Convergence and periodicity in a discrete-time network of two neurons with self-connections (Q597265): Difference between revisions

The authors study the difference system \[ x_n=\lambda x_{n-1}+(1-\lambda) \left\{ \tfrac {a}{2} [f(x_{n-1})+f(y_{n-1})] + \tfrac {c}{2} [f(x_{n-1})-f(y_{n-1})] \right\} \] \[ y_n=\lambda y_{n-1}+(1-\lambda) \left\{ \tfrac{b}{2} [f(x_{n-1})+f(y_{n-1})] + \tfrac{d}{2} [f(x_{n-1})-f(y_{n-1})] \right\} \] modeling a two-neuron network with feedback and McCulloch-Pitts type activation function given by \(f(\xi)=-1\) for \(\xi>\sigma\) and \(f(\xi)=1\) for \(\xi\leq\sigma\), where \(\sigma\) is a real constant which acts as the threshold. They prove several results guaranteeing the convergence of solutions of the above system as well as the existence of periodic solutions.