| 摘要 |
<p>PROBLEM TO BE SOLVED: To efficiently classify input information having an unknown distribution without setting complicated parameters. SOLUTION: Information is inputted as multi-dimensional vectors by an information input part 11. Input information is arranged on a two-dimensional self-organizing map in the descending order of the degree of similarity by a mapping part 12. Degrees of similarity between units constituting the self- organizing map are calculated by a distance map generation part 13 to generate a distance map representing degrees of similarity between units. Only boundaries of clusters are taken out from the distance map by a clustering map generation part 14 to generate a clustering map where cluster boundaries are described. Finally, input information is recognized for each cluster to perform clustering by a cluster recognition part 15.</p> |
