Back to first pageBack to first page Centre for Artificial Intelligence of UNL
Browse our site
You are here:

Publication details

Go back
Publication details
Main information
A Hybrid Parallel SOM Algorithm for Large Maps in Data-Mining
December 2007
sm07
We propose a method for a parallel implementation of the Self-Organizing Map (SOM) algorithm. We call this method Hybrid in the sense that it combines the advantages of the common network-partition and data-partition approaches, and is particularly efective when dealing with large maps. Our method relies on the fact that global topological ordering of the map is achieved during the first iterations of the SOM algorithm. The input data histogram over the map is then used to seg- ment the input vectors per processing nodes. Data may be moved among partitions by repeating this segmentation regularly. Our experimental re- sults show an average speed-up of 1.27 compared to the classical Batch data-partition method, while maintaining the topological information of the maps.
In proceedings
Bruno Silva, Nuno Marques
José Neves, Manuel Filipe Santos, José Machado
New Trends in Artificial Intelligence
-
Associação Portuguesa para a Inteligência Artificial (APPIA)
Guimarães. Portugal
-
-
ISBN-13 978-989-9561
-
-
http://ssdi.di.fct.unl.pt/~nmm/MyPapers/SM2007.pdf
Export formats
Bruno Silva and Nuno Marques, A Hybrid Parallel SOM Algorithm for Large Maps in Data-Mining, in: José Neves and Manuel Filipe Santos and José Machado (eds), New Trends in Artificial Intelligence, Associação Portuguesa para a Inteligência Artificial (APPIA), Guimarães. Portugal, ISBN ISBN-13 978-989-9561, (http://ssdi.di.fct.unl.pt/~nmm/MyPapers/SM2007.pdf), December 2007.
<a href="/people/members/view.php?code=877b33046c8fe81a875041f9b0889d62" class="author">Bruno Silva</a> and <a href="/people/members/view.php?code=9ac1f1682cb91b20d9fbfb8e659b0819" class="author">Nuno Marques</a>, <b>A Hybrid Parallel SOM Algorithm for Large Maps in Data-Mining</b>, in: José Neves, Manuel Filipe Santos and José Machado (eds), <u>New Trends in Artificial Intelligence</u>, <a href="http://www.appia.pt/" title="Link to external entity..." target="_blank" class="publisher">Associação Portuguesa para a Inteligência Artificial (APPIA)</a>, Guimarães. Portugal, ISBN ISBN-13 978-989-9561, (<a href="http://ssdi.di.fct.unl.pt/~nmm/MyPapers/SM2007.pdf" target="_blank">url</a>), December 2007.
@inproceedings {sm07, author = {Bruno Silva and Nuno Marques}, editor = {Jos{\'e} Neves and Manuel Filipe Santos and Jos{\'e} Machado}, title = {A Hybrid Parallel SOM Algorithm for Large Maps in Data-Mining}, booktitle = {New Trends in Artificial Intelligence}, publisher = {Associa\c{c}{\~a}o Portuguesa para a Intelig{\^e}ncia Artificial (APPIA)}, address = {Guimar{\~a}es. Portugal}, isbn = {ISBN-13 978-989-9561}, url = {http://ssdi.di.fct.unl.pt/~nmm/MyPapers/SM2007.pdf}, abstract = {We propose a method for a parallel implementation of the Self-Organizing Map (SOM) algorithm. We call this method Hybrid in the sense that it combines the advantages of the common network-partition and data-partition approaches, and is particularly efective when dealing with large maps. Our method relies on the fact that global topological ordering of the map is achieved during the first iterations of the SOM algorithm. The input data histogram over the map is then used to seg- ment the input vectors per processing nodes. Data may be moved among partitions by repeating this segmentation regularly. Our experimental re- sults show an average speed-up of 1.27 compared to the classical Batch data-partition method, while maintaining the topological information of the maps.}, month = {December}, year = {2007}, }
Publication's urls
/publications/view.php?code=04663ebd9e4012cf6e8adee108a8b8e7
/publications/view.php?code=sm07
Go back
Centre for Artificial Intelligence of UNL
Departamento de Informática, FCT/UNL
Quinta da Torre 2829-516 CAPARICA - Portugal
Tel. (+351) 21 294 8536 FAX (+351) 21 294 8541

Fundacao_FCT