Sigmoidal approximations of a delay neural lattice model with heaviside functions (Q2175719)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Sigmoidal approximations of a delay neural lattice model with heaviside functions
scientific article

    Statements

    Sigmoidal approximations of a delay neural lattice model with heaviside functions (English)
    0 references
    0 references
    29 April 2020
    0 references
    The paper investigates autonomous systems of differential equations on the lattice \(\mathbb{Z}^d\) with delays, of the type \[(3) \quad \dot u_i(t) = f_i(u_i(t)) + \sum_{j \in \mathbb{Z}^d} k_{i,j} \sigma_{\varepsilon}[u_j(t - \tau_{i,j}) - \theta] + g_i \quad (i \in \mathbb{Z}^d)\] Here \(\sigma_{\varepsilon}\) is a continuous `sigmoidal' response function which converges pointwise to the Heaviside function \(H\) as \(\varepsilon \to 0\). The system is related to the corresponding system with the discontinuous function \(H\) instead of \(\sigma_{\varepsilon}\), formulated as a differential inclusion (with [0,1] as value at zero), and to an `inflated' inclusion system, where the response function is interval-valued (with \(\varepsilon-\)dependent intervals) at every point of \(\mathbb{R}\). The space \( C^0([-h,0], \ell^2_{\rho}(\mathbb{Z}^d, \mathbb{R}))\) is taken as state space for the equation, where \(h\) is the maximal delay and \( \ell^2_{\rho}(\mathbb{Z}^d, \mathbb{R})\) is the \(\ell^2 \)-space with weight \( \rho\). Seven assumptions (mainly on boundedness, especially of the \(f_i\) and their derivatives) allow to prove the following results: Sequences of solutions to the above system have sequences converging to a solution of the `Heaviside' differential inclusion (the argument is of ArzelĂ -Ascoli type). Further, both the `Heaviside' and the inflated inclusion generate suitable set-valued dynamical systems, with attractors \(\mathfrak{A}\) and \(\mathfrak{A}^{\varepsilon}\), respectively. For \(\varepsilon \to 0\), these attractors, and the attractor \( \mathcal A^{\varepsilon}\) of the above lattice system, are related by \(\text{dist}(\mathcal A^{\varepsilon},\mathfrak{A}) \to 0 \) (in the space \( C^0([-h,0], \ell^2_{\rho}(\mathbb{Z}^d, \mathbb{R}))\)), and even \( \text{Dist}_{H} [\mathfrak{A}^{\varepsilon}, \mathfrak{A}] \to 0 \), where \( \text{Dist}_{H}\) stands for Hausdorff distance.
    0 references
    neural lattice model with delay
    0 references
    differential delay inclusion
    0 references
    sigmoidal function
    0 references
    delay dynamical system
    0 references
    set-valued dynamical system
    0 references
    global attractors
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references