Neural Nets with Markers and Gaussian-distributed Connectivities
E. Fournou, P. Argyrakis, Photios A. Anninos · Connection Science · 1993
We investigate probabilistic neural nets with the inclusion of chemical markers and Gaussian distribution of the connectivities of the constituent neurons. We start from the studied case of Poisson distributions and extend it to an analogous formalism for the activity of a netlet with Gaussian characteristics. The analytical formulae are somewhat more complicated due to the presence of markers. The results show that the change from a Poisson to a Gaussian distribution may cause a net to change class if it belongs to class A, making it class B. This trend is similar to the one observed in the absence of markers. We also observe interesting trends in the variation of the sizes of the markers by isolating the contribution of each subnet to the overall activity. Finally, the general repercussions of the present work to our understanding of the dynamics of the brain network are discussed.