Multistability and instability of competitive neural networks with non-monotonic piecewise linear activation functions (Q1729198)

In this paper, a class of competitive neural networks with non-monotonic piecewise linear activation functions is studied. Stability results for the equilibrium points of such networks are obtained by means of a fixed point theorem, the contraction mapping theorem and the eigenvalue properties of a strict diagonal dominance matrix. The main result shows that the competitive neural networks under consideration have exactly $5^n$ equilibrium points, among which $3^n$ equilibrium points are locally exponentially stable and the others are unstable. At the end of the paper, one example is provided and numerical simulations are made in order to illustrate the obtained theoretical results.