Title :
Adaptive regularization of neural classifiers
Author :
Andersen, L. Nonboe ; Larsen, J. ; Hansen, L.K. ; Hintz-Madsen, M.
Author_Institution :
Dept. of Math. Modelling, Tech. Univ., Lyngby, Denmark
Abstract :
We present a regularization scheme which iteratively adapts the regularization parameters by minimizing the validation error. It is suggested to use the adaptive regularization scheme in conjunction with optimal brain damage pruning to optimize the architecture and to avoid overfitting. Furthermore, we propose an improved neural classification architecture eliminating an inherent redundancy in the widely used SoftMax classification network. Numerical results demonstrate the viability of the method
Keywords :
adaptive signal processing; iterative methods; neural net architecture; pattern classification; SoftMax classification network; adaptive regularization; architecture optimization; iterative adaptation; neural classification architecture; neural classifiers; optimal brain damage pruning; validation error minimization; Assembly; Bayesian methods; Biological neural networks; Feedforward systems; Mathematical model; Neurons; Pattern recognition; Probability; Redundancy; World Wide Web;
Conference_Titel :
Neural Networks for Signal Processing [1997] VII. Proceedings of the 1997 IEEE Workshop
Conference_Location :
Amelia Island, FL
Print_ISBN :
0-7803-4256-9
DOI :
10.1109/NNSP.1997.622380
