Asymptotical Optimality of Adaptive Nearest Neighbour Discrimination

This paper has discovered that the commonly used k-N N discrimination is not Bayesian risk consistent. A new adaptive discrimination procedure has been proposed. It has been proved that such procedure is Bayesian risk consistent, i.e. it has asymptotical optimality. Finally, an open problem is posed...

Full description

Saved in:
Bibliographic Details
Published inLecture notes-monograph series Vol. 24; pp. 149 - 162
Main Authors Gui-Jing Chen, Wang, Y. H.
Format Journal Article
LanguageEnglish
Published Institute of Mathematical Statistics 01.01.1994
Subjects
Ballots
Conditional probabilities
Invited Papers
Mathematical inequalities
Random variables
Rectangles
Online AccessGet full text

Cover

More Information
Summary:This paper has discovered that the commonly used k-N N discrimination is not Bayesian risk consistent. A new adaptive discrimination procedure has been proposed. It has been proved that such procedure is Bayesian risk consistent, i.e. it has asymptotical optimality. Finally, an open problem is posed.
ISSN:0749-2170
2328-3874