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

Publication details

Publication details
Main information
A Study of a Hybrid Parallel SOM Algorithm for Large Maps in Data Mining
December 2008
Silva2008
The present thesis addresses a particular kind of neural network called Self-Organized Map (SOM), in particular the use of large maps for Data Mining purposes and the parallel training of the algorithm. [...] The SOM algorithm can be made parallel in several ways, featuring different granularity in execution. In general, parallel implementations of the SOM are developed by either partitioning the map (Network-Partitioning), partitioning the input data (Data-Partitioning) or partitioning the weights of the neurons (Weight-Partitioning) among the available processing units. By exploiting particular characteristics of the SOM training process, a algorithm that combines two of these common approaches, therefore called Hybrid, was devise, tested and studied.
M. Sc. dissertation
Bruno Silva
Nuno Marques
DI-FCT/UNL
-
-
Export formats
Bruno Silva, A Study of a Hybrid Parallel SOM Algorithm for Large Maps in Data Mining, Nuno Marques (superv.), DI-FCT/UNL, December 2008.
<b><a href="http://centria.di.fct.unl.pt/people/members/view.php?code=877b33046c8fe81a875041f9b0889d62_amp_cscd=c7fd03415bbdb8d1037483e2e02df742" class="author">Bruno Silva</a></b>, <u>A Study of a Hybrid Parallel SOM Algorithm for Large Maps in Data Mining</u>, <a href="http://centria.di.fct.unl.pt/people/members/view.php?code=9ac1f1682cb91b20d9fbfb8e659b0819_amp_cscd=c7fd03415bbdb8d1037483e2e02df742" class="supervisor">Nuno Marques</a> (superv.), DI-FCT/UNL, December 2008.
@mastersthesis {Silva2008, author = {Bruno Silva}, title = {A Study of a Hybrid Parallel SOM Algorithm for Large Maps in Data Mining}, school = {DI-FCT/UNL}, note = {Nuno Marques (superv.); }, abstract = {The present thesis addresses a particular kind of neural network called Self-Organized Map (SOM), in particular the use of large maps for Data Mining purposes and the parallel training of the algorithm. [...] The SOM algorithm can be made parallel in several ways, featuring different granularity in execution. In general, parallel implementations of the SOM are developed by either partitioning the map (Network-Partitioning), partitioning the input data (Data-Partitioning) or partitioning the weights of the neurons (Weight-Partitioning) among the available processing units. By exploiting particular characteristics of the SOM training process, a algorithm that combines two of these common approaches, therefore called Hybrid, was devise, tested and studied.}, month = {December}, year = {2008}, }
Publication's urls
http://centria.di.fct.unl.pt/publications/view.php?code=5368b525b4402bfca7c83ec329abb9ed
http://centria.di.fct.unl.pt/publications/view.php?code=Silva2008
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

POS_Conhecimento