抄録
Self-organizing map (SOM) is a type of artificial
neural network. SOM is trained using unsupervised learning to
produce low dimensional representation of the training samples
while preserving the topological information of the input space.
There are three problems when applying SOM for clustering:
map initialization, computational cost, and limited capabilities
for the representation. Hierarchical SOM and tree structured
SOM have been previously proposed to solve the problems of map
initialization or computational cost.
In this paper we propose an adaptive tree structured
clustering method using SOM in order to improve the
classification capability. In our proposed method, separate SOMs
are arranged to correspond to nodes of a binary tree structure.
The binary tree structure is generated by recursive child node
creation that is determined by the classification results of the
corresponding parent node SOM. The proposed method utilizes
the competitive learning feature of SOM, and the relationships in
the data set are shown as the generated tree structure.
