SVD-NET: an algorithm that automatically selects network structure
- 1 May 1994
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks
- Vol. 5 (3), 513-515
- https://doi.org/10.1109/72.286929
Abstract
An algorithm is developed for training feedforward neural networks that uses singular value decomposition (SVD) to identify and eliminate redundant hidden nodes. Minimizing redundancy gives smaller networks, producing models that generalize better and thus eliminate the need of using cross-validation to avoid overfitting. The method is demonstrated by modeling a chemical reactor.Keywords
This publication has 6 references indexed in Scilit:
- PLS/neural networksComputers & Chemical Engineering, 1992
- Nonlinear PLS modeling using neural networksComputers & Chemical Engineering, 1992
- Direct and indirect model based control using artificial neural networksIndustrial & Engineering Chemistry Research, 1991
- A simple procedure for pruning back-propagation trained neural networksIEEE Transactions on Neural Networks, 1990
- Improvement of the backpropagation algorithm for training neural networksComputers & Chemical Engineering, 1990
- The use of biased least-squares estimators for parameters in discrete-time pulse-response modelsIndustrial & Engineering Chemistry Research, 1988