论文信息 - Dynamics of Learning in Recurrent Feature-Discovery Networks

Dynamics of Learning in Recurrent Feature-Discovery Networks

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.

Todd K. Leen

[1] J. Rubner,et al. Development of feature detectors by self-organization. A network model. , 1990, Biological cybernetics.

[2] E. Oja,et al. On stochastic approximation of the eigenvectors and eigenvalues of the expectation of a random matrix , 1985 .

[3] Sun-Yuan Kung,et al. A neural network learning algorithm for adaptive principal component extraction (APEX) , 1990, International Conference on Acoustics, Speech, and Signal Processing.

[4] M. Golubitsky,et al. Singularities and groups in bifurcation theory , 1985 .

[5] Erkki Oja,et al. Neural Networks, Principal Components, and Subspaces , 1989, Int. J. Neural Syst..

[6] T. Leen. Dynamics of learning in linear feature-discovery networks , 1991 .

[7] Terence D. Sanger,et al. An Optimality Principle for Unsupervised Learning , 1988, NIPS.

[8] P. Foldiak,et al. Adaptive network for optimal linear feature extraction , 1989, International 1989 Joint Conference on Neural Networks.

[9] E. Oja. Simplified neuron model as a principal component analyzer , 1982, Journal of mathematical biology.

[10] M. Golubitsky,et al. Singularities and Groups in Bifurcation Theory: Volume I , 1984 .