Journal of Japan Society for Fuzzy Theory and Intelligent Informatics
Online ISSN : 1881-7203
Print ISSN : 1347-7986
ISSN-L : 1347-7986
Original Papers
Self-Organization of Shape Space by the Higher-Rank SOM-Topology-Free Shape Space Method
Makoto OTANITetsuo FURUKAW
Author information
JOURNAL FREE ACCESS

2013 Volume 25 Issue 2 Pages 701-720

Details
Abstract

The aims of this paper are to establish the shape space theory which deals with a set of shapes, and to develop a shape space estimation method which does not require the prior knowledge about the topology of shapes. To achieve these aims, first we developed a Self-Organizing Map( SOM) that is free from topological restriction, namely, Topology-Free SOM( TFSOM). Then we built TFSOM into the higher-rank SOM, which is called TFSOM×SOM in this paper. TFSOM×SOM was applied to handwritten digit recognition task, and it showed about 95% accuracy in recognizing digits written by 1000 people, after the TFSOM×SOM was trained by the digits of only 10 people. Since the proposal method is not specialized to line-drawing, it can be applied to wider tasks relevant to shapes.

Content from these authors
© 2013 Japan Society for Fuzzy Theory and Intelligent Informatics
Previous article
feedback
Top