"Generalized radial basis function networks for classification and novelty detection: Selforganization of optimal Bayesian decision" S. Albrecht, J. Busch, M. Kloppenburg, F. Metze und P. Tavan
Neural Networks 13: 10751093 (2000)
Abstract: By adding reverse connections from the output layer to the
central layer it is shown how a generalized radial basis
functions (GRBF) network can selforganize to form a Bayesian
classifier which is also capable of novelty detection.
For that purpose three stochastic sequential learning rules
are introduced from biological considerations which pertain to
the centers, the shapes, and the widths of the receptive fields
of the neurons and allow a joint optimization of all network
parameters. The rules are shown to generate maximum
likelihood estimates of the classconditional probability
density functions of labeled data in terms of multivariate
normal mixtures. Upon combination with a hierarchy of
deterministic annealing procedures, which implement a
multiple scale approach, the learning process can avoid the
convergence problems hampering conventional
expectationmaximization algorithms. Using an example
from the field of speech recognition
the stages of the learning process and the capabilities of the
selforganizing GRBF classifier are illustrated.
BMO authors (in alphabetic order): Sebastian Albrecht Jan Busch Martin Kloppenburg Florian Metze Paul Tavan
Assoziierte Projekte: Selforganized pattern classification and its application to molecular structures Speech recognition in neural architecture
