Incremental Self-Organizing Maps for Collaborative Clustering

摘要

Collaborative clustering aims at revealing the common structures of data distributed on different sites using local clustering methods such as Self-Organizing Maps (SOM). To face the ever growing quantity of data available, incremental clustering methods are needed. This paper presents an algorithm to perform incremental SOM-based collaborative clustering without topological modifications of the map. The experiments conducted on several datasets demonstrate the validity of the method and present the influence of the batch size on the learning.