Dynamics of Learning in Recurrent Feature-Discovery Networks

Part of Advances in Neural Information Processing Systems 3 (NIPS 1990)

Bibtex Metadata Paper

Authors

Todd Leen

Abstract

The self-organization of recurrent feature-discovery networks is studied from the perspective of dynamical systems. Bifurcation theory reveals pa(cid:173) rameter regimes in which multiple equilibria or limit cycles coexist with the equilibrium at which the networks perform principal component analysis.