Dynamics of Learning in Recurrent Feature-Discovery Networks

Leen, Todd K.

Neural Information Processing Systems 

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

Similar Docs  Excel Report  more

TitleSimilaritySource
None found