Least third-order cumulant method with adaptive regularization parameter selection for neural networks

Artificial Intelligence

Volumn 127, Issue 2, 2001, Pages 169-197

  Leung, Chi Tat ^a   Chow, Tommy W S ^b  

^a HONG KONG POLYTECHNIC UNIVERSITY   (Hong Kong)

^b CITY UNIVERSITY OF HONG KONG   (Hong Kong)

Author keywords

Generalization capability; Neural networks; Regularization; Third order cumulant

Indexed keywords

CONVERGENCE OF NUMERICAL METHODS; ERROR ANALYSIS; NEURAL NETWORKS; REGRESSION ANALYSIS; VECTORS;

LEAST THIRD-ORDER CUMULANT METHOD; MEAN SQUARE ERRORS;

ARTIFICIAL INTELLIGENCE;

EID: 0035310522     PISSN: 00043702     EISSN: None     Source Type: Journal    
DOI: 10.1016/S0004-3702(01)00061-3     Document Type: Article

References (19)

1. Hints
2. First- and second-order methods for learning: Between steepest descent and Newton's method
3. Neural network based short-term load forecasting using weather compensation
4. A nonlinear autoregressive integrated neural network model for short-term load forecasting
5. Training multilayer neural networks using fast global learning algorithm - Least squares and penalized optimization methods
6. Unsupervised mutual information criterion for elimination of overtraining in supervised multilayer networks
7. Optimization of network structure using genetic techniques
8. Learning translation invariant recognition in massively parallel networks
9. Comparing biases for minimal network construction with back-propagation
10. A simple weight decay can improve generalization
11. A least third-order cumulants objective function
12. Adaptive regularization parameter selection method for enhancing generalization capability of neural networks
13. Global optimization for neural network training
14. Generalization by weight elimination with application to forecasting