Effects of Lévy noise in a neuronal competition model

The effects of non-Gaussian Lévy noise in a competitive neural model of binocular rivalry are analyzed numerically. We first investigate how Lévy noise added to the input strength changes the period of oscillations and leads to the disappearance of the winner-take-all behavior. Noise-induced alternations are then studied in terms of their statistical measures. We are interested in the overlapping of the regions where the model’s behavior satisfies the experimental constraints on the mean dominance duration and the coefficient of variation of the dominance durations. The results indicate that changes in each of the parameters of Lévy noise produce changes in the overlap regions. The differences with Gaussian noise are emphasized. Finally we plot the distributions of the perceptual dominance durations for different parameters of Lévy noise and compare these distributions with gamma, log-normal and Weibull distributions using the Kolmogorov–Smirnov goodness-of-fit test.