A Markovian study of recurrent neural networks with stochastic dynamics

Article type: Research Article

Authors: Zaharie, Daniela

Affiliations: Faculty of Mathematics, The West University of Timişoara, V. Pârvan bv. 4, 1900 Timişoara, Romania

Abstract: Recurrent neural networks of binary stochastic units with a general distribution function are studied using Markov chains theory. Sufficient conditions for ergodicity are established and under some assumptions, the stationary distribution is determined. The relation between fixed points and absorbing states is studied both theoretically and through simulations. For numerical studies the notion of almost absorbing state is introduced.

Keywords: stochastic neural network, ergodicity, absorbing state

DOI: 10.3233/INF-1996-7206

Journal: Informatica, vol. 7, no. 2, pp. 255-267, 1996