Application of SOM neural network in clustering by ProQuest

VIEWS: 136 PAGES: 8

More Info
									J. Biomedical Science and Engineering, 2009, 2, 637-643                                                                     JBiSE
doi: 10.4236/jbise.2009.28093 Published Online December 2009 (http://www.SciRP.org/journal/jbise/).




Application of SOM neural network in clustering
Soroor Behbahani1, Ali Moti Nasrabadi2
1
 Biomedical Engineering Department, Science and Research Branch, Islamic Azad University, Tehran, Iran;
2
 Biomedical Engineering Department, Faculty of Engineering, Shahed University, Tehran, Iran.
Email: soroor_behbahani@yahoo.com; a_m_nasrabadi@yahoo.com

Received 11 June 2009; revised 29 June 2009; accepted 27 July 2009.

ABSTRACT                                                              widespread use is the identification and visualization of
                                                                      natural groupings in the data. The process of finding
The Self-Organizing Map (SOM) is an unsupervised
                                                                      similar items is generally referred to as clustering.
neural network algorithm that projects high-dimen-
                                                                      Compared to the k-means clustering algorithm, the SOM
sional data onto a two-dimensional map. The projec-
                                                                      exemplifies a robust and structured self-organizing neu-
tion preserves the topology of the data so that similar
                                                                      ral networks are based on the principle of transforming a
data items will be mapped to nearby locations on the
                                                                      set of p-variate observations into a spatial representation
map. One of the SOM neural network’s applications
                                                                      of smaller dimensionality, which may allow a more ef-
is clustering of animals due their features. In this
                                                                      fective visualization of correlations in the original data
paper we produce an experiment to analyze the SOM
                                                                      [4].
in clustering different species of animals.
                                                                      2. SELF-ORGANIZING MAP
Keywords: SOM Neural Network; Feature; Clustering;
Animal                                                                The Self-Organizing Map belongs to the class of unsu-
                                                                      pervised and competitive learning algorithms. It is a
1. INTRODUCTION                                                       sheet-like neural network, with nodes arranged as a
                                                                      regular, usually two-dimensional grid. As explained in
The Self-Organizing Map (SOM) is a fairly well-known                  the previous section on Neural Networks, we usually
neural network and indeed one of the most popular un-                 think of the node connections as being associated with a
supervised learning algorithms. Since its invention by
                                                                      vector of weights. In the case of Self-Organizing Maps,
Finnish Professor Teuvo Kohonen in the early 1980s,
                                                                      it is easier to think of each node as being directly associ-
more than 4000 research articles have been published on
                                                                      ated with a weight vector.
the algorithm, its visualization and applications. The
                                                                          The items in the input data set are assumed to be in a
maps comprehensively visualize natural groupings and
relationships in the data and have been successfully ap-              vector format. If n is the dimension of the input space,
plied in a broad spectrum of research areas ranging from              then every node on the map grid holds an n-dimensional
speech recognition to financial analysis. The Self-organ-             vector of weights:
izing Map performs a non-linear projection of multidi-                              mi = [mi1, mi2, mi3, . , min]               (1)
mensional data onto a two-dimensional display. The
                                                                      The basic principle of the Self-Organizing Map is to
mapping is topology-preserving, meaning that the more
                                                                      adjust these weight vectors until the map represents a
alike two data samples are in the input space, the closer
they will appear together on the final map. The SOM                   picture of the input data set. Since the number of map
belongs to the class of Neural Network algorithms. This               nodes is significantly smaller than the number of items
is a group of algorithms based on analogies to the neural             in the dataset, it is needless to say that it is impossible to
structures of the brain. The SOM in particular was in-                represent every input item from the
								
To top