Attractors in continuous-time switching networks (Q1424088)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Attractors in continuous-time switching networks
scientific article

    Statements

    Attractors in continuous-time switching networks (English)
    0 references
    0 references
    8 March 2004
    0 references
    The author studies the dynamics of equations with discontinuous right-hand side \[ \dot x_i = -x_i + \Lambda_i(x_1,\dots,x_n), \quad i=1,\dots,n. \tag{1} \] The functions \(\Lambda_i\) are assumed to be constant in the interior of every orthant in \(\mathbb R^n\). This system arises as a model of gene networks, where the effect of each gene on another depends on two states -- the ON state and an OFF state. Due to the piecewise linear structure, system (1) can be explicitly solved in every orthant in \(\mathbb R^n\). This helps under some conditions to associate the dynamics of (1) with the structure of some graph \(G\). The author considers the attractors in \(\mathbb R^4\), which lie in a set of orthants in the form of figure eight. In particular, it is found that if the attractor is symmetric with respect to the two loops of this figure eight, then the only possible attractor is a periodic orbit which intersects both loops once. In the case without symmetry, the set of admissible attractors includes periodic orbits which follow one loop \(k\) times and other loop once, for any \(k\). It is also shown that no trajectory in the attractor can intersect both loops more then once in a row.
    0 references
    attractors
    0 references
    continuous-time switching networks
    0 references
    additive neural networks
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references