Estimation of the basin of attractions of stable equilibrium points in CNNs

Abstract

We present an approach to estimate the basin of attraction of stable equilibrium points in cellular neural networks (CNNs). The approach is based on the determination of the so called tree of regions connected with each stable equilibrium points described in our previous work (1997). The new contribution is connected with additional separation of the regions where the boundaries between different basins are located. The suggested separation with internal hyperplanes will help to estimate more precisely the boundaries between different basins, because the previous algorithm to obtain the trees could not give the exact description of the basin of attractions.