Adaptive Scaling of Codebook Vectors
In this paper we introduce a vector quantization algorithm in which the codebook vectors are extended with a scale parameter to let them represent Gaussian functions. The means of these functions are determined by a standard vector quantization algorithm; and for their scales we have derived a learning rule. Our algorithm estimates probability densities efficiently. The main application is pattern classification.
KeywordsLearning Rule Vector Quantization Radial Basis Function Network Equilibrium Constraint Code Book
Unable to display preview. Download preview PDF.
- J. Moody and C.J. Darken. Fast learning in networks of locallv-tuned processing units. Neural Computation,Google Scholar
- B. Kosko. Neural Networks and Fuzzy Systems. Prentice-Hall International Editions, 1992.Google Scholar
- T. Kohonen. Self-Organization and Associative Memory. Springer-Verlag, 1984.Google Scholar
- H. Ritter, T. Martinetz, and Klaus Schulten. Neural Computation and Self-Organizing Maps. Addison Wesley, 1992.Google Scholar
- D. DeSieno. Adding a conscience to competitative learning. In Proceedings of the International Joint Conference on Neural Networks, pages 117–124. IEEE, 1988.Google Scholar