Loading [MathJax]/jax/output/SVG/jax.js
Theroy article Special Issues

Projective synchronization for quaternion-valued memristor-based neural networks under time-varying delays

  • In this paper, the projective synchronization of quaternion-valued memristor-based neural networks with time-varing delays was studied. First, by utilizing set-valued map and differential inclusion theories, we reformulated the networks as an uncertain system with interval parameters. Then, through designing a novel controller and utilizing Lyapunov function and Young's inequality, several new synchronization conditions for projection synchronization of quaternion-valued memristor-based neural networks were obtained. Finally, the effectiveness of this method was demonstrated through a numerical example, underscoring its practical applicability.

    Citation: Jun Guo, Yanchao Shi, Yanzhao Cheng, Weihua Luo. Projective synchronization for quaternion-valued memristor-based neural networks under time-varying delays[J]. Networks and Heterogeneous Media, 2024, 19(3): 1156-1181. doi: 10.3934/nhm.2024051

    Related Papers:

    [1] Lili Jia, Changyou Wang, Zongxin Lei . Synchronization of nonautonomous neural networks with Caputo derivative and time delay. Networks and Heterogeneous Media, 2023, 18(1): 341-358. doi: 10.3934/nhm.2023013
    [2] Wenlian Lu, Fatihcan M. Atay, Jürgen Jost . Consensus and synchronization in discrete-time networks of multi-agents with stochastically switching topologies and time delays. Networks and Heterogeneous Media, 2011, 6(2): 329-349. doi: 10.3934/nhm.2011.6.329
    [3] Tingting Zhu . Synchronization of the generalized Kuramoto model with time delay and frustration. Networks and Heterogeneous Media, 2023, 18(4): 1772-1798. doi: 10.3934/nhm.2023077
    [4] Chenchen Li, Chunyan Zhang, Lichao Feng, Zhihui Wu . Note on prescribed-time stability of impulsive piecewise-smooth differential systems and application in networks. Networks and Heterogeneous Media, 2024, 19(3): 970-991. doi: 10.3934/nhm.2024043
    [5] Young-Pil Choi, Seung-Yeal Ha, Seok-Bae Yun . Global existence and asymptotic behavior of measure valued solutions to the kinetic Kuramoto--Daido model with inertia. Networks and Heterogeneous Media, 2013, 8(4): 943-968. doi: 10.3934/nhm.2013.8.943
    [6] Seung-Yeal Ha, Se Eun Noh, Jinyeong Park . Practical synchronization of generalized Kuramoto systems with an intrinsic dynamics. Networks and Heterogeneous Media, 2015, 10(4): 787-807. doi: 10.3934/nhm.2015.10.787
    [7] Yinlin Ye, Yajing Li, Hongtao Fan, Xinyi Liu, Hongbing Zhang . SLeNN-ELM: A shifted Legendre neural network method for fractional delay differential equations based on extreme learning machine. Networks and Heterogeneous Media, 2023, 18(1): 494-512. doi: 10.3934/nhm.2023020
    [8] Mahendra Piraveenan, Mikhail Prokopenko, Albert Y. Zomaya . On congruity of nodes and assortative information content in complex networks. Networks and Heterogeneous Media, 2012, 7(3): 441-461. doi: 10.3934/nhm.2012.7.441
    [9] Yacine Chitour, Guilherme Mazanti, Mario Sigalotti . Stability of non-autonomous difference equations with applications to transport and wave propagation on networks. Networks and Heterogeneous Media, 2016, 11(4): 563-601. doi: 10.3934/nhm.2016010
    [10] Pau Erola, Albert Díaz-Guilera, Sergio Gómez, Alex Arenas . Modeling international crisis synchronization in the world trade web. Networks and Heterogeneous Media, 2012, 7(3): 385-397. doi: 10.3934/nhm.2012.7.385
  • In this paper, the projective synchronization of quaternion-valued memristor-based neural networks with time-varing delays was studied. First, by utilizing set-valued map and differential inclusion theories, we reformulated the networks as an uncertain system with interval parameters. Then, through designing a novel controller and utilizing Lyapunov function and Young's inequality, several new synchronization conditions for projection synchronization of quaternion-valued memristor-based neural networks were obtained. Finally, the effectiveness of this method was demonstrated through a numerical example, underscoring its practical applicability.



    Notations: Throughout this paper, R, C, and Q denote the real, complex, and quaternion fields, respectively. The notation adheres to standard mathematical conventions.

    The memristor, a pioneering circuit component that encapsulates the interplay between magnetic flux and electric charge, made its debut in scientific literature through the visionary work of Chua [1] in 1971. This innovative element distinguishes itself from traditional circuit components due to its nonlinear resistance and inherent memory capabilities. Its mnemonic attribute bears a striking resemblance to the synaptic plasticity observed in the neural connections of the human brain. Leveraging this distinctive feature, the memristor has transcended its role as an electrical resistor to become a cornerstone in simulating the cognitive functions of the human brain. Consequently, the exploration of memristor-based neural networks (MBNNs) has emerged as a vibrant field of study, with researchers delving into the intricate dynamics that these networks exhibit [2,3,4]. MBNNs are characterized as state-dependent dynamical systems, with coefficients that are inherently linked to their instantaneous state. The discontinuous nature of these systems can give rise to a spectrum of complex nonlinear phenomena, including but not limited to chaos, oscillations, and instability. The investigation of these dynamical traits is not merely an academic pursuit; it holds profound significance in both theoretical exploration and practical applications, paving the way for advancements in neuromorphic engineering and computational neuroscience.

    The quaternion, a mathematical construct first articulated by the Irish mathematician Sir William Rowan Hamilton in 1843, stands as an innovative extension of the real and complex number systems. Unlike their simpler counterparts, quaternions introduce a non-commutative multiplication, a feature that has historically posed challenges and led to a period of relative dormancy in their study. However, in recent years, there has been a resurgence of interest and a broadening scope of applications for quaternions. They have demonstrated remarkable utility across a spectrum of disciplines, including artificial intelligence [5], image processing [6], quantum mechanics [7], and aerospace technology [8]. These applications have not only reinvigorated the study of quaternions but also unveiled their potential to address complex problems in a multidimensional context. Especially in image processing, the nuanced representational power of quaternions has been harnessed. They are used to encode the three primary color channels within the imaginary components of a quaternion, while the real part is often reserved for the alpha channel or other metadata. This elegant mapping of color images to pure quaternions has opened new avenues for the representation and manipulation of visual data.

    Quaternion-valued neural networks (QVNNs) distinguish themselves from their complex-numbered counterparts by employing quaternions in every aspect of their architecture—states, connection weights, and activation functions. This holistic approach to quaternion integration endows QVNNs with a unique capacity for handling multidimensional data representations. In recent scholarly discourse, a surge of interest has been directed towards the dynamics of QVNNs, as evidenced by a burgeoning body of literature [9,10,11,12,13,14]. Notably, the exploration of robust stability within the fractional-order realm of QVNNs has garnered significant attention, as demonstrated by the contributions of [9]. Additionally, innovative control methodologies, such as sampled-data approaches, have been applied to stabilize QVNNs, as explored in [10]. The investigation into the robust stability of these networks continues to evolve, with offering fresh insights into the fractional-order context [13]. Furthermore, the stability analysis of quaternion-valued memristive neural networks has been enriched by the application of Lagrangian mechanics, as discussed in [14], and so on [15,16,17,18,19]. These studies collectively contribute to a deeper understanding of the intricate dynamics that govern QVNNs, paving the way for advancements in network stability and synchronization.

    The recent research on quaternion-valued memristive neural networks (QVMNNs) has attempted to understand their periodic solutions, as highlighted in [15]. This work has been complemented by an examination of dissipativity in neutral-type memristor-based networks, which provides a fresh perspective on stability and energy dynamics, as discussed in [16]. Moreover, the exploration of finite-time stabilization through implicit function methods, as introduced in [17], has opened new avenues for the rapid and reliable control of these networks, underscoring the growing sophistication in the field of QVNNs.

    Synchronization, a fundamental and pivotal phenomenon in the realm of neural networks, plays a crucial role in managing and orchestrating the inherent chaotic dynamics that are often observed in natural systems. It serves as a powerful tool for harnessing order from chaos, offering a mechanism to regulate and predict the behavior of complex neural systems. Through synchronization, we can explore unknown dynamical systems from known ones. So far, the synchronization of neural networks includes quasi-uniform synchronization [20,21,22], anti-synchronization [23,24], finite-time synchronization [25,26,27], projection synchronization [28,29], exponential synchronization [30,31], and global Mittag-Leffler synchronization [32,33], and others [34,35].

    Among the spectrum of synchronization phenomena, projection synchronization emerges as an inclusive and versatile approach within the domain of neural network systems. This method, underpinned by proportional dynamics, facilitates accelerated communication pathways, a feature that sets it apart from other synchronization modalities [36,37,38]. In this work, we extend the discourse to the projective synchronization of quaternion-valued memristor-based neural networks, a topic that gains relevance amidst time-varying delays. Our exploration is anchored in the following pivotal contributions:

    (ⅰ) QVMNNs offer distinctive benefits over traditional real- and complex-valued neural networks, particularly in their adept handling of multi-dimensional data through low-dimensional constructs and enhanced computational efficiency.

    (ⅱ) The synchronization criteria established within this study are not only capable of orchestrating complete synchronization and anti-synchronization but also of embracing the general projection synchronization paradigm. These findings are posited as universally applicable and representative of the broader scope of neural network synchronization.

    (ⅲ) By designing a novel controller and using Lyapunov function and Young's inequality, some new synchronization conditions for projection synchronization of quaternion-valued memristor-based neural networks are obtained.

    The quaternion constitutes a realm of hypercomplex numbers, encapsulating a singular real component alongside three imaginary elements, thereby extending the numerical system beyond the conventional real and complex planes. The quaternion mQ can be described as

    m=mR+mIi+mJj+mKk,

    where mR,mI,mJ,mKR, the imaginary parts i,j,k obey the Hamilton rule:

    i2=j2=k2=1,ij=ji=k,jk=kj=i,ki=ik=j.

    Remark 1. Quaternions, a class of hypercomplex numbers, diverge from the algebraic properties of real and complex numbers. Unlike their commutative counterparts, the multiplication of any two quaternions, denoted as m,nQ, does not necessarily adhere to the associative law mn=nm. This departure signifies that several principles governing real and complex number systems are inapplicable to quaternions. Consequently, it mandates the innovation of novel methodologies and theoretical frameworks to effectively harness and comprehend the intricacies of quaternion arithmetic.

    For the quaternion m and n, where n=nR+nIi+nJj+nKk, then we can denote m+n as follows:

    m+n=mR+nR+(mI+nI)i+(mJ+nJ)j+(mK+nK)k.

    By Hamilton rule, we can denote mn as

    mn=(mRnRmInImJnJmKnK)+(mRnI+mInR+mJnKmKnJ)i+(mRnJ+mJnR+mKnJmInK)j+(mRnK+mKnR+mInJmJnI)k.

    The modulus of m is written as

    |m|=m¯m=(mR)2+(ml)2+(ml)2+(mK)2.

    Furthermore, for m=(m1,m2,,mn)T, m=ni=1|mi| denotes the norm of m.

    In this study, we delve into the dynamics of a memristor-driven neural network enriched with quaternion-valued parameters and subject to time-varying delays. The system's evolution is intricately captured by a set of differential equations that account for these temporal lags.

    ˙mq(t)=cq(t)mq(t)+ns=1bqs(mq(t))fs(ms(t))+ns=1dqs(mq(t))fs(ms(tτ(t)))+Iq,mq(s)=ψq(s),s[τ,0], (2.1)

    with each neuron's state vector represented by the components m(t)=(m1(t),m2(t),,mn(t))TQn, q=1,2,,n, and mq(t)Q corresponds to the individual neuron's state. The positive self-feedback coefficient is denoted by cq>0, and the quaternion-valued connection weights, stemming from memristor dynamics, are given by bqs(mq(t)) and dqs(mq(t)). The vector fs(ms(t))=(f1(m1(t)),f2(m2(t)),fn(mn(t))) encapsulates the activation functions that govern the neurons' firing patterns. The external stimuli to the network are captured by the input vector Iq=(I1,I2,,In)TQn. Additionally, the network's signal transmission is subject to delays characterized by τ(t), which are constrained to be non-negative and less than a maximum delay, satisfying 0τ(t)<τ. For the initial setup of the system (2.1), we select an initial condition that is continuously differentiable over the interval [τ,0], This is mathematically expressed as mq(s)=ψq(s)=(ψ1(s),ψ2(s),,ψn(s))C(1)([τ,0],Qn),τs0, laying down a well-defined starting point for the system's trajectory.

    Incorporating the framework of differential inclusion and set-valued mappings, coupled with our preceding discourse, the representation of system (2.1) is delineated as:

    ˙mq(t)cqmq(t)+ns=1co(bqs,b+qs)fs(ms(t))+ns=1co(dqs,d+qs)fs(ms(tτ(t)))+Iq, (2.2)

    where bqs=min{ˊbqs,ˊbqs},b+qs=max{ˊbqs,ˊbqs}, dqs=min{ˊdqs,ˊdqs},d+qs=max{ˊdqs,ˊdqs}. The essence of differential inclusion encapsulates the existence of a collection of differential equations where the terms bqs(t)co(bqs,b+qs), dqs(t)co(dqs,d+qs), such that

    ˙mq(t)=cqmq(t)+ns=1bqs(t)fs(ms(t))+ns=1dqs(t)fs(ms(tτ(t)))+Iq. (2.3)

    Assumption 1. Let m=mR+mIi+mJj+mKk, where mR,mI,mJ,mKR. fs(m) can be decomposed into its real and imaginary constituents, denoted by fs(m)=fRs(mR)+fIs(mI)i+fJs(mJ)j+fKs(mK)k.

    Assumption 2. For any mϵq(t),nϵq(t)Rn, we identify the existence of certain positive constants ϵ=R,I,J,K,q=1,2,,n, such that

    |fϵq(nϵq)fϵq(mϵq)|lϵq|nϵqmϵq|.

    Under Assumption 1, we divide the network (2.1) into one real part and three imaginary parts, respectively.

    ˙mRq(t)=cqmRq(t)+ns=1bRqs(t)fRs(mRs(t))ns=1bIqs(t)fIs(mIs(t))ns=1bJqs(t)fJs(mJs(t))ns=1bKqs(t)fKs(mKs(t))+ns=1dRqs(t)fRs(mRs(tτ(t)))ns=1dIqs(t)fIs(mIs(tτ(t)))ns=1dJqs(t)fJs(mJs(tτ(t)))ns=1dKqs(t)fKs(mKs(tτ(t)))+IRq, (2.4)
    ˙mIq(t)=cqmIq(t)+ns=1bRqs(t)fIs(mIs(t))+ns=1bIqs(t)fRs(mRs(t))+ns=1bJqs(t)fKs(mKs(t))ns=1bKqs(t)fJs(mJs(t))+ns=1dRqs(t)fIs(mIs(tτ(t)))+ns=1dIqs(t)fRs(mRs(tτ(t)))+ns=1dJqs(t)fKs(mKs(tτ(t)))ns=1dKqs(t)fJs(mJs(tτ(t)))+IIq, (2.5)
    ˙mJq(t)=cqmJq(t)+ns=1bRqs(t)fJs(mJs(t))+ns=1bJqs(t)fRs(mRs(t))+ns=1bKqs(t)fIs(mIs(t))ns=1bIqs(t)fKs(mKs(t))+ns=1dRqs(t)fJs(mJs(tτ(t)))+ns=1dJqs(t)fRs(mRs(tτ(t)))+ns=1dKqs(t)fIs(mIs(tτ(t)))ns=1dIqs(t)fKs(mKs(tτ(t)))+IJq, (2.6)
    ˙mKq(t)=cqmKq(t)+ns=1bRqs(t)fKs(mKs(t))+ns=1bKqs(t)fRs(mRs(t))+ns=1bIqs(t)fJs(mJs(t))ns=1bJqs(t)fIs(mIs(t))+ns=1dRqs(t)fKs(mKs(tτ(t)))+ns=1dKqs(t)fRs(mRs(tτ(t)))+ns=1dIqs(t)fJs(mJs(tτ(t)))ns=1dJqs(t)fIs(mIs(tτ(t)))+IKq. (2.7)

    Based on the characteristics of the memristor, the quaternion-valued memristive connective weights are defined as

    bqs()={ˊbqs=bR1qs+bI1qsi+bJ1qsj+bK1qsk||<Tq,ˊbqs=bR2qs+bI2qsi+bJ2qsj+bK2qsk||Tq,
    dqs()={ˊdqs=dR1qs+dI1qsi+dJ1qsj+dK1qsk||<Tq,ˊdqs=dR2qs+dI2qsi+dJ2qsj+dK2qsk||Tq,

    where the switching jump Tq>0.

    Consider the system (2.1) as the drive system; then, the response system is given as

    ˙nq(t)=cqnq(t)+ns=1bqs(nq(t))fs(ns(t))+ns=1dqs(nq(t))fs(ns(tτ(t)))+Iq+uq(t),nq(s)=ϕq(s),s[τ,0], (2.8)

    where q=1,2,,n; n(t)=(n1(t),n2(t),,nn(t))TQn. nq(t)Q stand for the state vector of the neuron. The initial condition of system (2.4) is chosen to be nq(s)=ϕq(s)=(ϕ1(s),ϕ2(s),,ϕn(s))C(1)([τ,0],Qn),τs0. uq(t) is the controller.

    Based on the theory of differential inclusion set-valued map, it yields from Eq (2.8) that

    ˙nq(t)cqnq(t)+ns=1co(bqs,b+qs)fs(ns(t))+ns=1co(dqs,d+qs)fs(ns(tτ(t)))+Iq+uq(t). (2.9)

    Differential inclusion means that there exist bqs(t)co(bqs,b+qs), dqs(t)co(dqs,d+qs) such that

    ˙nq(t)=cqnq(t)+ns=1bqs(t)fs(ns(t))+ns=1dqs(t)fs(ns(tτ(t)))+Iq+uq(t). (2.10)

    Similarly, we divide the network (2.10) into one real part and three imaginary parts, respectively

    ˙nRq(t)=cqnRq(t)+ns=1bRqs(t)fRs(nRs(t))ns=1bIqs(t)fIs(nIs(t))ns=1bJqs(t)fJs(nJs(t))ns=1bKqs(t)fKs(nKs(t))+ns=1dRqs(t)fRs(nRs(tτ(t)))ns=1dIqs(t)fIs(nIs(tτ(t)))ns=1dJqs(t)fJs(nJs(tτ(t)))ns=1dKqs(t)fKs(nKs(tτ(t)))+IRq+uRq(t), (2.11)
    ˙nIq(t)=cqnIq(t)+ns=1bRqs(t)fIs(nIs(t))+ns=1bIqs(t)fRs(nRs(t))+ns=1bJqs(t)fKs(nKs(t))ns=1bKqs(t)fJs(nJs(t))+ns=1dRqs(t)fIs(nIs(tτ(t)))+ns=1dIqsfRs(nRs(tτ(t)))+ns=1dJqs(t)fKs(nKs(tτ(t)))ns=1dKqs(t)fJs(nJs(tτ(t)))+IIq+uIq(t), (2.12)
    ˙nJq(t)=cqnJq(t)+ns=1bRqs(t)fJs(nJs(t))+ns=1bJqs(t)fRs(nRs(t))+ns=1bKqs(t)fIs(nIs(t))ns=1bIqs(t)fKs(nKs(t))+ns=1dRqs(t)fJs(nJs(tτ(t)))+ns=1dJqs(t)fRs(nRs(tτ(t)))+ns=1dKqs(t)fIs(nIs(tτ(t)))ns=1dIqs(t)fKs(nKs(tτ(t)))+IJq+uJq(t), (2.13)
    ˙nKq(t)=cqnKq(t)+ns=1bRqs(t)fKs(nKs(t))+ns=1bKqs(t)fRs(nRs(t))+ns=1bIqs(t)fJs(nJs(t))ns=1bJqs(t)fIs(nIs(t))+ns=1dRqs(t)fKs(nKs(tτ(t)))+ns=1dKqs(t)fRs(nRs(tτ(t)))+ns=1dIqs(t)fJs(nJs(tτ(t)))ns=1dJqs(t)fIs(nIs(tτ(t)))+IKq+uKq(t). (2.14)

    Let e(t)=(e1(t),e2(t),,en(t)) be the synchronization error. The synchronization error between the drive system (2.2) and the response system (2.9) is defined as eq(t)=nq(t)βmq(t), expressed as one real part and three imaginary parts

    {eRq(t)=nRq(t)βmRq(t),eIq(t)=nIq(t)βmIq(t),eJq(t)=nJq(t)βmJq(t),eKq(t)=nKq(t)βmKq(t), (2.15)

    with the initial value ϕq(s)ψq(s), τs0.

    In order to synchronize the drive system and the response system, we choose the following controller

    uRq(t)=ki(nRq(t)βmRq(t))+ns=1(bRqs(t)βfRs(mRs(t))bRqs(t)fRs(βmRs(t)))+ns=1(bIqs(t)fIs(βmIs(t))bIqs(t)βfIs(mIs(t)))+ns=1(bJqs(t)fJs(βmJs(t))bJqs(t)βfJs(mJs(t)))+ns=1(bKqs(t)fKs(βmKs(t))bKqs(t)βfKs(mKs(t)))+ns=1(dRqs(t)βfRs(mRs(tτ(t)))dRqs(t)fRs(βmRs(tτ(t))))+ns=1(dIqs(t)fIs(βmIs(tτ(t)))dIqs(t)βfIs(mIs(tτ(t))))+ns=1(dJqs(t)fJs(βmJs(tτ(t)))dJqs(t)βfJs(mJs(tτ(t))))+ns=1(dKqs(t)fKs(βmKs(tτ(t)))dKqs(t)βfKs(mKs(tτ(t))))+(1β)IRq,uIq(t)=ki(nIq(t)βmIq(t))+ns=1(bRqs(t)β(fIs(mIs(t))bRqs(t)fIs(βmIs(t)))+ns=1(bIqs(t)β(fRs(mRs(t))bIqs(t)fRs(βmRs(t)))+ns=1(bKqs(t)(fJs(βmJs(t))bKqs(t)βfJs(mJs(t)))+ns=1(bJqs(t)β(fKs(mKs(t))bJqs(t)fKs(βmKs(t)))+ns=1(dRqs(t)β(fIs(mIs(tτ(t)))dRqs(t)fIs(βmIs(tτ(t))))+ns=1(dIqs(t)β(fRs(mRs(tτ(t)))dIqs(t)fRs(βmRs(tτ(t))))+ns=1(dKqs(t)(fJs(βmJs(tτ(t)))dKqs(t)βfJs(mJs(tτ(t))))+ns=1(dJqs(t)β(fKs(mKs(tτ(t)))dJqs(t)fKs(βmKs(tτ(t))))+(1β)IIq,uJq(t)=ki(nJq(t)βmJq(t))+ns=1(bRqs(t)β(fJs(mJs(t))bRqs(t)fJs(βmJs(t)))+ns=1(bIqs(t)(fKs(βmKs(t))bIqs(t)βfKs(mKs(t)))+ns=1(bKqs(t)β(fIs(mIs(t))bKqs(t)fIs(βmIs(t)))+ns=1(bJqs(t)β(fRs(mRs(t))bJqs(t)fRs(βmRs(t)))+ns=1(dRqs(t)β(fJs(mJs(tτ(t)))dRqs(t)fJs(βmJs(tτ(t))))+ns=1(dIqs(t)(fKs(βmKs(tτ(t)))dIqs(t)βfKs(mKs(tτ(t))))+ns=1(dKqs(t)β(fIs(mIs(tτ(t)))dKqs(t)fIs(βmIs(tτ(t))))+ns=1(dJqs(t)β(fRs(mRs(tτ(t)))dJqs(t)fRs(βmRs(tτ(t))))+(1β)IJq,uKq(t)=ki(nKq(t)βmKq(t))+ns=1(bKqs(t)β(fRs(mRs(t))bKqs(t)fRs(βmRs(t)))+ns=1(bRqs(t)β(fKs(mKs(t))bRqs(t)fKs(βmKs(t)))+ns=1(bIqs(t)β(fJs(mJs(t))bIqs(t)fJs(βmJs(t)))+ns=1(bJqs(t)(fIs(βmIs(t))bJqs(t)βfIs(mIs(t)))+ns=1(dKqs(t)β(fRs(mRs(tτ(t)))dKqs(t)fRs(βmRs(tτ(t))))+ns=1(dRqs(t)β(fKs(mKs(tτ(t)))dRqs(t)fKs(βmKs(tτ(t))))+ns=1(dIqs(t)β(fJs(mJs(tτ(t)))dIqs(t)fJs(βmJs(tτ(t))))+ns=1(dJqs(t)(fIs(βmIs(tτ(t)))dJqs(t)βfIs(mIs(tτ(t))))+(1β)IKq. (2.16)

    Let eq(t)=eRq(t)+eIq(t)i+eJq(t)j+eKq(t)k; then, according to the controller (2.16), the error system (2.15) can be separated into four real parts as below

    ˙eRq(t)=cqeRq(t)kqeRq(t)ns=1bRqs(t)˜fRs(eRs(t))ns=1bIqs(t)˜fIs(eIs(t))ns=1bJqs(t)˜fJs(eJs(t))ns=1bKqs(t)˜fKs(eKs(t))+ns=1dRqs(t)˜fRs(eRs(tτ(t)))ns=1dIqs(t)˜fIs(eIs(tτ(t)))ns=1dJqs(t)˜fJs(eJs(tτ(t)))ns=1dKqs(t)˜fKs(eKs(tτ(t))), (2.17)
    ˙eIq(t)=cqeIq(t)kqeIq(t)+ns=1bRqs(t)˜fIs(eIs(t))+ns=1bIqs(t)˜fRs(eRs(t))ns=1bKqs(t)˜fJs(eJs(t))+ns=1bJqs(t)˜fKs(eKs(t))+ns=1dRqs(t)˜fIs(eIs(tτ(t)))+ns=1dIqs(t)˜fRs(eRs(tτ(t)))ns=1dKqs(t)˜fJs(eJs(tτ(t)))+ns=1dJqs(t)˜fKs(eKs(tτ(t))), (2.18)
    ˙eJq(t)=cqeJq(t)kqeJq(t)+ns=1bRqs(t)˜fJs(eJs(t))ns=1bIqs(t)˜fKs(eKs(t))+ns=1bKqs(t)˜fIs(eIs(t))+ns=1bJqs(t)˜fRs(eRs(t))+ns=1dRqs(t)˜fJs(eJs(tτ(t)))ns=1dIqs(t)˜fKs(eKs(tτ(t)))+ns=1dKqs(t)˜fIs(eIs(tτ(t)))+ns=1dJqs(t)˜fRs(eRs(tτ(t))), (2.19)
    ˙eKq(t)=cqeKq(t)kqeKq(t)+ns=1bKqs(t)˜fRs(eRs(t))+ns=1bRqs(t)˜fKs(eKs(t))+ns=1bIqs(t)˜fJs(eJs(t))ns=1bJqs(t)˜fIs(eIs(t))+ns=1dKqs(t)˜fRs(eRs(tτ(t)))+ns=1dRqs(t)˜fKs(eKs(tτ(t)))+ns=1dIqs(t)˜fJs(eJs(tτ(t)))ns=1dJqs(t)˜fIs(eIs(tτ(t))), (2.20)

    where ˜fϵs(eϵs(t))=fs(nϵs(t))fs(βmϵs(t)), ˜fϵs(eϵs(tτ(t)))=fs(nϵs(tτ(t)))fs(βmϵs(tτ(t))), ϵ=R,I,J,K.

    The following notations will be used:

    |bRqs|=supt0|bRqs(t)|, |bIqs|=supt0|bIqs(t)|, |bJqs|=supt0|bJqs(t)|, |bKqs|=supt0|bKqs(t)|,

    |dRqs|=supt0|dRqs(t)|, |dIqs|=supt0|dIqs(t)|, |dJqs|=supt0|dJqs(t)|, |dKqs|=supt0|dKqs(t)|.

    Before deriving the result, the definitions and lemmas are given to facilitate the subsequent derivation.

    Definition 2.1. [38] The driving network (2.1) and the response network (2.8) are said to achieve projection synchronization if

    limtnq(t)βmq(t)=0,q=1,2,,n,

    where βR is a nonzero constant.

    Remark 2. When the projection factor β=1, complete synchronization is achieved. When the projection factor β=1, anti-synchronization is obtained.

    Lemma 2.1. [28] Let m>0, n>0, r>1 and 1r+1s=1. Then, the following inequality holds

    mn1rmr+1sns.

    Lemma 2.2. [39] Suppose that function V(t) is non-negative when t(τ,) and satisfies the following inequality

    ˙V(t)aV(t)bV(tτ(t)),t0,

    where a and b are positive constants with a>b. Then,

    V(t)supτs0V(s)ert,

    where r is the unique positive solution of the following equation

    aberτ=r.

    Theorem 3.1. Under the controller (2.16), if Assumptions 1 and 2 hold, and

    λ>ζ>0. (3.1)

    Then the projection synchronization of quaternion-valued memristor-based neural networks (2.1) and (2.8) is obtained, where λ=min{λ1,λ2,λ3,λ4}, ζ=max{ζ1,ζ2,ζ3,ζ4} and

    λ1=min1qn{rcq+rkqns=1(lRs|bRqs|+lRs|bIqs|+lRs|bJqs|+lRs|bKqs|+(r1)lRs|bRqs|+(r1)lIs|bIqs|+(r1)lJs|bJqs|+(r1)lKs|bKqs|+(r1)lRs|dRqs|+(r1)lIs|dIqs|+(r1)lJs|dJqs|+(r1)lKs|dKqs|)},λ2=min1qn{rcq+rkqns=1(lIs|bRqs|+lIs|bIqs|+lIs|bJqs|+lIs|bKqs|+(r1)lIs|bRqs|+(r1)lRs|bIqs|+(r1)lKs|bJqs|+(r1)lJs|bKqs|+(r1)lIs|dRqs|+(r1)lRs|dIqs|+(r1)lKs|dJqs|+(r1)lJs|dKqs|)},λ3=min1qn{rcq+rkqns=1(lJs|bRqs|+lJs|bIqs|+lJs|bJqs|+lJs|bKqs|+(r1)lJs|bRqs|+(r1)lRs|bJqs|+(r1)lKs|bIqs|+(r1)lIs|bKqs|+(r1)lJs|dRqs|+(r1)lRs|dJqs|+(r1)lKs|dIqs|+(r1)lIs|dKqs|)},λ4=min1qn{rcq+rkqns=1(lKs|bRqs|+lKs|bIqs|+lKs|bJqs|+lKs|bKqs|+(r1)lKs|bRqs|+(r1)lRs|bKqs|+(r1)lJs|bIqs|+(r1)lIs|bJqs|+(r1)lKs|dRqs+(r1)lRs|dKqs|+(r1)lJs|dIqs|+(r1)lIs|dJqs|)}.
    ζ1=max1qnns=1(lRs|dRqs|+lRs|dIqs|+lRs|dJqs|+lRs|dKqs|),ζ2=max1qnns=1(lIs|dRqs|+lIs|dIqs|+lIs|dJqs|+lIs|dKqs|),ζ3=max1qnns=1(lJs|dRqs|+lJs|dIqs|+lJs|dJqs|+lJs|dKqs|),ζ1=max1qnns=1(lKs|dRqs|+lKs|dIqs|+lKs|dJqs|+lKs|dKqs|).

    Proof. Considering the following Lyapunov function

    V(e(t))=1rnq=1|eRq(t)|r+1rnq=1|eIq(t)|r+1rnq=1|eJq(t)|r+1rnq=1|eKq(t)|r. (3.2)

    Computing the time derivative of V(e(t)) along the trajectory (2.17)–(2.20), from Lemma 2.1, we have

    ˙V(e(t)nq=1|eRq(t)|r1[cq|eRq(t)|+ns=1|bRqs(t)||˜fRs(eRs(t))|+ns=1|bIqs(t)||˜fIs(eIs(t))|+ns=1|bJqs(t)||˜fJs(eJs(t))|+ns=1|bKqs(t)||˜fKs(eKs(t))|+ns=1|dRqs(t)||˜fRs(eRs(tτ(t)))|+ns=1|dIqs(t)||˜fIs(eIs(tτ(t)))|+ns=1|dJqs(t)||˜fJs(eJs(tτ(t)))|+ns=1|dKqs(t)||˜fKs(eKs(tτ(t)))|kq|eRq(t)|]+nq=1|eIq(t)|r1[cq|eIq(t)|+ns=1|bRqs(t)||˜fIs(eIs(t))|+ns=1|bIqs(t)||˜fRs(eRs(t))|+ns=1|bKqs(t)||˜fJs(eJs(t))|+ns=1|bJqs(t)||˜fKs(eKs(t))|+ns=1|dRqs(t)||˜fIs(eIs(tτ(t)))|+ns=1|dIqs(t)||˜fRs(eRs(tτ(t)))|+ns=1|dKqs(t)||˜fJs(eJs(tτ(t)))|+ns=1|dJqs(t)||˜fKs(eKs(tτ(t)))|kq|eIq(t)|]+nq=1|eJq(t)|r1[cq|eJq(t)|+ns=1|bRqs(t)||˜fJs(eJs(t))|+ns=1|bIqs(t)||˜fKs(eKs(t))|+ns=1|bKqs(t)||˜fIs(eIs(t))|+ns=1|bJqs(t)||˜fRs(eRs(t))|+ns=1|dRqs(t)||˜fJs(eJs(tτ(t)))|+ns=1|dIqs(t)||˜fKs(eKs(tτ(t)))|+ns=1|dKqs(t)||˜fIs(eIs(tτ(t)))|+ns=1|dJqs(t)||˜fRs(eRs(tτ(t)))|kq|eJq(t)|]+nq=1|eKq(t)|r1[cq|eKq(t)|+ns=1|bRqs(t)||˜fKs(eKs(t))|+ns=1|bKqs(t)||˜fRs(eRs(t))|+ns=1|bIqs(t)||˜fJs(eJs(t))|+ns=1|bJqs(t)||˜fIs(eIs(t))|+ns=1|dRqs(t)||˜fKs(eKs(tτ(t)))|+ns=1|dIqs(t)||˜fJs(eJs(tτ(t)))|+ns=1|dKqs(t)||˜fRs(eRs(tτ(t)))|+ns=1|dJqs(t)||˜fIs(eIs(tτ(t)))|kq|eKq(t)|]. (3.3)

    According to Assumption 2 and the notations defined previously, we can get

    ˙V(e(t)nq=1|eRq(t)|r1[cq|eRq(t)|+ns=1|bRqs|lRs|eRs(t)|+ns=1|bIqs|lIs|eIs(t)|+ns=1|bJqs|lJs|eJs(t)|+ns=1|bKqs|lKs|eKs(t)|+ns=1|dRqs|lRs|eRs(tτ(t))|+ns=1|dIqs|lIs|eIs(tτ(t))|+ns=1|dJqs|lJs|eJs(tτ(t))|+ns=1|dKqs|lKs|eKs(tτ(t))|kq|eRq(t)|]+nq=1|eIq(t)|r1[cq|eIq(t)|+ns=1|bRqs|lIs|eIs(t)|+ns=1|bIqs|lRs|eRs(t)|+ns=1|bKqs|lJs|eJs(t)|+ns=1|bJqs|lKs|eKs(t)|+ns=1|dRqs|lIs|eIs(tτ(t))|+ns=1|dIqs|lRs|eRs(tτ(t))|+ns=1|dJqs|lKs|eKs(tτ(t))|+ns=1|dKqs|lJs|eJs(tτ(t))|kq|eIq(t)|]+nq=1|eJq(t)|r1[cq|eJq(t)|+ns=1|bRqs|lJs|eJs(t)|+ns=1|bIqs|lKs|eKs(t)|+ns=1|bKqs|lIs|eIs(t)|+ns=1|bJqs|lRs|eRs(t)|+ns=1|dRqs|lJs|eJs(tτ(t))|+ns=1|dIqs|lKs|eKs(tτ(t))|+ns=1|dKqs|lIs|eIs(tτ(t))|+ns=1|dJqs|lRs|eRs(tτ(t))|kq|eJq(t)|]+nq=1|eKq(t)|r1[cq|eKq(t)|+ns=1|bRqs|lKs|eKs(t)|+ns=1|bKqs|lRs|eRs(t)|+ns=1|bIqs|lJs|eJs(t)|+ns=1|bJqs|lIs|eIs(t)|+ns=1|dRqs|lKs|eKs(tτ(t))|+ns=1|dJqs|lIs|eIs(tτ(t))|kq|eKq(t)|+ns=1|dKqs|lRs|eRs(tτ(t))|+ns=1|dIqs|lJs|eJs(tτ(t))|]. (3.4)

    Then, according to Lemma 2.2, we have

    |eRq(t)|r1|eRs(t)|1r|eRs(t)|r+r1r|eRq(t)|r,  |eRq(t)|r1|eIs(t)|1r|eIs(t)|r+r1r|eRq(t)|r,|eRq(t)|r1|eJs(t)|1r|eJs(t)|r+r1r|eRq(t)|r,  |eRq(t)|r1|eKs(t)|1r|eKs(t)|r+r1r|eRq(t)|r,|eRq(t)|r1|eRs(tτ(t))|1r|eRs(tτ(t))|r+r1r|eRq(t)|r,|eRq(t)|r1|eIs(tτ(t))|1r|eIs(tτ(t))|r+r1r|eRq(t)|r,|eRq(t)|r1|eJs(tτ(t))|1r|eJs(tτ(t))|r+r1r|eRq(t)|r,|eRq(t)|r1|eKs(tτ(t))|1r|eKs(tτ(t))|r+r1r|eRq(t)|r,|eIq(t)|r1|eIs(t)|1r|eIs(t)|r+r1r|eIq(t)|r,|eIq(t)|r1|eRs(t)|1r|eRs(t)|r+r1r|eIq(t)|r,  |eIq(t)|r1|eJs(t)|1r|eJs(t)|r+r1r|eIq(t)|r,|eIq(t)|r1|eKs(t)|1r|eKs(t)|r+r1r|eIq(t)|r,  |eIq(t)|r1|eRs(tτ(t))|1r|eRs(tτ(t))|r+r1r|eIq(t)|r,|eIq(t)|r1|eIs(tτ(t))|1r|eIs(tτ(t))|r+r1r|eIq(t)|r,|eIq(t)|r1|eJs(tτ(t))|1r|eJs(tτ(t))|r+r1r|eIq(t)|r,|eIq(t)|r1|eKs(tτ(t))|1r|eKs(tτ(t))|r+r1r|eIq(t)|r,|eJq(t)|r1|eIs(t)|1r|eIs(t)|r+r1r|eJq(t)|r,  |eJq(t)|r1|eRs(t)|1r|eRs(t)|r+r1r|eJq(t)|r,|eJq(t)|r1|eJs(t)|1r|eJs(t)|r+r1r|eJq(t)|r,  |eJq(t)|r1|eKs(t)|1r|eKs(t)|r+r1r|eJq(t)|r,|eJq(t)|r1|eRs(tτ(t))|1r|eRs(tτ(t))|r+r1r|eJq(t)|r,|eJq(t)|r1|eIs(tτ(t))|1r|eIs(tτ(t))|r+r1r|eJq(t)|r,|eJq(t)|r1|eJs(tτ(t))|1r|eJs(tτ(t))|r+r1r|eJq(t)|r,|eJq(t)|r1|eKs(tτ(t))|1r|eKs(tτ(t))|r+r1r|eJq(t)|r,|eKq(t)|r1|eIs(t)|1r|eIs(t)|r+r1r|eKq(t)|r,  |eKq(t)|r1|eRs(t)|1r|eRs(t)|r+r1r|eKq(t)|r,|eKq(t)|r1|eJs(t)|1r|eJs(t)|r+r1r|eKq(t)|r,  |eKq(t)|r1|eKs(t)|1r|eKs(t)|r+r1r|eKq(t)|r,|eKq(t)|r1|eRs(tτ(t))|1r|eRs(tτ(t))|r+r1r|eKq(t)|r,|eKq(t)|r1|eIs(tτ(t))|1r|eIs(tτ(t))|r+r1r|eKq(t)|r,|eKq(t)|r1|eJs(tτ(t))|1r|eJs(tτ(t))|r+r1r|eKq(t)|r,|eKq(t)|r1|eKs(tτ(t))|1r|eKs(tτ(t))|r+r1r|eKq(t)|r.

    Therefore, one has

    ˙V(e(t)nq=1(cq+kq)|eRq(t)|rnq=1(cq+kq)|eIq(t)|rnq=1(cq+kq)|eJq(t)|rnq=1(cq+kq)|eKq(t)|r+nq=1ns=1|bRqs|lRs(1r|eRs(t)|r+r1r|eRq(t)|r)+nq=1ns=1|bIqs|lIs(1r|eIs(t)|r+r1r|eRq(t)|r)+nq=1ns=1|bJqs|lJs(1r|eJs(t)|r+r1r|eRq(t)|r)+nq=1ns=1|bKqs|lKs(1r|eKs(t)|r+r1r|eRq(t)|r)+nq=1ns=1|dRqs|lRs(1r|eRs(tτ(t))|r+r1r|eRq(t)|r)+nq=1ns=1|dIqs|lIs(1r|eIs(tτ(t))|r+r1r|eRq(t)|r)+nq=1ns=1|dJqs|lJs(1r|eJs(tτ(t))|r+r1r|eRq(t)|r)+nq=1ns=1|dKqs|lKs(1r|eKs(tτ(t))|r+r1r|eRq(t)|r)+nq=1ns=1|bRqs|lIs(1r|eIs(t)|r+r1r|eIq(t)|r)+nq=1ns=1|bIqs|lRs(1r|eRs(t)|r+r1r|eIq(t)|r)+nq=1ns=1|bJqs|lKs(1r|eKs(t)|r+r1r|eIq(t)|r)+nq=1ns=1|bKqs|lJs(1r|eJs(t)|r+r1r|eIq(t)|r)+nq=1ns=1|dRqs|lIs(1r|eIs(tτ(t))|r+r1r|eIq(t)|r)+nq=1ns=1|dIqs|lRs(1r|eRs(tτ(t))|r+r1r|eIq(t)|r)+nq=1ns=1|dKqs|lJs(1r|eJs(tτ(t))|r+r1r|eIq(t)|r)+nq=1ns=1|dJqs|lKs(1r|eKs(tτ(t))|r+r1r|eIq(t)|r)+nq=1ns=1|bRqs|lJs(1r|eJs(t)|r+r1r|eJq(t)|r)+nq=1ns=1|bIqs|lKs(1r|eKs(t)|r+r1r|eJq(t)|r)+nq=1ns=1|bKqs|lIs(1r|eIs(t)|r+r1r|eJq(t)|r)+nq=1ns=1|bJqs|lRs(1r|eRs(t)|r+r1r|eJq(t)|r)+nq=1ns=1|dRqs|lJs(1r|eJs(tτ(t))|r+r1r|eJq(t)|r)+nq=1ns=1|dIqs|lKs(1r|eKs(tτ(t))|r+r1r|eJq(t)|r)+nq=1ns=1|dKqs|lIs(1r|eIs(tτ(t))|r+r1r|eJq(t)|r)+nq=1ns=1|dJqs|lRs(1r|eRs(tτ(t))|r+r1r|eJq(t)|r)+nq=1ns=1|bRqs|lKs(1r|eKs(t)|r+r1r|eKq(t)|r)+nq=1ns=1|bKqs|lRs(1r|eRs(t)|r+r1r|eKq(t)|r)+nq=1ns=1|bIqs|lJs(1r|eJs(t)|r+r1r|eKq(t)|r)+nq=1ns=1|bJqs|lIs(1r|eIs(t)|r+r1r|eKq(t)|r)+nq=1ns=1|dRqs|lKs(1r|eKs(tτ(t))|r+r1r|eKq(t)|r)+nq=1ns=1|dKqs|lRs(1r|eRs(tτ(t))|r+r1r|eKq(t)|r)+nq=1ns=1|dIqs|lJs(1r|eJs(tτ(t))|r+r1r|eKq(t)|r)+nq=1ns=1|dJqs|lIs(1r|eIs(tτ(t))|r+r1r|eKq(t)|r).

    Then, we have

    ˙V(e(t))nq=1r(cqkq+1rns=1lRs|bRqs|+1rns=1lRs|bIqs|+r1rns=1lIs|bIqs|+r1rns=1lJs|bJqs|+1rns=1lRs|bJqs|+1rns=1lRs|bKqs|+r1rns=1lRs|bRqs|+r1rns=1lKs|bKqs|+r1rns=1lRs|dRqs|+r1rns=1lIs|dIqs|+r1rns=1lJs|dJqs|+r1rns=1lKs|dKqs|)1r|eRq(t)|r+nq=1r(cqkq+1rns=1lIs|bRqs|+1rns=1lIs|bIqs|+1rns=1lIs|bJqs|+1rns=1lIs|bKqs|+r1rns=1lIs|bRqs|+r1rns=1lRs|bIqs|+r1rns=1lKs|bJqs|+r1rns=1lJs|bKqs|+r1rns=1lIs|dRqs|+r1rns=1lRs|dIqs|+r1rns=1lKs|dJqs|+r1rns=1lJs|dKqs|)1r|eIq(t)|r+nq=1r(cqkq+1rns=1lJs|bRqs|+1rns=1lJs|bIqs|+1rns=1lJs|bJqs|+1rns=1lJs|bKqs|+r1rns=1lJs|bRqs|+r1rns=1lRs|bJqs|+r1rns=1lKs|bIqs|+r1rns=1lIs|bKqs|+r1rns=1lJs|dRqs|+r1rns=1lRs|dJqs|+r1rns=1lKs|dIqs|+r1rns=1lIs|dKqs|)1r|eJq(t)|r+nq=1r(cqkq+1rns=1lKs|bRqs|+1rns=1lKs|bIqs|+1rns=1lKs|bJqs|+1rns=1lKs|bKqs|+r1rns=1lKs|bRqs|+r1rns=1lRs|bKqs|+r1rns=1lJs|bIqs|+r1rns=1lIs|bJqs|+r1rns=1lKs|dRqs|+r1rns=1lRs|dKqs|+r1rns=1lJs|dIqs|+r1rns=1lIs|dJqs|)1r|eKq(t)|r+nq=1ns=1r(1rlRs|dRqs|+1rlRs|dIqs|+1rlRs|dJqs|+1rlRs|dKqs|)1r|eRq(tτ(t))|r+nq=1ns=1r(1rlIs|dRqs|+1rlIs|dIqs|+1rlIs|dJqs|+1rlIs|dKqs|)1r|eIq(tτ(t))|r+nq=1ns=1r(1rlJs|dRqs|+1rlJs|dIqs|+1rlJs|dJqs|+1rlJs|dKqs|)1r|eJq(tτ(t))|r+nq=1ns=1r(1rlKs|dRqs|+1rlKs|dIqs|+1rlKs|dJqs|+1rlKs|dKqs|)1r|eKq(tτ(t))|r. (3.5)

    We suppose that

    λ1=min1qn{rcq+rkqns=1(lRs|bRqs|+lRs|bIqs|+lRs|bJqs|+lRs|bKqs|+(r1)lRs|bRqs|+(r1)lIs|bIqs|+(r1)lJs|bJqs|+(r1)lKs|bKqs|+(r1)lRs|dRqs|+(r1)lIs|dIqs|+(r1)lJs|dJqs|+(r1)lKs|dKqs|)},λ2=min1qn{rcq+rkqns=1(lIs|bRqs|+lIs|bIqs|+lIs|bJqs|+lIs|bKqs|+(r1)lIs|bRqs|+(r1)lRs|bIqs|+(r1)lKs|bJqs|+(r1)lJs|bKqs|+(r1)lIs|dRqs|+(r1)lRs|dIqs|+(r1)lKs|dJqs|+(r1)lJs|dKqs|)},λ3=min1qn{rcq+rkqns=1(lJs|bRqs|+lJs|bIqs|+lJs|bJqs|+lJs|bKqs|+(r1)lJs|bRqs|+(r1)lRs|bJqs|+(r1)lKs|bIqs|+(r1)lIs|bKqs|+(r1)lJs|dRqs|+(r1)lRs|dJqs|+(r1)lKs|dIqs|+(r1)lIs|dKqs|)},λ4=min1qn{rcq+rkqns=1(lKs|bRqs|+lKs|bIqs|+lKs|bJqs|+lKs|bKqs|+(r1)lKs|bRqs|+(r1)lRs|bKqs|+(r1)lJs|bIqs|+(r1)lIs|bJqs|+(r1)lKs|dRqs|+(r1)lRs|dKqs|+(r1)lJs|dIqs|+(r1)lIs|dJqs|)},
    ζ1=max1qnns=1(lRs|dRqs|+lRs|dIqs|+lRs|dJqs|+lRs|dKqs|),  ζ2=max1qnns=1(lIs|dRqs|+lIs|dIqs|+lIs|dJqs|+lIs|dKqs|),ζ3=max1qnns=1(lJs|dRqs|+lJs|dIqs|+lJs|dJqs|+lJs|dKqs|),  ζ1=max1qnns=1(lKs|dRqs|+lKs|dIqs|+lKs|dJqs|+lKs|dKqs|).

    Then,

    ˙V(e(t)λ11r|eRq(t)|rλ21r|eIq(t)|rλ31r|eJq(t)|rλ41r|eKq(t)|r+ζ11r|eRq(tτ(t))|r+ζ21r|eIq(tτ(t))|r+ζ31r|eJq(tτ(t))|r+ζ41r|eKq(tτ(t))|rλV(e(t))+ζV(e(tτ(t))), (3.6)

    where λ=min{λ1,λ2,λ3,λ4}, ζ=max{ζ1,ζ2,ζ3,ζ4}.

    Therefore, according to Lemma 2.4 and Theorem 3.1, the projection synchronization between drive system (2.1) and response system (2.8) can be achieved.

    Remark 3. In recent years, the research of real-valued fractional-order neural networks and complex-valued fractional-order neural networks in projection synchronization[28,29,36,37] has also achieved excellent results. Compared with previous studies, QVNN is superior in handling multi-dimensional problems. Therefore, our results are more general.

    Considering the two dimensional quaternion-valued memristor-based neural networks model of Eq (2.1) with the memristive connection weights are

    b11(m1(t))={2.31.6i+2.3j1.5k,|m1(t)|<1,2.02.7i+2.0j0.7k,|m1(t)|1,
    b12(m1(t))={0.50.4i0.5j0.7k,|m1(t)|<1,0.10.9i0.1j0.3k,|m1(t)|1,
    b21(m2(t))={1.1+0.7i+1.0j+0.6k,|m2(t)|<1,1.60.3i+1.5j0.4k,|m2(t)|1,
    b22(m2(t))={1.20.3i0.7j0.3k,|m2(t)|<1,0.80.1i1.3j0.2k,|m2(t)|1,
    d11(m1(t))={1.4+3.1i1.4j+3.0k,|m1(t)|<1,1.5+2.6i1.5j+2.3k,|m1(t)|1,
    d12(m1(t))={0.51.5i0.5j1.6k,|m1(t)|<1,0.10.9i0.1j0.6k,|m1(t)|1,
    d21(m2(t))={0.80.2i0.6j0.1k,|m2(t)|<1,1.21.1i1.3j1.3k,|m2(t)|1,
    d22(m2(t))={0.50.8i+0.4j0.7k,|m2(t)|<1,1.30.5i+1.2j0.4k,|m2(t)|1.

    The response system of (2.8) is

    ˙n1(t)=c1n1(t)+2s=1b1s(n1(t))fs(ns(t))+2s=1d1s(n1(t))fs(ns(tτ(t)))+I1+u1(t),˙n2(t)=c2n2(t)+2s=1b2s(n2(t))fs(ns(t))+2s=1d2s(n2(t))fs(ns(tτ(t)))+I2+u2(t).

    Take the time delay τ(t)=0.750.25cos(t) such that τ=1 and the activation function f(mq(t))=0.23(|mRq(t)+1||mRq(t)1|)+0.23(|mIq(t)+1||mIq(t)1|)i+0.23(|mJq(t)+1||mJq(t)1|)j+0.23(|mKq(t)+1||mKq(t)1|)k (q=1,2), and external inputs I1 = I2 = 0. According to the Assumption 2, we can get lεq=0.46, where q=1,2. The control gain parameter is taken as k1=k2=40 and the connection weights are taken as c1=c2=1. Then, the trajectories of error systems with controllers and without controllers are obtained.

    Besides, we take r=2. According to the above parameters, we can directly calculate the direct result of the condition in Theorem 3.1: λ1=λ2=λ3=λ4=53.526 and ζ1=ζ2=ζ3=ζ4=10.166. Then, we can obtain λ=53.526>ζ=10.166.

    The above result satisfies the condition in Theorem 3.1, so systems (2.1) and (2.8) achieve projective synchronization. Figures 1 and 2 show the synchronization error curves eq (q=1,2) for β=0.8 without the controller. Figures 3 and 4 show the synchronization error curves eq (q=1,2) for β=0.8 with the controller (2.16). Figures 5 and 6 show the synchronization error curves eq (q=1,2) for β=1 with the controller (2.16). Figures 7 and 8 show the synchronization error curves eq (q=1,2) for β=2 with the controller (2.16). From the above simulation results, we know that the derive system (2.1) and the response system (2.8) are synchronized, which verifies the effectiveness of Theorem 3.1.

    Figure 1.  The errors e1 with β=0.8 without the controller.
    Figure 2.  The errors e2 with β=0.8 without the controller.
    Figure 3.  The errors e1 with β=0.8 under the controller.
    Figure 4.  The errors e2 with β=0.8 under the controller.
    Figure 5.  The errors e1 with β=1 under the controller.
    Figure 6.  The errors e2 with β=1 under the controller.
    Figure 7.  The errors e1 with β=2 under the controller.
    Figure 8.  The errors e2 with β=2 under the controller.

    This study delves into the dynamics of projective synchronization within the realm of quaternion-valued memristor-based neural networks, which are subject to time-varying delays. Employing the theoretical underpinnings of set-valued mappings and differential inclusion, we formulate a hybrid control approach to dissect the projection synchronization dilemma of the network. By harnessing the stability assurances of a Lyapunov function and the quantitative bounds provided by Young's inequality, we formulate a novel criterion for synchronization. This leads to the achievement of projective synchronization in the context of the aforementioned neural networks. The efficacy and practicality of our proposed strategy are substantiated through rigorous numerical simulations.

    Jun Guo, Yanchao Shi and Yanzhao Cheng wrote the main manuscript text. All authors reviewed the manuscript.

    Supported by the National Natural Science Foundation of China under Grant 61703354; Key Laboratory of Numerical Simulation of Sichuan Provincial Universities KLNS-2023SZFZ002; the Scientific Research Foundation of Chengdu University of Information Technology KYQN202324.

    The authors declare that they have not used Artificial Intelligence (AI) tools in the creation of this article.

    The authors declare there is no conflict of interest.



    [1] L. O. Chua, Memristor-the missing circuit element, IEEE Trans. Circuit Theory, 18 (1971), 507–519. https://doi.org/10.1109/TCT.1971.1083337 doi: 10.1109/TCT.1971.1083337
    [2] H. Bao, Y. Zhang, W. Liu, B. Bao, Memristor synapse-coupled memristive neuron network: Synchronization transition and occurrence of chimera, Nonlinear Dyn., 100 (2020), 937–950. https://doi.org/10.1007/s11071-020-05529-2 doi: 10.1007/s11071-020-05529-2
    [3] F. Wei, G. Chen, W. Wang, Finite-time synchronization of memristor neural networks via interval matrix method, Neural Networks, 127 (2020), 7–18. https://doi.org/10.1016/j.neunet.2020.04.003 doi: 10.1016/j.neunet.2020.04.003
    [4] D. Ding, Z. You, Y. Hu, Z. Yang, L. Ding, Finite-time synchronization for fractional-order memristor-based neural networks with discontinuous activations and multiple delays, Mod. Phys. Lett. B, 34 (2020), 2050162. https://doi.org/10.1142/S0217984920501626 doi: 10.1142/S0217984920501626
    [5] U. Bhatti, Z. Yu, L. Yuan, Z. Zeeshan, M. Bhatti, M. Anum, et al., Geometric algebra applications in geospatial artificial intelligence and remote sensing image processing, IEEE Access, 8 (2020), 155783–155796. https://doi.org/10.1109/ACCESS.2020.3018544 doi: 10.1109/ACCESS.2020.3018544
    [6] L. Hua, Y. Qiang, J. Gu, L. Chen, X. Zhang, H. Zhu, Mechanical fault diagnosis using color image recognition of vibration spectrogram based on quaternion invariable moment, Math. Probl. Eng., 2015 (2015), 1–11. https://doi.org/10.1155/2015/702760 doi: 10.1155/2015/702760
    [7] M. Hasan, B. P. Mandal, New scattering features of quaternionic point interaction in non-Hermitian quantum mechanics, J. Math. Phys., 61 (2020), 032104. https://doi.org/10.1063/1.5117873 doi: 10.1063/1.5117873
    [8] R. J. Goodman, Digital simulation of aerospace vehicle flight path dynamics using quaternions, in Prague International Astronautical Federation Congress, 1977.
    [9] H. Wang, G. Wei, S. Wen, T. Huang, Impulsive disturbance on stability analysis of delayed quaternion-valued neural networks, Neurocomputing, 390 (2021), 125680. https://doi.org/10.1016/j.amc.2020.125680 doi: 10.1016/j.amc.2020.125680
    [10] J. Shu, B. Wu, L. Xiong, T. Wu, H. Zhang, Stochastic stabilization of Markov jump quaternion-valued neural network using sampled-data control, Appl. Math. Comput., 400 (2021), 1260414. https://doi.org/10.1016/j.amc.2021.126041 doi: 10.1016/j.amc.2021.126041
    [11] Y. Zhang, L. Zhou, Novel global polynomial stability criteria of impulsive complex-valued neural networks with multi-proportional delays, Neural Comput. Appl., 34 (2022), 2913–2924. https://doi.org/10.1007/s00521-021-06555-w doi: 10.1007/s00521-021-06555-w
    [12] S. Wang, Y. Shi, J. Guo, Exponential stability of a class of quaternion-valued memristor-based neural network with time-varying delay via M-matrix, Math. Methods Appl. Sci., 2024. https://doi.org/10.1002/mma.10486
    [13] Q. Song, Y. Chen, Z. Zhao, Y. Liu, F. Alsaadi, Robust stability of fractional-order quaternion-valued neural networks with neutral delays and parameter uncertainties, Neurocomputing, 420 (2021), 70–81. https://doi.org/10.1016/j.neucom.2020.08.059 doi: 10.1016/j.neucom.2020.08.059
    [14] R. Li, J. Cao, N. Li, Stop and go strategy for Lagrange stability of quaternion-valued memristive neural networks, Math. Methods Appl. Sci., 46 (2023), 6578–6589. https://doi.org/10.1002/mma.8926 doi: 10.1002/mma.8926
    [15] W. Liu, J. Huang, Q. Yao, Stability analysis for quaternion-valued inertial memristor-based neural networks with time delays, Neurocomputing, 448 (2021), 67–81. https://doi.org/10.1016/j.neucom.2021.03.106 doi: 10.1016/j.neucom.2021.03.106
    [16] Y. Shi, X. Chen, P. Zhu, Dissipativity for a class of quaternion-valued memristor-based neutral-type neural networks with time-varying delays, Math. Methods Appl. Sci., 46 (2023), 18166–18184. https://doi.org/10.1002/mma.9551 doi: 10.1002/mma.9551
    [17] T. Peng, J. Lu, Z. Tu, J. Lou, Finite-time stabilization of quaternion-valued neural networks with time delays: An implicit function method, Inf. Sci., 613 (2022), 747–762. https://doi.org/10.1016/j.ins.2022.09.014 doi: 10.1016/j.ins.2022.09.014
    [18] G. Tan, Z. Wang, Z. Shi, Proportional-integral state estimator for quaternion-valued neural networks with time-varying delays, IEEE Trans. Neural Networks Learn. Syst., 34 (2023), 1074–1079. https://doi.org/10.1109/TNNLS.2021.3103979 doi: 10.1109/TNNLS.2021.3103979
    [19] J. Hu, G. Tan, L. Liu, A new result on H state estimation for delayed neural networks based on an extended reciprocally convex inequality, IEEE Trans. Circuits Syst. II Express Briefs, 71 (2024), 1181–1185. https://doi.org/10.1109/TCSII.2023.3323834 doi: 10.1109/TCSII.2023.3323834
    [20] J. Cai, J. Feng, J. Wang, Y. Zhao, Quasi-synchronization of neural networks with diffusion effects via intermittent control of regional division, Neurocomputing, 409 (2020), 146–156. https://doi.org/10.1016/j.neucom.2020.05.037 doi: 10.1016/j.neucom.2020.05.037
    [21] R. Li, X. Gao, J. Cao, Quasi-state estimation and quasi-synchronization control of quaternion-valued fractional-order fuzzy memristive neural networks: Vector ordering approach, Appl. Math. Comput., 362 (2019), 124572. https://doi.org/10.1016/j.amc.2019.124572 doi: 10.1016/j.amc.2019.124572
    [22] X. Song, X. Li, S. Song, Y. Zhang, Z. Ning, Quasi-synchronization of coupled neural networks with reaction-diffusion terms driven by fractional brownian motion, J. Franklin Inst., 358 (2021), 2482–2499. https://doi.org/10.1016/j.jfranklin.2021.01.023 doi: 10.1016/j.jfranklin.2021.01.023
    [23] Z. Zhang, T. Zheng, S. Yu, Finite-time anti-synchronization of neural networks with time-varying delays via inequality skills, Neurocomputing, 356 (2019), 60–68. https://doi.org/10.1016/j.neucom.2019.05.012 doi: 10.1016/j.neucom.2019.05.012
    [24] X. Liu, Z. Li, Finite time anti-synchronization of complex-valued neural networks with bounded asynchronous time-varying delays, Neurocomputing, 387 (2020), 129–138. https://doi.org/10.1016/j.neucom.2020.01.035 doi: 10.1016/j.neucom.2020.01.035
    [25] Y. Qiao, H. Yan, L. Duan, J. Miao, Finite-time synchronization of fractional-order gene regulatory networks with time delay, Neural Networks, 126 (2020), 1–10. https://doi.org/10.1016/j.neunet.2020.02.004 doi: 10.1016/j.neunet.2020.02.004
    [26] T. Peng, J. Zhong, Z. Tu, J. Lu, J. Lou, Finite-time synchronization of quaternion-valued neural networks with delays: A switching control method without decomposition, Neural Networks, 148 (2022), 37–47. https://doi.org/10.1016/j.neunet.2021.12.012 doi: 10.1016/j.neunet.2021.12.012
    [27] T. Peng, J. Qiu, J. Lu, Z. Tu, J. Cao, Finite-time and fixed-time synchronization of quaternion-valued neural networks with/without mixed delays: An improved one-norm method, IEEE Trans. Neural Networks Learn. Syst., 12 (2022), 7475–7487. https://doi.org/10.1109/TNNLS.2021.3085253 doi: 10.1109/TNNLS.2021.3085253
    [28] D. Ding, X. Yao, H. Zhang, Complex projection synchronization of fractional-order complex-valued memristive neural networks with multiple delays, Neural Process. Lett., 51 (2020), 325–345. https://doi.org/10.1007/s11063-019-10093-x doi: 10.1007/s11063-019-10093-x
    [29] Y. Zhang, S. Deng, Finite-time projective synchronization of fractional-order complex-valued memristor-based neural networks with delay, Chaos, Solitons Fractals, 128 (2019), 176-190. https://doi.org/10.1016/j.chaos.2019.07.043 doi: 10.1016/j.chaos.2019.07.043
    [30] Y. Cheng, Y. Shi, The exponential synchronization and asymptotic synchronization of quaternion-valued memristor-based Cohen-Grossberg neural networks with time-varying delays, Neural Process. Lett., 55 (2023), 6637–6656. https://doi.org/10.1007/s11063-023-11152-0 doi: 10.1007/s11063-023-11152-0
    [31] Y. Cheng, Y. Shi, J. Guo, Exponential synchronization of quaternion-valued memristor-based Cohen-Grossberg neural networks with time-varying delays: Norm method, Cognit. Neurodyn., 18 (2024), 1943–1953. https://doi.org/10.1007/s11571-023-10057-x doi: 10.1007/s11571-023-10057-x
    [32] H. L. Li, L. Zhang, C. Hu, H. Jiang, J. Cao, Global Mittag-Leffler synchronization of fractional-order delayed quaternion-valued neural networks: Direct quaternion approach, Appl. Math. Comput., 373 (2020), 125020. https://doi.org/10.1016/j.amc.2019.125020 doi: 10.1016/j.amc.2019.125020
    [33] Y. Kao, Y. Li, J. H. Park, X. Chen, Mittag-Leffler synchronization of delayed fractional memristor neural networks via adaptive control, IEEE Trans. Neural Networks Learn. Syst., 32 (2020), 2279–2284. https://doi.org/10.1109/TNNLS.2020.2995718 doi: 10.1109/TNNLS.2020.2995718
    [34] J. Cheng, L. Xie, D. Zhang, H. Yan, Novel event-triggered protocol to sliding mode control for singular semi-Markov jump systems, Automatica, 151 (2023), 110906. https://doi.org/10.1016/j.automatica.2023.110906 doi: 10.1016/j.automatica.2023.110906
    [35] J. Cheng, J. H. Park, Z. Wu, Observer-based asynchronous control of nonlinear systems with dynamic event-based try-once-discard protocol, IEEE Trans. Cybern., 52 (2022), 12638–12648. https://doi.org/10.1109/TCYB.2021.3104806 doi: 10.1109/TCYB.2021.3104806
    [36] Y. Gu, Y. Yu, H. Wang, Projective synchronization for fractional-order memristor-based neural networks with time delays, Neural Comput. Appl., 31 (2019), 6039–6054. https://doi.org/10.1007/s00521-018-3391-7 doi: 10.1007/s00521-018-3391-7
    [37] G. Velmurugan, R. Rakkiyappan, Hybrid projective synchronization of fractional-order memristor-based neural networks with time delays, Nonlinear Dyn., 83 (2016), 419–432. https://doi.org/10.1007/s11071-015-2337-1 doi: 10.1007/s11071-015-2337-1
    [38] H. Bao, J. Cao, Projective synchronization of fractional-order memristor-based neural networks, Neural Networks, 63 (2015), 1–9. https://doi.org/10.1016/j.neunet.2014.10.007 doi: 10.1016/j.neunet.2014.10.007
    [39] R. Li, X. Gao, J. Cao, K. Zhang, Exponential stabilization control of delayed quaternion-valued memristive neural networks: Vector ordering approach, Circuits Syst. Signal Process., 39 (2020), 1353–1371. https://doi.org10.1007/s00034-019-01225-8 doi: 10.1007/s00034-019-01225-8
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(679) PDF downloads(40) Cited by(0)

Figures and Tables

Figures(8)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog