We present a sample re-weighting scheme inspired by recent results in margin theory. The basic idea is to add to the training set replicas of samples which are not classified with a sufficient margin. We prove the convergence of the input distribution obtained in this way. As study case, we consider an instance of the scheme involving a 1-NN classifier implementing a Vector Quantization algorithm that accommodates tangent distance models. The tangent distance models created in this way have shown a significant improvement in generalization power with respect to the standard tangent models. More-over, the obtained models were able to outperform state of the art algorithms, such as SVM.

A Simple Additive Re-weighting Strategy for Improving Margins

AIOLLI, FABIO;SPERDUTI, ALESSANDRO
2001

Abstract

We present a sample re-weighting scheme inspired by recent results in margin theory. The basic idea is to add to the training set replicas of samples which are not classified with a sufficient margin. We prove the convergence of the input distribution obtained in this way. As study case, we consider an instance of the scheme involving a 1-NN classifier implementing a Vector Quantization algorithm that accommodates tangent distance models. The tangent distance models created in this way have shown a significant improvement in generalization power with respect to the standard tangent models. More-over, the obtained models were able to outperform state of the art algorithms, such as SVM.
Proceedings of the Seventeenth International Joint Conference on Artificial Intelligence, IJCAI 2001
1558608125
9781558608122
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11577/2474201
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact