next up previous
Next: はじめに



Fast learning algorithm for Bayesian self-organizing maps

Akio Utsugi
National Institute of Bioscience and Human-Technology

Abstract The Bayesian self-organizing map (BSOM) is a statistical method to extract intrinsic topological structures hidden in multidimensional data. Unlike the traditional self-organizing map (SOM), the learning of BSOM is regarded as an estimation algorithm of a stochastic model, and thus statistical techniques for high-level inference are available. In this paper, a fast search algorithm for optimal parameters and hyperparameters of BSOM is presented.

Utsugi Akio;(6730)
Thu Aug 8 18:07:48 JST 1996