Training Single Hidden Layer Feedforward Neural Networks by Singular Value Decomposition
Authors:
- Hieu Trung HuynhHHHieu Trung Huynh
- Yonggwan WonYWYonggwan Won
- 1 January 2009
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1300-1304
Abstract
Training neural networks has attracted many researchers for a long time. Many training algorithms and their improvements have been proposed. However, up to now, improving performance of training algorithms for neural networks is still a challenge. In this paper, we investigate a new training method for single hidden layer feedforward neural networks (SLFNs) which use `tansig' activation function. The proposed training algorithm uses SVD (singular value decomposition) to calculate the network parameters. It is simple and has low computational complexity. Experimental results show that the proposed approach can obtain good performance with a compact network which has small number of hidden units.Keywords
This publication has 14 references indexed in Scilit:
- AN IMPROVEMENT OF EXTREME LEARNING MACHINE FOR COMPACT SINGLE-HIDDEN-LAYER FEEDFORWARD NEURAL NETWORKSInternational Journal of Neural Systems, 2008
- Small Number of Hidden Units for ELM with Two-Stage Linear ModelIEICE Transactions on Information and Systems, 2008
- Extreme learning machine: Theory and applicationsNeurocomputing, 2006
- Evolutionary extreme learning machinePattern Recognition, 2005
- A HIERARCHICAL METHOD FOR FINDING OPTIMAL ARCHITECTURE AND WEIGHTS USING EVOLUTIONARY LEAST SQUARE BASED LEARNINGInternational Journal of Neural Systems, 2003
- Avoiding overfitting caused by noise using a uniform training mode2003
- Classification ability of single hidden layer feedforward neural networksIEEE Transactions on Neural Networks, 2000
- Upper bounds on the number of hidden neurons in feedforward networks with arbitrary bounded nonlinear activation functionsIEEE Transactions on Neural Networks, 1998
- Differential Evolution – A Simple and Efficient Heuristic for global Optimization over Continuous SpacesJournal of Global Optimization, 1997
- Improving the learning speed of 2-layer neural networks by choosing initial values of the adaptive weights1990