Stochastic and deterministic neural networks with a continuous state space and a connectivity greater than two
Jérôme Lacaille · 2002
This article is divided into two parts, which both give a detailed observation of a particular type of recurrent network, presenting cells which activities continuously evolve in an interval of R. The first part of this article shows a stochastic type of network derived from Boltzmann machines, whereas the second part is devoted to determinist process. In both cases, we will formalize the dynamic system, and give an algorithm of relaxation and learning.>