Self-Organizing map and Points of Interest

Abstract

We propose in this paper a new manner to learn topographic clustering which uses points of interest. This new approach introduces in the learning phase a new concept of referent of interest. These referents are automatically detected during the learning phase or deducted from knowledge of the database. Referents of interest will have more active role in the topological map organization. At the end, we evaluate the performance of our new approach on several databases with different difficulties. The obtained results are encouraging and promising.