Title :
The minimum entropy network
Author :
Brause, Rüdiger W.
Author_Institution :
J.W. Goethe-Univ., Frankfurt, Germany
Abstract :
It is shown that, using as basic building block a linear neuron with an anti-Hebb rule and restricted weights, an asymmetric network which computes the eigenvectors in the ascending order of their corresponding eigenvalues can be built. The conditions for their convergence are obtained and demonstrated by simulations
Keywords :
Hebbian learning; eigenvalues and eigenfunctions; entropy; neural nets; anti-Hebb rule; asymmetric network; convergence; eigenvalues; eigenvectors; linear neuron; minimum entropy network; restricted weights; Autocorrelation; Clouds; Eigenvalues and eigenfunctions; Entropy; Mean square error methods; Neural networks; Neurons; Pattern recognition; Prototypes; Vectors;
Conference_Titel :
Tools with Artificial Intelligence, 1992. TAI '92, Proceedings., Fourth International Conference on
Conference_Location :
Arlington, VA
Print_ISBN :
0-8186-2905-3
DOI :
10.1109/TAI.1992.246369
