
i1 i1
e żt „ żs
ai pξ qd ξ
şt
mi p ξ q d ξ
1
n ´ ¯ ż `8 ¯ j ´ ai pt q e´ s ai psq
s´αi psq
Q j pξ qeλξ dξ L j
´
0
ÿ
` w`
i j
` di`j ` e`
i j
1 1 1
0 n
j“1
bi j psq g j p p j ps ´ τ j psqqq ´ g j p p˚j ps ´ τ j psqqq
ÿ ´ ¯
`
e´λt1
ˆ ˙
pλ´a´ qt1 j“1
ˆ 1´e i 1
a´
i1
´λ n ż `8
Q j pξ qg j p p j ps ´ ξ qqdξ
ľ
` di j psq
1 pλ´a´ 1
"
“ qN }ψ ´ ψ ˚ }0 e´λt1
qt1 0
e i1
` ´ j“1
qN ai ´ λ n ż `8
1
Q j pξ qg j p p˚j ps ´ ξ qqdξ
ľ
ˆ
pλ´a´
˙„ ´ di j psq
` λαi1
`
ˆ 1´e i
qt1
1 a`
i1
α i1
e j“1 0
n ż `8
n
Q j pξ qg j p p j ps ´ ξ qqdξ
¯ ż `8 ¯ j* ł
Q j pξ qeλξ dξ L j ei j psq
´ ´
`
ÿ
` w`
i1 j
` di`j ` e`
i1 j 0
1
0 j“1
j“1
n ż `8 j
1 pλ´a´ 1
"
ă qN }ψ ´ ψ ˚ }0 e´λt1
qt1
Q j pξ qg j p p˚j ps ´ ξ qqdξ ds.
ł
e i1
` ´ ´ ei j psq (34)
N ai ´ λ j“1 0
1
14
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
n ż `8 ¯ j*
|wi2 j pt2 q|L j | p j pt2 ´ τ j pt2 qq| Q j pξ qeλξ dξ L j
ÿ
` ˆ
0
j“1
" a`
pλ´a´
ă qN }ψ ´ ψ ˚ }0 e´λt2
n ż `8 i2 qt2
e i
Q j pξ q| p j pt2 ´ ξ q|dξ
ÿ 2
` |di2 j pt2 q|L j N
j“1 0
a` ˆ
pλ´a´
˙¯„
λαi`
i qt2
αi` e
´
n ż `8 ` 1` 2
1´e i 2 a` 2
Q j pξ q| p j pt2 ´ ξ q|dξ a´ ´λ i
ÿ 2 2
` |ei2 j pt2 q|L j i2
j“1 0
n ´
ÿ ´ ¯
ż t2
´pt2 ´sqa´
„ żs ` w` ` di`j ` e`
|mi2 pξ q|dξ
1 i j
2 i j 2 2
` |ai2 pt2 q| e i 2 |ai2 psq| j“1
0 s´αi2 psq ż `8 ¯ j*
n ˆ Q j pξ qeλξ dξ L j
|wi2 j psq|L j | p j ps ´ τ j psqq|
ÿ
` 0
1 1
"ˆ „
j“1 λα `
“ qN }ψ ´ ψ ˚ }0 e´λt2 ´ ´ a` α ` e i2
n ż `8 N ai ´ λ i2 i2
Q j pξ q| p j ps ´ ξ q|dξ
ÿ
` |di2 j psq|L j 2
0 n ´ ¯ ż `8 ¯ j˙
Q j pξ qeλξ dξ L j
j“1 ÿ ´
` w`
i j
` di`j ` e`
i j
n ż `8 j 2 2 2
0
Q j pξ q| p j ps ´ ξ q|dξ
ÿ j“1
` |ei2 j psq|L j
j“1 0
pλ´a` a` „
λαi`
ˆ a`
qt2 i2
a` αi` e
` ˘
e i 2 ` 1` 2
ď a` }ψ ´ ψ ˚ }0 e
´t2 a´
i 2
i 2
a´
i2
´ λ i 2 2
i 2
n ´ ¯ ż `8
αi` e´λpt2 ´αi2 pt2 qq qN}ψ ´ ψ ˚ }0
¯ j*
` a` Q j pξ qeλξ dξ L j
ÿ ´
i 2 2
` w`
i j
` di`j ` e`
i j
2 2 2
j“1 0
n
´λpt2 ´τ j pt2 qq
ψ ´ψ
ÿ
` w` L qN}
i2 j j
˚
}0 e a` ¯„
λαi`
ă qN }ψ ´ ψ ˚ }0 e´λt2 1 ` αi` e
´
i
j“1 2
a` 2
n ż `8 a´
i2
´λ i 2 2
„ n ´ ă qN }ψ ´ ψ ˚ }0 e´λt1 ,
` λαi2
`
α
´ ¯
ˆ a`
ÿ
i i
e ` w`
i j
` di`j ` e`
i j
2 2 2 2 2 which contradicts the equality (35).
j“1
ż `8 ż t2 Case 4. There must be some t4 ą 0 and some i4 P vnw, such that
¯ j
pt2 ´sqpλ´a´
Q j pξ qeλξ dξ L j a`
q
ˆ i
e i 2 ds
| pi4 pt4 q| “ qNe´λt4 }ψ ´ ψ ˚ }0 ,
0 2
0 1
(37)
" a`
pλ´a´
“ qN }ψ ´ ψ ˚ }0 e´λt2
i2 qt2
e i 2 and
qN
ż t2
pt2 ´sqpλ´a´ ›pmpt q, ppt qq ´ pm˚ pt q, p˚ pt qq›
´ ¯ › ›
q
` 1 ` a`
i
e i 2 ds ˚
2
ă qNe´λt }ψ ´ ψ ˚ }0 ,
0
„ n ´
for t ă t4 . (38)
` λαi2
`
α
ÿ ´ ¯
ˆ a`
i2 i2
e ` w`
i2 j
` di`j ` e`
i2 j By a similar calculus like case 2., we obtain
2
j“1
1
| pi4 pt4 q| ď qN }ψ ´ ψ ˚ }0 e´λt4 ` λq
1
pa˜´ ` a˜`
a˜´
i
´ λ 4
i4 i4
15
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
Consider the following FGRNs with leakage and mixed time vary-
ing delays:
2
$
’m1i pt q “ ´ai pt qmi pt ´ αi pt qq ` bi j pt qg j p p j pt ´ τ j pt qqq
’ ÿ
’
’
’
j“1
’
’
’
2
’
’ żt
Q j pt ´ ξ qg j p p j pξ qqdξ
’
’ ľ
` di j pt q
’
’
’
’
’
’
’ j“1 ´8
’
’
’ 2 żt
Q j pt ´ ξ qg j p p j pξ qqdξ
’ ł
ei j pt q
’
’
’
& `
´8
j“1 (39)
’
’ `Ji pt q,
Fig. 2. Numerical solution of Model 39 with initial values p´0.2, 0.1, ´0.1, 0.2q .
T
´a˜i pt q pi pt ´ α
’
p1i pt q “
’
˜ i pt qq ` ci pt qmi pt ´ ri pt qq
’
’
’
’
2 żt
’
’
d˜i j pt q Q j pt ´ ξ qm j pξ qdξ
’
’ ľ
’
’
’ `
´8
’
j“1
’
’
’
2
’
’ żt
Q j pt ´ ξ qm j pξ qdξ ,
’
i P v2 w.
’ ł
` e˜i j pt q
’
’
’
%
j“1 ´8
x2
The regulatory function is taken as g j pxq “ 1`x2
for j P v2w, that
is, the Hill coefficient is 2. Then the upper bound of g1j p.q can be
calculated as L j “ 0.65.
The delay kernels Q j ptq “ e´t for j P v2w, which implies that
`
Q j “ 1.
Fig. 3. Phase-plane behavior of the state variables mpt q and ppt q.
The remaining parameters are defined as follows:
j « ? ff
a1 pt q 0.1p3 ` sinp3t q ` cosp ?3t qq
„
“ ,
„ n ´ a˜1 pt q 0.1p4 ` cosp7t q ` cosp 7t qq
α˜ i` eλα˜ i4 ` ci` eλri4 `
` ` ¯
d˜i`j ` e˜`
ÿ
ˆ a˜`
i i j j « ? ff
4 4 4 4 4
a2 pt q 0.1p3 ` cosp3t q ` sinp? 3t qq
„
j“1
“ ,
ż `8 j a˜2 pt q 0.1p4 ` sinp7t q ` sinp 7t qq
ˆ Lj Q j psqeλs ds
0 ?
0.1pcosp π t q ` e´t q
« 2
ff
J1 pt q
„ j
ă qN }ψ ´ ψ ˚ }0 e´λt4 , “ ? ,
0.2pcosp π t q ` e´t q
2
J2 pt q
?
α1 pt q 0.001p2 ` cosp3t q ` sinp 3t q ` 1
„ j « ff
which contradicts the equality (37). 1`t 2
q
“ ? ,
Consequently, (28) hold. Letting q Ñ 1, then (27) hold. Hence, α˜ 1 pt q 0.001p2 ` cosp2t q ` sinp 2t q ` 1
1`t 2
q
the pν1 , ν2 q-p.a.p. solution of the Model (1) is globally exponen-
α2 pt q 0.001p1 ` cosp2t q ` 1`1t 2 q
„ j « ff
tially stable. ˝
“
α˜ 2 pt q 0.001p1 ` sinp2t q ` 1`1t 2 q
τ1 pt q
« ff
1
r pt q | cosp5t q|
„ j „ j
“ 1 “ 3 ,
τ2 pt q r2 pt q 1
3
| sinp5t q|
6. Example
« 2
ff
e11 pt q e˜11 pt q 0.001e´t
„ j „ j
In this section, we give an example to illustrate the feasibility “ “ 2
e21 pt q e˜21 pt q 0.002e´t
and effectiveness of our main results.
Let the measures ν1 “ ν2 “ ν where the Radon-Nikodym
« 2
ff
e12 pt q e˜12 pt q 0.002psinp0.1t q ` e´t q
„ j „ j
derivative of ν is given by “ “ 2 ,
e22 pt q e˜22 pt q 0.001pcosp0.5t q ` e´t q
16
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
` 0.001 ` 0.004q0.65
CRediT authorship contribution statement
“ 0.0278
2 ´ Moez Ayachi: Conceptualization, Writing – original draft.
ς2 “ a` α` `
ÿ ¯
2 2
b`
2j
` d2`j Q `
j
` e` Q` Lj
2j j
j“1 References
“ 0.5 ˆ 0.003 ` p0.009 ` 0.015 ` 0.006 ` 0.008 [1] Abbas S, Dhama S, Pinto M, Sepúlveda D. Pseudo compact almost automorphic
` 0.002 ` 0.002q0.65 solutions for a family of delayed population model of nicholson type. J Math
Anal Appl 2021;495(1):124722.
“ 0.0288. [2] Amdouni M, Chérif F, Tunç C. On the weighted piecewise pseudo almost
automorphic solutions mackey–glass model with mixed delays and harvest-
2 ´
ing term. Iranian Journal of Science and Technology, Transactions A: Science
ς˜1 “ a˜` α˜ ` ` c1` `
¯
d˜1`j ` e˜`
ÿ
1 1 1j
Q`
j
Lj 2021;45(2):619–34.
j“1 [3] Aouiti C, Dridi F. Delayed fuzzy genetic regulatory networks: novel results. In-
ternational Journal of Biomathematics 2020.
“ 0.6 ˆ 0.005 ` 0.002 ` p0.003 ` 0.004 ` 0.001 [4] Aouiti C, Dridi F. (μ, ν )-Pseudo-almost automorphic solutions for high-order
` 0.004q0.65 hopfield bidirectional associative memory neural networks. Neural Computing
and Applications 2020;32(5):1435–56.
“ 0.0148. [5] Aouiti C, Gharbia IB, Cao J, M’hamdi MS, Alsaedi A. Existence and global ex-
ponential stability of pseudo almost periodic solution for neutral delay BAM
2 ´ neural networks with time-varying delay in leakage terms. Chaos, Solitons &
ς˜2 “ a˜` α˜ ` ` c2` `
¯
d˜2`j ` e˜`
ÿ
2 2 2j
Q`
j
Lj Fractals 2018;107:111–27.
j“1 [6] Arbi A. Dynamics of BAM neural networks with mixed delays and leakage
time-varying delays in the weighted pseudo–almost periodic on time-space
“ 0.6 ˆ 0.003 ` 0.004 ` p0.006 ` 0.008 ` 0.002 scales. Math Methods Appl Sci 2018;41(3):1230–55.
[7] Ayachi M. Existence and exponential stability of weighted pseudo-almost pe-
` 0.002q0.65 riodic solutions for genetic regulatory networks with time-varying delays. In-
“ 0.0175. ternational Journal of Biomathematics 2021;14(2):2150 0 06.
[8] Blot J, Cieutat P, Ezzinbi K. New approach for weighted pseudo-almost periodic
´ J` ´ a` ¯ ¯ functions under the light of measure theory, basic results and applications.
i i
L “ max , 1` ´ i
J` Appl Anal 2013;92(3):493–526.
iPv2w a´i
ai [9] Bohr B. Almost periodic functions. Courier Dover Publications; 2018.
[10] Chen L, Aihara K. Stability of genetic regulatory networks with time delay. IEEE
“ max p2, 4, 1.2, 2.4q “ 4. Transactions on circuits and systems I: Fundamental Theory and Applications
2002;49(5):602–8.
Thus [11] Chen T, He HL, Church GM. Modeling gene expression with differential equa-
ςi a` ς˜i a˜` tions. Biocomputing’99 1999;4:29–40.
" *
“ max ςi , ς˜i
´ ¯ ´ ¯
i i
´
, 1` , 1` [12] Chen X, Lin F. Almost periodic solutions of neutral functional differential equa-
iPv2w ai a´
i
´
a˜i a˜´
i
tions. Nonlinear Anal Real World Appl 2010;11(2):1182–9.
[13] Chen Z, Jiang H. Stability analysis of genetic regulatory networks with mixed
“ maxt0.278, 0.288, 0.167, 0.173, 0.074, 0.087, time-delays. International Symposium on Neural Networks 2011:280–9.
[14] Chérif F, Miraoui M. New results for a lasota–wazewska model. International
0.06, 0.07u Journal of Biomathematics 2019;12(02):1950019.
“ 0.173 ă 1, [15] Diagana T. Weighted pseudo almost periodic functions and applications. CR
Math 2006;343(10):643–6.
which implies that pC5q holds. [16] Diagana T. Weighted pseudo-almost periodic solutions to some differ-
ential equations. Nonlinear Analysis: Theory, Methods & Applications
We can see that, (C1)-(C6) are satisfied, then Model (39) has a 2008;68(8):2250–60.
unique doubly-measure pseudo-almost periodic solution, which is [17] Diagana T, Ezzinbi K, Miraoui M. Pseudo-almost periodic and pseudo-almost
globally exponentially stable, in the region automorphic solutions to some evolution equations involving theoretical mea-
sure theory. Cubo (Temuco) 2014;16(2):01–32.
Y “ ψ P PAP 1 pR2n , ν1 , ν2 q : }ψ ´ ψ0 }1 ď 0.836 .
! )
[18] Duan L, Di F, Wang Z. Existence and global exponential stability of almost peri-
odic solutions of genetic regulatory networks with time-varying delays. Journal
of Experimental & Theoretical Artificial Intelligence 2020;32(3):453–63.
The results are verified by the numerical simulations in Figs. 2 and [19] Duan L, Fang X, Huang C. Global exponential convergence in a delayed almost
3 periodic nicholson’s blowflies model with discontinuous harvesting. Math
Methods Appl Sci 2018;41(5):1954–65.
[20] Duan L, Shi M, Huang L. New results on finite-/fixed-time synchronization of
7. Conclusion delayed diffusive fuzzy HNNs with discontinuous activations. Fuzzy Sets Syst
2021;416:141–51.
[21] Dudley RM. Real analysis and probability. Cambridge University Press; 2002.
It is well known that many mathematicians proposed differ-
[22] Emmert-Streib F, de Matos Simoes R, Mullan P, Haibe-Kains B, Dehmer M. The
ent classes of functions that are more appropriate than periodic gene regulatory network for breast cancer: integrated regulatory landscape of
to explain complex processes or phenomena that show behav- cancer hallmarks. Front Genet 2014;5:15.
[23] Fink AM. Almost periodic differential equations. Springer; 2006.
ior that is not regular such as almost periodic functions, almost
[24] Friedman N, Linial M, Nachman I. Using bayesian networks to analyze expres-
automorphic functions, and their different extensions. In this pa- sion data. J Comput Biol 20 0 0;7(3–4):601–20.
per, doubly-measure pseudo-almost periodic functions are used [25] Goldbeter A. A model for circadian oscillations in the drosophila period protein
to model a class of FGRNs with leakage delays, time-varying de- (PER). Proceedings of the Royal Society of London Series B: Biological Sciences
1995;261(1362):319–24.
lays, and distributed delays. We initiate some sufficient condi- [26] Goodwin BC. Oscillatory behavior in enzymatic control processes. Adv Enzyme
tions that contribute to the existence, uniqueness, and stability of Regul 1965;3:425–37.
17
M. Ayachi Chaos, Solitons and Fractals 154 (2022) 111659
[27] Gopalsamy K. Leakage delays in BAM. J Math Anal Appl 2007;325(2):1117–32. [45] Miraoui M, Repovš DD. Dynamics and oscillations of models for differential
[28] Graudenzi A, Serra R, Villani M, Damiani C, Colacci A, Kauffman SA. Dynami- equations with delays. Boundary Value Problems 2020;2020(1):1–17.
cal properties of a boolean model of gene regulatory network with memory. J [46] Monk NAM. Oscillatory expression of hes1, p53, and NF-κ b driven by tran-
Comput Biol 2011;18(10):1291–303. scriptional time delays. Current Biology 2003;13(16):1409–13.
[29] He W, Cao J. Robust stability of genetic regulatory networks with distributed [47] Rakkiyappan R, Balasubramaniam P, Balachandran K. Delay-dependent global
delay. Cogn Neurodyn 2008;2(4):355. asymptotic stability criteria for genetic regulatory networks with time delays
[30] Hirata H, Yoshiura S, Ohtsuka T, Bessho Y, Harada T, Yoshikawa K, Kageyama R. in the leakage term. Phys Scr 2011;84(5):055007.
Oscillatory expression of the bHLH factor hes1 regulated by a negative feed- [48] Ratnavelu K, Kalpana M, Balasubramaniam P. Stability analysis of fuzzy genetic
back loop. Science 2002;298(5594):840–3. regulatory networks with various time delays. Bulletin of the Malaysian Math-
[31] Hirata H, Yoshiura S, Ohtsuka T, Bessho Y, Harada T, Yoshikawa K, Kageyama R. ematical Sciences Society 2018;41(1):491–505.
Oscillatory expression of the bHLH factor hes1 regulated by a negative feed- [49] Ren F, Cao J. Asymptotic and robust stability of genetic regulatory networks
back loop. Science 2002;298(5594):840–3. with time-varying delays. Neurocomputing 2008;71(4–6):834–42.
[32] Hu J, Liang J, Cao J. Stabilization of genetic regulatory networks with [50] Smolen P, Baxter DA, Byrne JH. Mathematical modeling of gene networks. Neu-
mixed time-delays: an adaptive control approach. IMA J Math Control Inf ron 20 0 0;26(3):567–80.
2015;32(2):343–58. [51] Torrejón R. Positive almost periodic solutions of a nonlinear integral equation
[33] Johnson CH. Circadian clocks and cell division: what’s the pacemaker? Cell Cy- from the theory of epidemics. J Math Anal Appl 1991;156(2):510–34.
cle 2010;9(19):3864–73. [52] Wang Y, Ma Z, Shen J, Liu Z, Chen L. Periodic oscillation in delayed gene
[34] Kumar V. Biological rhythms. Springer Science & Business Media; 2002. networks with SUM regulatory logic and small perturbations. Math Biosci
[35] Lee TH, Lakshmanan S, Park JH, Balasubramaniam P. State estimation for ge- 2009;220(1):34–44.
netic regulatory networks with mode-dependent leakage delays, time-vary- [53] Yaghoobi H, Haghipour S, Hamzeiy H, Asadi-Khiavi M. A review of mod-
ing delays, and markovian jumping parameters. IEEE Trans Nanobioscience eling techniques for genetic regulatory networks. J Med Signals Sens
2013;12(4):363–75. 2012;2(1):61–70.
[36] Leloup JC, Goldbeter A. Modeling the molecular regulatory mechanism of cir- [54] Yang T, Yang LB. The global stability of fuzzy cellular neural network. IEEE
cadian rhythms in drosophila. Bioessays 20 0 0;22(1):84–93. transactions on circuits and systems I: fundamental theory and applications
[37] Lewis J. Autoinhibition with transcriptional delay: a simple mechanism for the 1996;43(10):880–3.
zebrafish somitogenesis oscillator. Current Biology 2003;13(16):1398–408. [55] Zang H, Zhang T, Zhang Y. Bifurcation analysis of a mathematical model
[38] Li Y, Huo N. (μ, ν )-Pseudo almost periodic solutions of clifford-valued high- for genetic regulatory network with time delays. Appl Math Comput
-order HNNs with multiple discrete delays. Neurocomputing 2020;414:1–9. 2015;260:204–26.
[39] Li Z, Chen D, Liu Y, Zhao Y. New delay-dependent stability criteria of ge- [56] Zhang C. Pseudo almost periodic solutions of some differential equations. J
netic regulatory networks subject to time-varying delays. Neurocomputing Math Anal Appl 1994;181(1):62–76.
2016;207:763–71. [57] Zhang C. Pseudo almost periodic solutions of some differential equations, ||. J
[40] Liang J, Han J. Stochastic boolean networks: an efficient approach to modeling Math Anal Appl 1995;192(2):543–61.
gene regulatory networks. BMC Syst Biol 2012;6(1):1–21. [58] Zhang X, Wang Y, Wu L. Analysis and design of delayed genetic regulatory
[41] Liu C, Wang X, Xue Y. Global exponential stability analysis of discrete-time networks. Springer; 2019.
genetic regulatory networks with time-varying discrete delays and unbounded [59] Zhang X, Wu L, Zou J. Globally asymptotic stability analysis for genetic reg-
distributed delays. Neurocomputing 2020;372:100–8. ulatory networks with mixed delays: an m-matrix-based approach. IEEE/ACM
[42] Madhamshettiwar PB, Maetschke SR, Davis MJ, Reverter A, Ragan MA. Gene Trans Comput Biol Bioinf 2015;13(1):135–47.
regulatory network inference: evaluation and application to ovarian cancer al- [60] Zhou H, Yang L, Jiang W. Weighted pseudo almost periodic solutions for
lows the prioritization of drug targets. Genome Med 2012;4(5):1–16. a class of hematopoiesis model with time-varying delay. Anal Theory Appl
[43] Manivannan R, Cao J, Chong KT. Generalized dissipativity state estimation for 2017;3:197–205.
genetic regulatory networks with interval time-delay signals and leakage de- [61] Ayachi M. Measure-pseudo almost periodic dynamical behaviors for BAM neu-
lays. Commun Nonlinear Sci Numer Simul 2020;89:105326. ral networks with D operator and hybrid time-varying delays. Neurocomputing
[44] M’hamdi MS, Aouiti C, Touati A, Alimi AM, Snasel V. Weighted pseudo al- 2021 In press. doi:10.1016/j.neucom.2021.11.020.
most-periodic solutions of shunting inhibitory cellular neural networks with
mixed delays. Acta Mathematica Scientia 2016;36(6):1662–82.
18