Modeling bistable perception with a network of chaotic neurons

When an ambiguous stimulus is observed, our perception undergoes dynamical changes between two states, a situation extensively explored in association with the Necker cube. Such phenomenon refers to bistable perception. Here, we present a model neural network composed of forced FitzHugh-Nagumo neurons, implemented also experimentally in an electronic circuit. We show, that under a particular coupling configuration, the neural network exhibit bistability between two configurations of clusters. Each cluster composed of two neurons undergoes independent chaotic spiking dynamics. As an appropriate external perturbation is applied to the system, the network undergoes changes in the clusters configuration, involving different neurons at each time. We hypothesize that the winning cluster of neurons, responsible for perception, is that exhibiting higher mean frequency. The clusters features may contribute to an increase of local field potential in the neural network.