Performance | Ψ4 | Ψ1 | Ψ2 | Ψ3 |
H∞ | 0 | -I | 0 | γ2I |
L2−L∞ | I | 0 | 0 | γ2I |
This paper explored the topic of extended dissipativity analysis for Markovian jump neural networks (MJNNs) that were influenced by time-varying delays. A distinctive Lyapunov functional, distinguished by a non-zero delay-product types, was presented. This was achieved by combining a Wirtinger-based double integral inequality with a flexible matrix set. This novel methodology addressed the limitations of the slack matrices found in earlier research. As a result, a fresh condition for extended dissipativity in MJNNs was formulated, utilizing an exponential type reciprocally convex inequality in conjunction with the newly introduced nonzero delay-product types. A numerical example was included to demonstrate the effectiveness of the proposed methodology.
Citation: Wenlong Xue, Yufeng Tian, Zhenghong Jin. A novel nonzero functional method to extended dissipativity analysis for neural networks with Markovian jumps[J]. AIMS Mathematics, 2024, 9(7): 19049-19067. doi: 10.3934/math.2024927
[1] | Zirui Zhao, Wenjuan Lin . Extended dissipative analysis for memristive neural networks with two-delay components via a generalized delay-product-type Lyapunov-Krasovskii functional. AIMS Mathematics, 2023, 8(12): 30777-30789. doi: 10.3934/math.20231573 |
[2] | Chengqiang Wang, Xiangqing Zhao, Yang Wang . Finite-time stochastic synchronization of fuzzy bi-directional associative memory neural networks with Markovian switching and mixed time delays via intermittent quantized control. AIMS Mathematics, 2023, 8(2): 4098-4125. doi: 10.3934/math.2023204 |
[3] | Xipan Zhang, Changchun Shen, Dingju Xu . Reachable set estimation for neutral semi-Markovian jump systems with time-varying delay. AIMS Mathematics, 2024, 9(4): 8043-8062. doi: 10.3934/math.2024391 |
[4] | Miao Zhang, Bole Li, Weiqiang Gong, Shuo Ma, Qiang Li . Matrix measure-based exponential stability and synchronization of Markovian jumping QVNNs with time-varying delays and delayed impulses. AIMS Mathematics, 2024, 9(12): 33930-33955. doi: 10.3934/math.20241618 |
[5] | Saravanan Shanmugam, R. Vadivel, Mohamed Rhaima, Hamza Ghoudi . Improved results on an extended dissipative analysis of neural networks with additive time-varying delays using auxiliary function-based integral inequalities. AIMS Mathematics, 2023, 8(9): 21221-21245. doi: 10.3934/math.20231082 |
[6] | Qiang Li, Weiqiang Gong, Linzhong Zhang, Kai Wang . Robust dissipativity and passivity of stochastic Markovian switching CVNNs with partly unknown transition rates and probabilistic time-varying delay. AIMS Mathematics, 2022, 7(10): 19458-19480. doi: 10.3934/math.20221068 |
[7] | Arthit Hongsri, Wajaree Weera, Thongchai Botmart, Prem Junsawang . Novel non-fragile extended dissipative synchronization of T-S fuzzy complex dynamical networks with interval hybrid coupling delays. AIMS Mathematics, 2023, 8(12): 28601-28627. doi: 10.3934/math.20231464 |
[8] | Mei Xu, Bo Du . Dynamic behaviors for reaction-diffusion neural networks with mixed delays. AIMS Mathematics, 2020, 5(6): 6841-6855. doi: 10.3934/math.2020439 |
[9] | Yakufu Kasimu, Gulijiamali Maimaitiaili . Non-fragile H∞ filter design for uncertain neutral Markovian jump systems with time-varying delays. AIMS Mathematics, 2024, 9(6): 15559-15583. doi: 10.3934/math.2024752 |
[10] | Yuxin Lou, Mengzhuo Luo, Jun Cheng, Xin Wang, Kaibo Shi . Double-quantized-based H∞ tracking control of T-S fuzzy semi-Markovian jump systems with adaptive event-triggered. AIMS Mathematics, 2023, 8(3): 6942-6969. doi: 10.3934/math.2023351 |
This paper explored the topic of extended dissipativity analysis for Markovian jump neural networks (MJNNs) that were influenced by time-varying delays. A distinctive Lyapunov functional, distinguished by a non-zero delay-product types, was presented. This was achieved by combining a Wirtinger-based double integral inequality with a flexible matrix set. This novel methodology addressed the limitations of the slack matrices found in earlier research. As a result, a fresh condition for extended dissipativity in MJNNs was formulated, utilizing an exponential type reciprocally convex inequality in conjunction with the newly introduced nonzero delay-product types. A numerical example was included to demonstrate the effectiveness of the proposed methodology.
Neural networks (NNs) are rooted in the emulation of the human brain's intricate architecture. These computational models have been effectively employed across various domains, including medical diagnostics, the financial industry, and signal processing [1,2], in which the versatility and efficacy underscore the transformative potential of NNs in advancing these fields. Stability holds significant importance as a prerequisite in various applications [3]. Nevertheless, the implementation of (large-scale) neural networks unavoidably introduces time-varying delays (TVDs) arising from finite amplifier switching speeds and inter-neuron communication rates. These delays can potentially lead to instability and oscillations within the NN [4,5]. Therefore, analyzing the dynamic performance of NNs with TVDs is a meaningful topic [6,7,8].
Additionally, abrupt parameter changes in the structures of many NNs may occur due to sudden environmental changes, components or interconnection failures. These NNs can exhibit distinct modes that transition or switch unpredictably between each other, and such switching can be determined by a Markov chain, leading to MJNNs [9,10]. It is noted that, dissipativity, recognized as a critical dynamic performance metric, has been proven to be a potent tool for the analysis of nonlinear systems, particularly when leveraging energy-related input-output characteristics, as highlighted in [11]. While the traditional notion of dissipativity encompasses both passivity and H∞ performances, it falls short in addressing the l2−l∞ performance. Addressing this lacuna, the seminal work in [12] marked the first stride toward reconciling this gap. Consequently, the concept of "extended dissipativity" emerged, which has been widely applied to practical systems, presenting a more comprehensive performance index [13,14]. This innovative index effectively amalgamates the features of passivity, traditional dissipativity, H∞ performance, and the l2−l∞ performance. To further harness the potential and expand the applicability of MJNNs, there's been an increasing scholarly focus on this domain. For example, the delay-product-type (DPT) functional approach is presented to analyze MJNNs [15,16,17]. Nevertheless, the matrices that were introduced had a constraint imposed upon them, requiring them to be positive definite. This constraint was imposed to guarantee the positive definiteness of the Lyapunov function, which in turn led to certain limitations in the conditions established. In response to this concern, an attempt was made in [18] to mitigate the situation. This involved creating two innovative DPT functionals, aiming to support the extended dissipativity analysis of MJNNs. Nevertheless, it's worth noting that only the DPT Lyapunov functional based on a single integral was developed in [18]. This, in turn, prompted the proposal of an enhanced approach in [19]. In this later work, a double integral-based DPT (DIDPT) functional was introduced. This new functional featured augmented forms of double integrals to provide a more comprehensive solution. The TVDs-dependent components in the DIDPT are efficient for capturing TVDs information but suffer from incomplete zero components, limiting their utilization in system information. This insufficiency motivates the current study.
Moreover, the traditional method for estimating the integral term −∫tt−ς˙§T(s)R˙§(s)ds involves handling the TVDs in the denominators using the traditional reciprocally convex inequality (TRCI) [20]. To improve the conservatism of the TRCI, an improved version known as the improved reciprocally convex inequality (IRCI) was proposed in [21,22]. Furthermore, to achieve even less conservative results without introducing additional decision variables, the exponential type reciprocally convex inequality (ETRCI) is first introduced in the exponential-type parameters introduced in [23], which quickly approaches the left-hand side of the inequality and also covers [21,22] as special cases. In the existing works, there have been no related studies exploring the use of ETRCI for reducing conservatism in the analysis of MJNNs, warranting further investigation in this direction.
Building upon the preceding analysis, this paper directs its attention toward the extended dissipativity analysis concerning MJNNs that exhibit TVDs. Through the application of the S-procedure lemma, coupled with the utilization of adaptable matrices, a fresh approach is formulated. Drawing upon the ETRCI and the innovative DIDPT functional, a novel criterion is introduced. This criterion serves to ensure both the stochastic stability and extended dissipativity of MJNNs. The main contributions are as follows:
● By employing the S-procedure lemma and integrating the use of adaptable matrices, we propose the DIDPT functional, which overcomes the zero components in [19]. As a result, the newly formulated DIDPT functional optimally captures the intricate interplay of coupling dynamics among states and TVD, offering a robust solution in the realm of state dynamics.
● Drawing upon the ETRCI and the innovative DIDPT functional, a novel criterion is introduced. This criterion serves to ensure both the stochastic stability and extended dissipativity of MJNNs.
Notation. In this paper, Rl means the l-dimensional Euclidean space; RT denotes the transpose of R; '*' represents the symmetric term in linear matrix inequalities (LMIs); He[R] represents R+RT; diag{...} and col[...] denote a block-diagonal matrix and a block-column vector, respectively; Co{...} represents a convex set; E[⋅] represents the expectation operator; λmax(X) and λmin(X) represent the maximum and minimum eigenvalue of X, respectively.
Consider the ensuing MJNNs exhibiting TVDs:
{˙§(t)=−A(rt)§(t)+W0(rt)π(W2ρ(t))+W1(rt)π(W2§(t−σ(t)))+B1ρ(t),†(t)=C§(t)+D1§(t−σ(t))+D2π(W2§(t))+B2ρ(t), | (2.1) |
where the state vector §(t)∈Rn, the disturbance input w(t)∈Rw, and the output †(t)∈Ry are defined for the system. The disturbance input ρ(t) belongs to L2[0,+∞). Additionally, the system parameters include the known interconnection weight matrices A(rt), W0(rt), W1(rt), B1, B2, C, D1, D2, and W2=col[W21,W22,⋯,W2n]. The Markov chain rt(t≥0), which operates for t greater than or equal to zero, exhibits right continuity and takes on values within a finite set S={1,2,...,m}. The transition rate matrix (TRM), denoted as Π and defined as [τij], characterizes this chain for δ>0, limδ→0o(δ)/δ=0, τij≥0, τii=−m∑j=1,j≠iτij(j≠i), and
Pr{rt+δ=j|rt=i}={τijδ+o(δ),j≠i,1+τiiδ+o(δ),j=i. | (2.2) |
The TVD σ(t) is subject to the following constraints:
0≤σ(t)≤ς,−ν≤˙σ(t)≤ν, | (2.3) |
where ς and ν are constants. For constants k−l and k+l(l=1,2,⋯,n), the neuron activation function π(W2(§(t))) satisfies the following conditions:
k−l≤πl(a1)−πl(a2)a1−a2≤k+l,a1≠a2, | (2.4) |
k−l≤πl(a)a≤k+l,a≠0. | (2.5) |
In this paper, a less conservative sufficient condition will be proposed to ensure the extended dissipativity of the MJNNs (2.1). To accomplish this, the subsequent preconditions are established.
From the standpoint of practical application, the extended disspativity can embody more performance indices for physical systems such as H∞, passivity, dissipativity, and l2−l∞ performance [13,14]. The definition is given as follows:
Definition 1. [12] MJNNs (2.1) are considered extended dissipative if there exist prescribed matrices Ψ1, Ψ2, Ψ3, and Ψ4, satisfying
Ψ1=ΨT1≤0,Ψ3=ΨT3,Ψ4=ΨT4, | (2.6) |
BT2Ψ1B2+He[BT2Ψ2]+Ψ3>0, | (2.7) |
(||Ψ1||+||Ψ2||)⋅||Ψ4||=0, | (2.8) |
and a scalar ϱ can be found such that the following condition holds:
∫T0(†T(t)Ψ1†(t)+2†T(t)Ψ2ρ(t)+ρT(t)Ψ3ρ(t))dt≥ρT(t)Ψ4†(t)+ϱ. | (2.9) |
Lemma 1. [24] Let us denote the set G={g}, and introduce the notations: F(g), X1(g), X2(g), ..., Xk(g) are defined as functionals or functions. The domain H is specified as
H={g∈Z:X1(g)≥0,X2(g)≥0,...,Xk(g)≥0}. |
We proceed to examine two specific conditions:
(I) The initial condition states that F(g)≥0 holds for every g∈H.
(II) Additionally, we consider the existence of nonnegative values λ1≥0,λ2≥0,...,λk≥0 such that the inequality holds:
It can be established that condition (II) implies condition (I).
Lemma 2. [23] For real constants β∈(0,1), κ1≥0, κ2≥0, and real symmetric positive definite matrices R1, R2∈Rm×m, the matrix inequality below is satisfied in the presence of a real matrix S∈Rm×m:
[1βR10011−βR2]≥[R1+(1−β)eκ1T1S∗R2+βeκ2T2], |
where T1=R1−SR−12ST,T2=R2−STR−11S.
Remark 1. Different from the unknown slack matrix S in [17], parameters κ1 and κ2 are prespecified, which can be freely chosen. Thus, the flexible tuning can lead to less conservatism.
In this section, our emphasis lies in introducing an adequate criterion to ensure both the stochastic stability and extended dissipativity of the system (2.1).
Let's begin by revisiting the DIDPT functional denoted as V∗(§t), which was introduced in [19]. The DIDPT functional is defined as follows:
V∗(§t)=VU(§t)+VT(§t), | (2.10) |
where
VU(§t)=σ(t)2∫0−σ(t)∫tt+θ˙§T(s)U1˙§(s)dsdθ+ςσ(t)2∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)U2˙§(s)dsdθ,VT(§t)=−σ(t)ς2χT4(t)T1χ4(t)−ςσ(t)ς2χT5(t)T2χ5(t),Tk=[32Uk0−3Uk∗3Uk−6Uk∗∗18Uk],k=1,2, |
where the positive definite matrices U1 and U2 are denoted with appropriate dimensions and
ςσ(t)=ς−σ(t),x1(t)=1σ(t)∫tt−σ(t)§(s)ds,x2(t)=1ςσ(t)∫t−σ(t)t−ς§(s)ds,x3(t)=1σ2(t)∫0−σ(t)∫tt+θ§(s)dsdθ,x4(t)=1ς2σ(t)∫−σ(t)−ς∫t−σ(t)t+θ§(s)dsdθ,χ1(t)=col[§(t),π(W2§(t)),˙§(s)],χ2(t)=col[§(t),§(t−σ(t)),x1(t)],χ3(t)=col[§(t−σ(t)),§(t−ς),x2(t)],χ4(t)=col[§(t),x1(t),x3(t)],χ5(t)=col[x(t−σ(t)),x2(t),x4(t)],ξ1(t)=[§(t),§(t−σ(t)),§(t−ς)],ξ2(t)=[π(W2§(t)),π(W2x(t−σ(t))),π(W2§(t−ς))],ξ3(t)=[x1(t),x2(t),x3(t),x4(t)],ξ4(t)=[˙§(t),˙§(t−σ(t)),˙§(t−ς)],η(t)=col[ξ1(t),ξ2(t),ξ3(t),ξ4(t)]. |
It should be pointed out that the DIDPT functional constructed in [19] increases the delay information. However, it lacks consideration of the coupling information. This is due to the existence of zero components in Tk, which results in insufficient cross terms among §(t), §(t−σ(t)), x1(t), and x2(t). To address this, we use the S-procedure lemma and introduce semi-positive definite matrices §k=[0XkXk0]. By deducing Gk, if there exist scalars μk≥0, we can express it as follows:
Gk=Tk−μk[§k000]=[32Uk−μkXk−3Uk∗3Uk−6Uk∗∗18Uk]≥0. | (2.11) |
Then, we have
σ(t)2∫0−σ(t)∫tt+θ˙§T(s)U1˙§(s)dsdθ≥1σ(t)χT4(t)G1χ4(t)≥σ(t)ς2χT4(t)G1χ4(t),ςσ(t)2∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)U2˙§(s)dsdθ≥1ςσ(t)χT5(t)G2χ5(t)≥ςσ(t)ς2χT5(t)G1χ5(t). |
Further, we can construct the following nonzeoro DPT (NDPT) functional.
Proposition 1. The provided functional V0(§t) is positive definite for the MJNNs (2.1) incorporating (2.3), under the conditions of matrices U1>0 and U2>0, along with matrices X1 and X2 that fulfill (2.11).
V0(§t)=VU(§t)+VG(§t), | (2.12) |
where
VU(§t)=σ(t)2∫0−σ(t)∫tt+θ˙§T(s)U1˙§(s)dsdθ+ςσ(t)2∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)U2˙§(s)dsdθ,VG(§t)=−σ(t)ς2χT4(t)G1χ4(t)−ςσ(t)ς2χT5(t)G2χ5(t). |
Remark 2. It is shown from [25] that the conservatism of stability criterion can be reduced by increasing the ply of integral terms in the LKF. Thus, a novel NDPT Lyapunov functional with double integrals is introduced in Proposition 1 with the subsequent benefits:
(i) The novel NDPT functional (2.12) strengthens the connections among the variables ˙σ(t), x3(t), and x4(t), as compared to the existing Lyapunov functionals in [9,15,18].
(ii) In addressing the challenge posed by zero components, this paper leverages the double integral condition presented in [5] combined with the S-procedure lemma as outlined in Lemma 1. By establishing a direct relationship between the functional VG(§t) and VU(§t), the introduction of supplementary variables becomes superfluous. Moreover, by incorporating nonintegral components possessing negative attributes, there is a reduction in the prerequisite for matrices to maintain strict positive definiteness. This methodological relaxation not only curtails conservatism but also lessens the computational complexity inherent in the analysis.
(iii) DIDPT functional presented in [19], the innovative NDPT formulation, as defined in Eq (2.12) circumvents the inclusion of zero components. This design choice facilitates the comprehensive integration of additional information among the variables §(t), §(t−σ(t)), x1(t), and x2(t). For example, the term −μk§T(t)X1x1(t) in (2.12) instead of 0 in [19], thereby preserving the connectivity information between §T(t) and x1(t). Consequently, the novel NDPT (2.12) yields less conservative results.
(iv) It is essential to highlight that the number of decision variables (NDVs) plays a pivotal role in the stability analysis, directly influencing the computational complexity. Within the NDPT framework, a delicate balance between conservatism and complexity can be achieved by judiciously selecting different values of μk. Opting for nonzero μk≠0 values aligns with a preference for conservatism, enhancing the robustness of the analysis. Conversely, setting μk=0 to zero caters to computational efficiency, streamlining the process. This flexibility allows for tailored approaches to stability analysis, accommodating diverse research needs and computational resources.
Based on Proposition 1 and Lemma 2, the following theorem provides a sufficient condition for the extended dissipativity of MJNN (2.1).
Theorem 1. Setting el=[0n×(l−1)n,In,0n×(14−l)n],l=1,2,⋯,14, the MJNNs (2.1) achieve stochastically stable and extended dissipative if there exist matrices Pi∈R3n×3n>0, Qk∈R3n×3n>0, Rk>0, Hk>0, Mk>0, Uk>0, Li∈Rn×n≥>0, Ni∈Rnw×nw>0, positive definite diagonal matrices Λc, Δc, Vk=diag{vk1,vk2,⋯,vkn}∈R3n×3n, and any matrices Xk satisfying the following conditions, including the inequalities for all i∈S, c=1,2,3, and k=1,2, with the provided scalars α∈(0,1), ς, and ν, and given μk≥0, κk≥1, as well as matrices Ψ1, Ψ2, Ψ3, Ψ4
(1−ν)R1+ςH1−νM1>0, | (2.13) |
R2−νM2>0,H2−ν2U1>0,H2−ν2U2>0, | (2.14) |
[(℘⊥i)TΩ(0,0)℘⊥i√κ1(℘⊥i)TΠT14S∗−ςℜ2(0,0)]<0, | (2.15) |
[(℘⊥i)TΩ(0,ν)℘⊥i√κ1(℘⊥i)TΠT14S∗−ςℜ2(0,ν)]<0, | (2.16) |
[(℘⊥i)TΩ(ς,0)℘⊥i√κ2(℘⊥i)TΠT15ST∗−ςℜ2(ς,0)]<0, | (2.17) |
[(℘⊥i)TΩ(ς,−ν)℘⊥i√κ2(℘⊥i)TΠT15ST∗−ςℜ2(ς,−ν)]<0, | (2.18) |
[Γ11Γ12−CTΨ4D2−CTΨ4B2∗Γ22−DT1Ψ4D2−DT1Ψ4B2∗∗Γ33−DT2Ψ4B2∗∗∗Γ44]>0, | (2.19) |
where
Ω(σ,˙σ)=7∑l=1Ωl,σ=σ(t),ςσ=ς−σ(t),Ω1=ΠT1∑j∈SτijPjΠ1+He[ΠT1PiΠ2],Ω2=ΠT3Q1Π3−ΠT5Q2Π5−(1−˙d)ΠT4(Q1−Q2)Π4,Ω3=He[eT4(V1−V2)W2e11+eT1WT2(K2V2−K1V1)W2e11],Ω4=eT11(σ(R1+M1)+ς22T+σς2G1)e11+eT12((1−˙σ)(ςσM2−σM1)+ςσ(1−˙σ)R2+ς(1−˙σ)ςσ2G2))e12−ςσeT13M2e13−˙σς(ΠT6U1Π6−ΠT7U2Π7)−1ςHe[ΠT6M1Π8+ΠT7M2Π9]−˙σς2(ΠT10G1Π10−ΠT11G2Π11)−1ς2He[ΠT10G1Π12+ΠT11G2Π13],Ω5=−1ςΠT16ℜΠ16−ΠT17T1(σ,˙σ)Π17−ΠT18T2(σ,˙σ)Π18,Ω6=3∑l=1He[(el+3−K1W2el)TΛl(K2W2el−el+3)]+2∑l=1He[((el+3−el+4)−K1W2(el−el+1))T×Δl(K2W2(el−el+1)−(el+3−el+4))]+He[((e4−e6)−K1W2(e1−e3))T×Δ3(K1W2(e1−e3)−(e4−e6))],Ω7=−eT0Ψ1e0−He[eT0Ψ2e14]−eT14Ψ3e14,ℜ=[ˉτℜ1(σ,˙σ)S∗(1+τeκ2ς)ℜ2(σ,˙σ)],ˉτ=(1+(ς−τ)eκ1ς),ℜ1(σ,˙σ)=diag{ℜ01(t),3ℜ01(t)},ℜ2(σ,˙σ)=diag{ℜ02(t),3ℜ02(t)},T1(σ,˙σ)=diag{2T01,4T01},T2(σ,˙σ)=diag{2T02,4T02},ℜ01(t)=(1−˙σ)R1+ςσT+(1−˙σ)σ2G1−˙σM1,ℜ02(t)=R2+˙σM2+ςσ2G2,T01(t)=H−˙σ2U1,T02(t)=H+˙σ2U2,K1=diag{k−1,k−2,⋯,k−n},K2=diag{k+1,k+2,⋯,k+n},P1i=[I,0,0]Pi[I,0,0]T,Γ11=αP1i−CTΨ4C,Γ12=−CTΨ4D1,Γ22=(1−α)P1i−DT1Ψ4D1,Γ33=Li−DT2Ψ4D2,Γ44=Ni−BT2Ψ4B2,Π1=col[e1,e2,e3],Π2=col[e11,(1−˙σ)e12,e13],Πl+3=col[el+1,el+4,el+11],l=0,1,2,Π6+l=col[el+1,el+2,el+7],l=0,1,Π8=col[σe11,d(1−˙σ)e12,e1−(1−˙σ)e2−˙σe7],Π9=col[ςσ(1−˙σ)e12,ςσe13,(1−˙σ)e2−e3+˙σe8],Πl+10=col[el+1,el+7,el+9],l=0,1,Π12=col[σe11,e1−(1−˙σ)e2−˙σe7,e1−(1−˙σ)e7−2˙σe9],Π13=col[ςσe12,(1−˙σ)e2−e3+˙σe8,(1−˙σ)e2−e8+2˙σe10],Π14=col[e1−e2,e1+e2−2e7],Π15=col[e2−e3,e2+e3−2e8],Π16=col[Π14,Π15],Π17=col[e1−e7,−12e1−e7+3e9],Π18=col[e2−e8,−12e2−e8+3e10],℘i=−Aie1+W0ie4+W1ie5+B1e14−e11,e0=Ce1+D1e2+D2e4+B2e14. |
Proof. Select the subsequent LKF candidate:
V(§t)=6∑l=0Vl(§t), | (2.20) |
where
V1(§t)=ηT1(t)Piη1(t),V2(§t)=∫tt−σ(t)χT1(s)Q1χ1(s)ds+∫t−σ(t)t−ςχT1(s)Q2χ1(s)ds,V3(§t)=2n∑l=1∫W2l§l(t)0[v1l(πl(s)−k−ls)+v2l(k+ls−πl(s))]ds,V4(§t)=∫0−σ(t)∫tt+θ˙§T(s)R1˙§(s)dsdθ+∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)R2˙§(s)dsdθ,V5(§t)=∫0−ς∫0θ∫tt+u˙§T(s)H˙§(s)dsdudθ,V6(§t)=σ(t)∫tt−σ(t)˙§T(s)M1˙§(s)ds+ςσ(t)∫t−σ(t)t−ς˙§T(s)M2˙§(s)ds−σ(t)ςχT2(t)M1χ2(t)−ςσ(t)ςχT3(t)M2χ3(t). |
The infinitesimal operator along system (2.1) is denoted by L. One has
3∑c=1LVc(§t)=ξT(t)(Ω1+Ω2+Ω3)ξ(t), | (2.21) |
5∑l=4LVl(§t)=˙§T(t)(σ(t)R1+ς22H)˙§(t)+ςσ(t)(1−˙σ(t))˙§T(t−σ(t))×R2˙§(t−σ(t))−(1−˙σ(t))∫tt−σ(t)˙§T(s)R1˙§(s)ds−∫t−σ(t)t−ς˙§T(s)R2˙§(s)ds−∫0−ς∫tt+θ˙§T(s)H˙§(s)dsdθ, | (2.22) |
LV6(§t)=˙σ(t)∫tt−σ(t)˙§T(s)M1˙§(s)ds−˙σ(t)∫t−σ(t)t−ς˙§T(s)M2˙§(s)ds+σ(t)˙§T(t)M1˙§(t)−ςσ(t)˙§T(t−h)×M2˙§(t−ς)−(1−˙σ(t))˙§T(t−σ(t))×(σ(t)M1−ςσ(t)M2)˙§(t−σ(t))−˙σ(t)ςχT2(t)M1χ2(t)−2σ(t)ςχT2(t)M1˙χ2(t)+˙σ(t)ςχT3(t)M2χ3(t)−2ςσ(t)ςχT3(t)M2˙χ3(t), | (2.23) |
LV0(§t)≤˙§T(t)σ(t)ς2G1˙§(t)+(1−˙σ(t))˙§T(t−σ(t))ςσ(t)ς2G2˙§(t−σ(t))−˙σ(t)ς2ξT4(t)Z1ξ4(t)−2σ(t)ς2ξT4(t)Z1˙ξ4(t)+˙σ(t)ς2ξT5(t)Z2ξ5(t)−2ςσ(t)ς2ξT5(t)Z2˙ξ5(t)−(1−˙σ(t))σ(t)2∫tt−σ(t)˙§T(s)G1˙§(s)ds−ςσ(t)2∫t−σ(t)t−ς˙§T(s)G2˙§(s)ds+˙σ(t)2∫0−σ(t)∫tt+θ˙§T(s)G1˙§(s)dsdθ−˙σ(t)2∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)G2˙§(s)dsdθ. | (2.24) |
Referring to (2.22), to enhance the incorporation of TVDs information, we divide [0,ς] into [0,σ(t)]∪[σ(t),ς], resulting in the following expression:
−∫0−ς∫tt+θ˙§T(s)H˙§(s)dsdθ=−∫0−σ(t)∫tt+θ˙§T(s)H˙§(s)dsdθ−ςσ(t)∫tt−σ(t)˙§T(s)H˙§(s)ds−∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)H˙§(s)dsdθ. | (2.25) |
Combining with (2.22)–(2.25), one has
6∑c=4LVc(§t)+LV∗(§t)≤ξT(t)Ω4ξ(t)−∫tt−σ(t)˙§T(s)R01(t)˙§(s)ds−∫t−σ(t)t−ς˙§T(s)R02(t)˙§(s)ds−∫0−σ(t)∫tt+θ˙§T(s)T01(t)˙§(s)dsdθ−∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)T02(t)˙§(s)dsdθ. | (2.26) |
Taking into account (2.3), (2.13), and (2.14), alongside the presence of positive definite matrices Rk, Mk, Uk, and Hk (for k=1,2), we obtain
R01(t)>0,R02(t)>0,T01(t)>0,T02(t)>0. |
By utilizing the inequality condition in [4], we can approximate the single integral terms in (2.26) that depend on R, M, U, and H as follows:
−∫tt−σ(t)˙§T(s)R01(t)˙§(s)ds≤−1σ(t)ξT(t)ΠT14ℜ1(σ(t),˙σ(t))Π14ξ(t) | (2.27) |
and
−∫t−σ(t)t−ς˙xT(s)R02(t)˙x(s)ds≤−1ςσ(t)ξT(t)ΠT15ℜ2(σ(t),˙σ(t))Π15ξ(t). | (2.28) |
By applying Lemma 2, one has
−1σ(t)ΠT14ℜ1(σ(t),˙σ(t))Π14−1ςσ(t)ΠT15×ℜ2(σ(t),˙σ(t))Π15≤−1ς[Π14Π15]T[ℜ1(σ(t),˙σ(t))+T1S∗T2][Π14Π15]≤−1ςΠT16(ℜ−ℑ)Π16, | (2.29) |
where
T1=(1−eκ1ςσ(t)ςT11),T2=ℜ2(σ(t),˙σ(t)+eκ2σ(t)ςT12,T11=ℜ1(σ(t),˙σ(t))−Sℜ−12(σ(t),˙σ(t))ST,T12=ℜ2(σ(t),˙σ(t))−STℜ−11(σ(t),˙σ(t))S,ℑ=diag{eκ1ςσ(t)ςSℜ−12(σ(t),˙σ(t))ST,eκ2σ(t)ςSTℜ−11(σ(t),˙σ(t))S}. |
By employing the inequality in [5], we can respectively estimate the H- and U-dependent terms in (2.26) as follows:
−∫0−σ(t)∫tt+θ˙§T(s)T01(t)˙§(s)dsdθ≤−ξT(t)Π17T1(σ(t),˙σ(t))Π17ξ(t), | (2.30) |
−∫−σ(t)−ς∫t−σ(t)t+θ˙§T(s)T02(t)˙§(s)dsdθ=−ξT(t)Π18T2(σ(t),˙σ(t))Π18ξ(t). | (2.31) |
Combining (2.26) and (2.29)–(2.31), we obtain:
6∑c=4LVc(§t)+LV0(§t)≤ξT(t)(Ω4+Ω5+1ςΠT16ℑΠ16)ξ(t). | (2.32) |
By considering (2.4) and (2.5), we can obtain the following inequalities for a=1,2,3.
λa(s)=2[π(W2x(s))−K1W2§(s)]T×Λc[K2W2§(s)−π(W2§(s))]≥0,δa(s1,s2)=2[π(W2§(s1))−π(W2§(s2))−K1W2(§(s1)−§(s2))]T×Δc[K2W2(§(s1)−§(s2))−π(W2§(s1))+π(W2§(s2))]≥0. |
Thus, we can obtain
λ1(t)+λ2(t−σ(t))+λ3(t−ς)=2[π(W2ρ(s))−K1W2ρ(s)]T×Λi[K2W2ρ(s)−π(W2ρ(s))]≥0,δ(t,t−σ(t))+δ(t−σ(t),t−ς)+δ(t,t−ς)=2[π(W2ρ(s1))−π(W2ρ(s2))−K1W2(ρ(s1)−ρ(s2))]T×Δi[K2W2(ρ(s1)−ρ(s2))−π(W2ρ(s1))+π(W2ρ(s2))]≥0. | (2.33) |
Further, the cost function can be expressed as follows:
JT=∫T0(†T(t)Ψ1†(t)+2†T(t)Ψ2ρ(t)+ρT(t)Ψ3ρ(t))dt. | (2.34) |
Considering (2.21) and (2.32)–(2.34), it yields
∫T0LV(§t)dt−JT≤∫T0ξT(t)Υ(σ(t),˙σ(t))ξ(t)dt, | (2.35) |
where
Υ(σ(t),˙σ(t))=7∑l=1Ωl+1ςΠT16ℑΠ16. |
In the context of the problem, ℓl(l=1,2,...,4) are real matrix combinations that do not depend on σ(t) and ˙σ(t). Therefore, following the approach presented in [26], we have Υ(σ(t),˙σ(t))<0.
Considering the condition ℘iξ(t)=0, and applying Finsler's Lemma [3], we can deduce
(℘⊥i)TΥ(σ(t),˙σ(t))℘⊥i<0⇒ξT(t)Υ(σ(t),˙σ(t))ξ(t)<0. |
When Ψ1≤0, a nonnegative scalar ϵ can be found, satisfying the subsequent inequality for the case of w(t)=0.
LV(§t)<−ϵ|ξ(t)|2. |
This indicates the stochastic stability of MJNNs (2.1).
From (2.35), one has
JT≥∫T0LV(§t)dt=V(§t)−V(§0)≥§T(T)P1i§(t)−V(§0). | (2.36) |
Based on (2.19), it can be observed that Γ33=Li−DT2Ψ4D2>0 and Γ44=Ni−BT2Ψ4B2>0. Given that Li and Ni stand as symmetric positive definite matrices, the condition Ψ4≥0 is assured. Furthermore, it becomes apparent that Ψ4<0 is valid for all instances involving matrices Li≥0 and Ni≥0. Regarding the unrestricted parameter Ψ4, we will explore two distinct scenarios as outlined below:
a: In the case of Ψ4=0, by setting ϱ≤−V(§0), we can derive from (2.9) and (2.36) the following result:
JT≥ϱ. | (2.37) |
b: In the case of Ψ4≠0, based on Ψ1=Ψ2=0 and Ψ3>0, along with (2.36), we obtain the following result:
JT=∫T0ρT(t)Ψ3ρ(t)dt≥0. | (2.38) |
For 0≤t≤T and 0≤t−σ(t)≤T, and considering matrices Li>0, Ni>0, and Ψ3>0, we have the following result for all i∈S:
JT≥Jt≥§T(t)P0i§(t)−V(§0)≥§T(t)P0i§(t)−V(§0)−πT(W2§(t))Liπ(W2§(t))−ρT(t)Niρ(t), | (2.39) |
JT≥Jt−σ(t)≥§T(t−σ(t))P0i§(t−σ(t))−V(§0)−πT(W2§(t))Liπ(W2§(t))−ρT(t)Niρ(t). | (2.40) |
Therefore, by choosing a constant 0<α<1, we obtain
JT≥(1−α)§T(t−σ(t))P0i§(t−σ(t))+α§T(t)P0i§(t)−V(§0)−πT(W2§(t))Liπ(W2§(t))−ρT(t)Niρ(t). | (2.41) |
Then, one has
†T(t)Ψ4†(t)=(1−α)§T(t−σ(t))P0i§(t−σ(t))+α§T(t)P0i§(t)+ρT(t)Niρ(t)+πT(W2§(t))Liπ(W2§(t))−φT(t)Θφ(t), | (2.42) |
where
φ(t)=col[§(t),§(t−σ(t)),π(W2§(t)),ρ(t)],Θ=[Γ11Γ12−CTΨ4D2−CTΨ4B2∗Γ22−DT1Ψ4D2−DT1Ψ4B2∗∗Γ33−DT2Ψ4B2∗∗∗Γ44]. |
Combining Θ>0, (2.41), and (2.42), one has
JT≥†T(ts)Ψ4†(t)−2πT(W2§(t))Liπ(W2§(t))−2ρT(t)Niρ(t)−V(§0). | (2.43) |
Recalling (2.5), we find that |π(W2§(t))|≤||K2W2||⋅|§(t)|. Now, let's consider two cases for Ψ4≠0:
b(1): For Ψ4>0, considering ρ(t)∈L2[0,∞), there exists a constant ϱ≤0 such that
†T(t)Ψ4†(t)≥sup0≤t≤T†T(t)Ψ4†(t)+ϱ. | (2.44) |
Recalling (2.43), one has
JT≥sup0≤t≤T†T(t)Ψ4†(t)+ϱ. | (2.45) |
b(2): For Ψ4<0, one has
sup0≤t≤T†T(t)Ψ4†(t)≤0. | (2.46) |
Combining (2.38) and (2.46), one has
JT=∫T0ρT(s)Ψ3ρ(s)ds≥sup0≤t≤T†T(t)Ψ4†(t)+ϱ. | (2.47) |
Recalling (2.45) and (2.47), we have
JT=∫T0ρT(s)Ψ3ρ(s)ds≥sup0≤t≤T†T(t)Ψ4†(t)+ϱ. | (2.48) |
Based on Definition 1, we can conclude the extended dissipativity of system (2.1), which completes the proof.
Remark 3. Theorem 1 proposes a new condition with the help of NDPT and ETRCI, ensuring the stochastically stable and extended dissipative behavior of MJNNs (2.1). The advantages of Theorem 1 can be highlighted.
(i) The condition captures more information about TVDs, involving its derivative in association with σ(t), ς, ˙σ(t), x1(t), x2(t), x3(t), and x4(t). Additionally, the incorporation of double integrals and the elimination of zero components result in outcomes that are less conservative when contrasted with those presented in [18,19].
(ii) The ETRCI is introduced to deal with the term −∫tt−ς˙§T(s)R˙§(s)ds, which distinguishes it from the approaches in [18,19]. The flexibility of two parameters κ1 and κ2 in Theorem 1 allows independent adjustments, leading to better solutions.
(iii) It is noted that extended dissipativity [13,14] includes strict dissipativity [11] and passivity [27], which enhances the generality of the proposed method. Furthermore, based on the NDPT and ETRCI, the dissipative characteristics are superior compared with some existing works [27,28].
(iv) The NDPT and ETRCI presented in this paper primarily enhances the interconnectedness between time-delay information and the slack matrices, thereby reducing conservativeness. Consequently, the approach is well-suited for time delayed systems such as delayed semi-Markovian NNs.
Remark 4. Note that theoretical proof in [25] shows that the conservatism of a stability criterion can be reduced by increasing the ply of integral terms in LKF. Thus, a suitable multiple integral-type LKF can further reduce the conservatism of the stability criterion. Consequently, employing a multiple integral-type LKF can further diminish the conservatism of the stability criterion. Additionally, the literature [8] reveals the introduction of exponentially extended dissipativity, which not only consolidates various performance measures-such as extended dissipative, exponential H∞, l2−l∞ performance within a cohesive framework, but also delineates both the extended dissipativity and the transient behavior of switched NNs with greater clarity upon establishing the decay rate. Therefore, future research endeavors should focus on the analysis of less conservative exponentially extended dissipativity employing multiple DPT, which promises to yield significant insights.
We consider MJNNs (2.1) with parameters presented in [18]
A1=[2002],A2=[2.2001.8],K1=D2=0,W01=[11−1−1],W02=[−1−10.5−1],W11=[0.88111],W12=[−0.50.60.70.8],W2=[1001],B1=[0.04030.6771],K2=diag{0.4,0.8},C=[−0.3775−0.2959],D1=[0.2532−0.1684]. |
We choose Π=[−445−5] and Ψl(l=1,2,3,4) from Table 1. For simplicity, we set μ=μ1=μ2 and κ=κ1=κ2, and consider the following two cases as special cases of extended dissipativity.
Performance | Ψ4 | Ψ1 | Ψ2 | Ψ3 |
H∞ | 0 | -I | 0 | γ2I |
L2−L∞ | I | 0 | 0 | γ2I |
Case 1. The H∞ performance is investigated by selecting B2=0.1184, and Table 2 presents the minimum H∞ performance of γ and NDVs computed using different techniques.
Criteria | ν=0.8 | ν=0.9 | NDVs |
[18, Theorem 3] | 0.8429 | 2.7968 | 125 |
[19, Theorem 1] | 0.6113 | 1.7642 | 252 |
Theorem 1(μ=0,κ=0.1) | 0.6541 | 1.7712 | 252 |
Theorem 1(μ=1,κ=0.1) | 0.5917 | 1.7504 | 266 |
Theorem 1(μ=0.4,κ=0.1) | 0.6571 | 1.8154 | 266 |
Theorem 1(μ=0.2,κ=0.1) | 0.6897 | 1.8401 | 266 |
Theorem 1(μ=0.2,κ=0.3) | 0.6994 | 1.8577 | 266 |
Case 2. The L2−L∞ performance is evaluated under the condition that B2=0. Table 3 presents the minimum L2−L∞ performance of γ and NDVs computed using different techniques.
Criteria | ν=0.8 | ν=0.9 | NDVs |
[18, Theorem 3] | 0.6684 | 1.3591 | 125 |
[19, Theorem 1] | 0.4241 | 0.8744 | 252 |
Theorem 1(μ=0,κ=0.1) | 0.4271 | 0.8794 | 252 |
Theorem 1(μ=2,κ=0.1) | 0.4012 | 0.8514 | 266 |
Theorem 1(μ=0.4,κ=0.1) | 0.4475 | 0.8651 | 266 |
Theorem 1(μ=0.2,κ=0.1) | 0.4794 | 0.8841 | 266 |
Theorem 1(μ=0.2,κ=0.3) | 0.5047 | 0.8924 | 266 |
Based on Cases 1 and 2, the subsequent observations can be made:
(i) Tables 2 and 3 effectively illustrate the adaptable nature of the NDPT approach (2.12), which can be tailored according to distinct values of τ.
(ii) The ETRCI assumes a pivotal role in mitigating excessive caution.
(iii) The proposition presented in this paper regarding MJNNs (2.1) showcases a comparatively diminished level of conservatism in contrast to the outcomes derived in [18,19].
Although MJNNs have been implemented in various fields such as biomolecules [29], this paper primarily focuses on theoretical methodological innovation. By introducing a novel time-delay product approach, it aims to reduce the conservativeness of time-delay Markovian neural systems. Therefore, extending the method presented in this paper to practical applications is not only meaningful but also constitutes a worthwhile direction for future research.
This paper delves into the issue of extended dissipative analysis of MJNNs using the delay-product-type functional method. The approach introduced here involves the creation of a novel NDPT functional, achieved through the utilization of Wirtinger-based double integral inequality combined with the S-procedure lemma. Unlike the approach in [19], this fresh functional overcomes the presence of incomplete components, thereby facilitating enhanced integration of supplementary system information. By melding the exponential type reciprocally convex inequality with the Wirtinger-based integral inequality, we are able to estimate the derivative of the constructed Lyapunov-Krasovskii function. This leads to the derivation of a delay-dependent extended dissipativity condition tailored to address the specificities of delayed MJNNs. The effectiveness of the proposed methodology is demonstrated convincingly through a numerical example.
Wenlong Xue: Writing-original draft; Yufeng Tian: Writing-review & editing; Zhenghong Jin: Validation. All authors have read and approved the final version of the manuscript for publication.
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
The authors declare no conflict of interest.
[1] |
N. Huo, B. Li, Y. Li, Global exponential stability and existence of almost periodic solutions in distribution for Clifford-valued stochastic high-order Hopfield neural networks with time-varying delays, AIMS Mathematics, 7 (2022), 3653–3679. http://dx.doi.org/10.3934/math.2022202 doi: 10.3934/math.2022202
![]() |
[2] |
H. Qiu, L. Wan, Z. Zhou, Q. Zhang, Q. Zhou, Global exponential periodicity of nonlinear neural networks with multiple time-varying delays, AIMS Mathematics, 8 (2023), 12472–12485. http://dx.doi.org/10.3934/math.2023626 doi: 10.3934/math.2023626
![]() |
[3] |
J. Geromel, J. da Cruz, On the robustness of optimal regulators for nonlinear discrete-time systems, IEEE T. Automat. Contr., 32 (1987), 703–710. http://dx.doi.org/10.1109/TAC.1987.1104696 doi: 10.1109/TAC.1987.1104696
![]() |
[4] |
A. Seuret, F. Gouaisbaut, Wirtinger-based integral inequality: application to time-delay systems, Automatica, 49 (2013), 2860–2866. http://dx.doi.org/10.1016/j.automatica.2013.05.030 doi: 10.1016/j.automatica.2013.05.030
![]() |
[5] |
M. Park, O. Kwon, J. Park, S. Lee, E. Cha, Stability of time-delay systems via Wirtinger-based double integral inequality, Automatica, 55 (2015), 204–208. http://dx.doi.org/10.1016/j.automatica.2015.03.010 doi: 10.1016/j.automatica.2015.03.010
![]() |
[6] |
Y. Tian, Y. Yang, X. Ma, X. Su, Stability of discrete-time delayed systems via convex function-based summation inequality, Appl. Math. Lett., 145 (2023), 108764. http://dx.doi.org/10.1016/j.aml.2023.108764 doi: 10.1016/j.aml.2023.108764
![]() |
[7] |
Z. Wang, S. Ding, Q. Shan, H. Zhang, Stability of recurrent neural networks with time-varying delay via flexible terminal method, IEEE T. Neur. Net. Lear., 28 (2017), 2456–2463. http://dx.doi.org/10.1109/TNNLS.2016.2578309 doi: 10.1109/TNNLS.2016.2578309
![]() |
[8] |
Y. Tian, X. Su, C. Shen, X. Ma, Exponentially extended dissipativity-based filtering of switched neural networks, Automatica, 161 (2024), 111465. http://dx.doi.org/10.1016/j.automatica.2023.111465 doi: 10.1016/j.automatica.2023.111465
![]() |
[9] |
W. Lin, Y. He, C. Zhang, M. Wu, Stability analysis of neural networks with time-varying delay: enhanced stability criteria and conservatism comparison, Commun. Nonlinear Sci., 54 (2018), 118–135. http://dx.doi.org/10.1016/j.cnsns.2017.05.021 doi: 10.1016/j.cnsns.2017.05.021
![]() |
[10] |
C. Zhang, Y. He, L. Jiang, Q. Wang, M.Wu, Stability analysis of discrete-time neural networks with time-varying delay via an extended reciprocally convex matrix inequality, IEEE T. Cybernetics, 47 (2017), 3040–3049. http://dx.doi.org/10.1109/TCYB.2017.2665683 doi: 10.1109/TCYB.2017.2665683
![]() |
[11] |
R. Vadivel, P. Hammachukiattikul, N. Gunasekaran, R. Saravanakumar, H. Dutta, Strict dissipativity synchronization for delayed static neural networks: an event-triggered scheme, Chaos Soliton. Fract., 150 (2021), 111212. http://dx.doi.org/10.1016/j.chaos.2021.111212 doi: 10.1016/j.chaos.2021.111212
![]() |
[12] |
B. Zhang, W. Zheng, S. Xu, Filtering of Markovian jump delay systems based on a new performance index, IEEE T. Circuits-I, 60 (2013), 1250–1263. http://dx.doi.org/10.1109/TCSI.2013.2246213 doi: 10.1109/TCSI.2013.2246213
![]() |
[13] |
S. Shanmugam, R. Vadivel, M. Rhaima, H. Ghoudi, Improved results on an extended dissipative analysis of neural networks with additive time-varying delays using auxiliary function-based integral inequalities, AIMS Mathematics, 8 (2023), 21221–21245. http://dx.doi.org/10.3934/math.20231082 doi: 10.3934/math.20231082
![]() |
[14] |
R. Anbuvithya, S. Dheepika Sri, R. Vadivel, N. Gunasekaran, P. Hammachukiattikul, Extended dissipativity and non-fragile synchronization for recurrent neural networks with multiple time-varying delays via sampled-data control, IEEE Access, 9 (2021), 31454–31466. http://dx.doi.org/10.1109/ACCESS.2021.3060044 doi: 10.1109/ACCESS.2021.3060044
![]() |
[15] |
R. Zhang, D. Zeng, X. Liu, S. Zhong, J. Cheng, New results on stability analysis for delayed Markovian generalized neural networks with partly unknown transition rates, IEEE T. Neur. Net. Lear., 30 (2019), 3384–3395. http://dx.doi.org/10.1109/TNNLS.2019.2891552 doi: 10.1109/TNNLS.2019.2891552
![]() |
[16] |
R. Zhang, D. Zeng, J. Park, Y. Liu, S. Zhong, A new approach to stochastic stability of Markovian neural networks with generalized transition rates, IEEE T. Neur. Net. Lear., 30 (2019), 499–510. http://dx.doi.org/10.1109/TNNLS.2018.2843771 doi: 10.1109/TNNLS.2018.2843771
![]() |
[17] |
W. Lin, Y. He, M. Wu, Q. Liu, Reachable set estimation for Markovian jump neural networks with time-varying delay, Neural Networks, 108 (2018), 527–532. http://dx.doi.org/10.1016/j.neunet.2018.09.011 doi: 10.1016/j.neunet.2018.09.011
![]() |
[18] |
W. Lin, Y. He, C. Zhang, M. Wu, S. Shen, Extended dissipativity analysis for Markovian jump neural networks with time-varying delay via delay-product-type functionals, IEEE T. Neur. Net. Lear., 30 (2019), 2528–2537. http://dx.doi.org/10.1109/TNNLS.2018.2885115 doi: 10.1109/TNNLS.2018.2885115
![]() |
[19] |
Y. Tian, Z. Wang, Extended dissipativity analysis for Markovian jump neural networks via double-integral-based delay-product-type Lyapunov functional, IEEE T. Neur. Net. Lear., 32 (2021), 3240–3246. http://dx.doi.org/10.1109/TNNLS.2020.3008691 doi: 10.1109/TNNLS.2020.3008691
![]() |
[20] |
P. Park, J. Ko, C. Jeong, Reciprocally convex approach to stability of systems with time-varying delays, Automatica, 47 (2011), 235–238. http://dx.doi.org/10.1016/j.automatica.2010.10.014 doi: 10.1016/j.automatica.2010.10.014
![]() |
[21] |
C. Zhang, Y. He, L. Jiang, M. Wu, Q. Wang, An extended reciprocally convex matrix inequality for stability analysis of systems with time-varying delay, Automatica, 85 (2017), 481–485. http://dx.doi.org/10.1016/j.automatica.2017.07.056 doi: 10.1016/j.automatica.2017.07.056
![]() |
[22] |
X. Zhang, Q. Han, A. Seuret, F. Gouaisbaut, An improved reciprocally convex inequality and an augmented Lyapunov-Krasovskii functional for stability of linear systems with time-varying delay, Automatica, 84 (2017), 222–226. http://dx.doi.org/10.1016/j.automatica.2017.04.048 doi: 10.1016/j.automatica.2017.04.048
![]() |
[23] |
Y. Tian, Z. Wang, Stability analysis and generalized memory controller design for delayed T-S fuzzy systems via flexible polynomial-based functions, IEEE T. Fuzzy Syst., 30 (2022), 728–740. http://dx.doi.org/10.1109/TFUZZ.2020.3046338 doi: 10.1109/TFUZZ.2020.3046338
![]() |
[24] | V. Yakubovich, The S-procedure in nonlinear control theory, Vestnik Leningrad Univ. Mathe., 4 (1977), 73–93. |
[25] |
Z. Wang, S. Ding, H. Zhang, Hierarchy of stability criterion for time-delay systems based on multiple integral approach, Appl. Math. Comput., 314 (2017), 422–428. http://dx.doi.org/10.1016/j.amc.2017.07.016 doi: 10.1016/j.amc.2017.07.016
![]() |
[26] |
A. Seuret, F. Gouaisbaut, Stability of linear systems with time-varying delays using Bessel-Legendre inequalities, IEEE T. Automat. Contr., 63 (2018), 225–232. http://dx.doi.org/10.1109/TAC.2017.2730485 doi: 10.1109/TAC.2017.2730485
![]() |
[27] |
R. Vadivel, P. Hammachukiattikul, Q. Zhu, N. Gunasekaran, Event-triggered synchronization for stochastic delayed neural networks: passivity and passification case, Asian J. Control, 25 (2023), 2681–2698. http://dx.doi.org/10.1002/asjc.2965 doi: 10.1002/asjc.2965
![]() |
[28] |
R. Sakthivel, S. Mohanapriya, C. Ahn, P. Selvaraj, State estimation and dissipative-based control design for vehicle lateral dynamics with probabilistic faults, IEEE T. Ind. Electron., 65 (2018), 7193–7201. http://dx.doi.org/10.1109/TIE.2018.2793253 doi: 10.1109/TIE.2018.2793253
![]() |
[29] |
M. Ghorbani, S. Prasad, J. Klauda, B. Brooks, GraphVAMPNet, using graph neural networks and variational approach to Markov processes for dynamical modeling of biomolecules, J. Chem. Phys., 156 (2022), 184103. http://dx.doi.org/10.1063/5.0085607 doi: 10.1063/5.0085607
![]() |
Performance | Ψ4 | Ψ1 | Ψ2 | Ψ3 |
H∞ | 0 | -I | 0 | γ2I |
L2−L∞ | I | 0 | 0 | γ2I |
Criteria | ν=0.8 | ν=0.9 | NDVs |
[18, Theorem 3] | 0.8429 | 2.7968 | 125 |
[19, Theorem 1] | 0.6113 | 1.7642 | 252 |
Theorem 1(μ=0,κ=0.1) | 0.6541 | 1.7712 | 252 |
Theorem 1(μ=1,κ=0.1) | 0.5917 | 1.7504 | 266 |
Theorem 1(μ=0.4,κ=0.1) | 0.6571 | 1.8154 | 266 |
Theorem 1(μ=0.2,κ=0.1) | 0.6897 | 1.8401 | 266 |
Theorem 1(μ=0.2,κ=0.3) | 0.6994 | 1.8577 | 266 |
Criteria | ν=0.8 | ν=0.9 | NDVs |
[18, Theorem 3] | 0.6684 | 1.3591 | 125 |
[19, Theorem 1] | 0.4241 | 0.8744 | 252 |
Theorem 1(μ=0,κ=0.1) | 0.4271 | 0.8794 | 252 |
Theorem 1(μ=2,κ=0.1) | 0.4012 | 0.8514 | 266 |
Theorem 1(μ=0.4,κ=0.1) | 0.4475 | 0.8651 | 266 |
Theorem 1(μ=0.2,κ=0.1) | 0.4794 | 0.8841 | 266 |
Theorem 1(μ=0.2,κ=0.3) | 0.5047 | 0.8924 | 266 |
Performance | Ψ4 | Ψ1 | Ψ2 | Ψ3 |
H∞ | 0 | -I | 0 | γ2I |
L2−L∞ | I | 0 | 0 | γ2I |
Criteria | ν=0.8 | ν=0.9 | NDVs |
[18, Theorem 3] | 0.8429 | 2.7968 | 125 |
[19, Theorem 1] | 0.6113 | 1.7642 | 252 |
Theorem 1(μ=0,κ=0.1) | 0.6541 | 1.7712 | 252 |
Theorem 1(μ=1,κ=0.1) | 0.5917 | 1.7504 | 266 |
Theorem 1(μ=0.4,κ=0.1) | 0.6571 | 1.8154 | 266 |
Theorem 1(μ=0.2,κ=0.1) | 0.6897 | 1.8401 | 266 |
Theorem 1(μ=0.2,κ=0.3) | 0.6994 | 1.8577 | 266 |
Criteria | ν=0.8 | ν=0.9 | NDVs |
[18, Theorem 3] | 0.6684 | 1.3591 | 125 |
[19, Theorem 1] | 0.4241 | 0.8744 | 252 |
Theorem 1(μ=0,κ=0.1) | 0.4271 | 0.8794 | 252 |
Theorem 1(μ=2,κ=0.1) | 0.4012 | 0.8514 | 266 |
Theorem 1(μ=0.4,κ=0.1) | 0.4475 | 0.8651 | 266 |
Theorem 1(μ=0.2,κ=0.1) | 0.4794 | 0.8841 | 266 |
Theorem 1(μ=0.2,κ=0.3) | 0.5047 | 0.8924 | 266 |