Title :
Parallel Implementation of Gradient-Based Neural Networks for SVM Training
Author :
Ferreira, Leonardo V. ; Kaszkurewicz, Eugenius ; Bhaya, Amit
Author_Institution :
Rio de Janeiro Fed. Univ., Rio de Janeiro
Abstract :
This paper presents the implementation of two neural networks for SVM training in parallel computers. The results obtained are compared with two well known packages for SVM training and the parallel implementation shows that the neural network approach can be as accurate as the traditional packages and, since the proposed gradient-based neural networks can be easily parallelized, the proposed approach is scalable and the training times can be considerably reduced.
Keywords :
gradient methods; learning (artificial intelligence); neural nets; parallel architectures; pattern classification; support vector machines; SVM Training; binary classification tasks; gradient-based neural networks; parallel computers; support vector machines; Computer networks; Concurrent computing; Electronic mail; Libraries; Machine learning; Neural networks; Packaging machines; Performance analysis; Support vector machine classification; Support vector machines;
Conference_Titel :
Neural Networks, 2006. IJCNN '06. International Joint Conference on
Conference_Location :
Vancouver, BC
Print_ISBN :
0-7803-9490-9
DOI :
10.1109/IJCNN.2006.246701