Improved Memory-Based Classification

Wlodzislaw Duch, Rafal Adamczak and Norbert Jankowski.
Department of Computer Methods, Nicholas Copernicus University, 
Grudziadzka 5, 87-100 Torun, Poland.
e-mail: duch,raad,norbert@phys.uni.torun.pl

Abstract: Memory-based methods and memory-based reasoning are powerful
classification methods but for a large number of data items they are quite
expensive. The number of memorized data items is reduced if some fuzziness
in the representation of the data is introduced. Memory-based techniques are
extended to deal with non-point objects in feature spaces. The effectiveness
of this approach is demonstrated in a number of examples.

in: Engineering Applications of Neural Networks, London 17-19.06.1996