Clustering Quality and Topology Preservation in Fast Learning Soms

Summary

---

The Self-Organizing Map (SOM) is a popular unsupervised neural network able to provide effective clustering and data visualization for data represented in multidimensional input spaces. In this paper, we describe Fast Learning SOM (FLSOM) which adopts a learning algorithm that improves the performance of the standard SOM with respect to the convergence time in the training phase. We show that FLSOM also improves the quality of the map by providing better clustering quality and topology preservation of multidimensional input data. Several tests have been carried out on different multidimensional datasets, which demonstrate better performances of the algorithm in comparison with the original SOM.

See the full content of this document

Extract

---

Clustering Quality and Topology Preservation in Fast Learning Soms

(ProQuest: ... denotes formulae omitted.)

1. Introduction

The Self-Organizing Map (SOM) [1] algorithm is used to build a mapping of an input manifold from a high-dimensional data space to a low-dimensional representation space. The Fast Learning SOM [FLSOM) [2] is an approach to speed up the learning process by introducing an optimization technique based on the Simulated Annealing (SA).

In this work, we demonstrate that the FLSOM algorithm is also able to perform a "good" clustering. We evaluate and compare the algorithm with a standard SOM over a number of artificial and real datasets. The experimental analysis shows that FLSOM provides better results in terms of both topology preservation and clustering criteria.

The paper has the following structure: in the next section, we briefly discuss some related works in the field of the data projection and SOM learning algorithms. The Section 3 shows the basic SOM algorithm and describes the proposed one; the Section 4 shows the advantage of using the proposed algorithm for clustering process; the Sections 5 and 6 report respectively the evaluation criteria and the experimental results. Finally in Section 7 some conclusions are reported....

See the full content of this document

---