Global analysis of planar neural networks (Q2492419): Difference between revisions

A two-dimensional learning model is presented as a cubic polynomial dynamical system in the following form \[ \dot{x}=\alpha x+\beta y-x(ax^2+2bxy+cy^2),\quad \dot{y}=\gamma x+\delta y-y(ax^2+2bxy+cy^2). \] The paper provides a global qualitative analysis of the system. In particular, it studies singular point bifurcations and the existence and number of limit cycles. The main tools employed include, among others, the Poincaré index theorems, Dulac criterion, and Wintner-Perko termination principle.