球面SOMのデータ構造と量子化誤差の考察およびインタラクティビティの向上(<特集>生体情報解析と複雑系科学)

概要

論文の詳細を見る
Implementing the Self-Organizing map (SOM) on a spherical lattice is an effective way to remove the "border effect". Several spherical SOMs based on a geodesic dome have been proposed. In this paper, we present a geodesic dome-based SOM (GeoSOM) and its quantization error analyses. Experimental results show that the GeoSOM runs almost at the same speed of the conventional 2D SOM and it also represents the data more accurately. Furthermore, for the purpose of Information Visualization, we present a method to convert the 3D GeoSOM into a 2D interactive map. This 2D visualization of the GeoSOM allows a user to examine every part of the sphere while maintaining the overall view of the sphere without retraining the SOM.
日本知能情報ファジィ学会の論文
2007-12-15