The weighted k-nn with selection of features and its neural realization.

Włodzisław Duch and Karol Grudziński
Department of Informatics, Nicolaus Copernicus University,
Grudziadzka 5, 87-100 Torun, Poland.
E-mails: duch,

Fourth Conference on Neural Networks and Their Applications, Zakopane, May 1999, pp. 191-196

As a step towards neural realization of various similarity based algorithms k-nn method has been extended to weighted nearest neighbor scheme. Experiments show that for some datasets significant improvements are obtained. As an alternative to the minimization procedures a best-first search weighted nearest neighbor scheme has been implemented. A feature selection method for \knn, based on a variant of the best-first search strategy, has also been implemented. This method is relatively fast and for some databases gives excellent results. Finally a natural neural network extension of k-nn method is described, including weights and other parameters as a part of the model.

Paper in PDF, 86 KB

Projects on similar subject and BACK to the on-line publications of W. Duch.