Ho-Kashyap Classifier with Early Stopping for Regularization
| Autor: | Fabien Lauer, Gérard Bloch |
|---|---|
| Přispěvatelé: | Centre de Recherche en Automatique de Nancy (CRAN), Université Henri Poincaré - Nancy 1 (UHP)-Institut National Polytechnique de Lorraine (INPL)-Centre National de la Recherche Scientifique (CNRS) |
| Jazyk: | angličtina |
| Rok vydání: | 2006 |
| Předmět: |
Generalization
Regularization perspectives on support vector machines Linear classifier 02 engineering and technology Overfitting Regularization (mathematics) [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG] Artificial Intelligence 020204 information systems Stopping time Learning rule 0202 electrical engineering electronic engineering information engineering Robustness Mathematics Early stopping business.industry Pattern recognition Ho-Kashyap classifier Support vector machine Classifier design Signal Processing 020201 artificial intelligence & image processing Computer Vision and Pattern Recognition Artificial intelligence business Software |
| Zdroj: | Pattern Recognition Letters Pattern Recognition Letters, Elsevier, 2006, 27 (9), pp.1037-1044. ⟨10.1016/j.patrec.2005.12.009⟩ |
| ISSN: | 0167-8655 |
| Popis: | 17 pages; International audience; This paper focuses on linear classification using a fast and simple algorithm known as the Ho-Kashyap learning rule (HK). In order to avoid overfitting and instead of adding a regularization parameter in the criterion, early stopping is introduced as a regularization method for HK learning, which becomes HKES (Ho-Kashyap with Early Stopping). Furthermore, an automatic procedure, based on generalization error estimation, is proposed to tune the stopping time. The method is then tested and compared to others (including SVM and LSVM), that use either $\ell_1$ or $\ell_2$-norm of the errors, on well-known benchmarks. The results show the limits of the early stopping for regularization with respect to the generalization error estimation and the drawbacks of low level hyperparameters such as a number of iterations. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect