Classifier selection from a totally bounded class of functions

✍ Scribed by Majid Mojirsheibani

• Book ID: 104301622
• Publisher: Elsevier Science
• Year: 2001
• Tongue: English
• Weight: 125 KB
• Volume: 52
• Category: Article
• ISSN: 0167-7152
• DOI: 10.1016/s0167-7152(01)00005-0

✦ Synopsis

This article deals with the two-class classiÿcation problem, where the class conditional probability (x)=P{Y =1 | X=x} belongs to some known class of functions F. Given a data-based skeleton estimate Fn of the class F, with respect to the empirical L1-norm, we consider methods of constructing classiÿers using the members of the class Fn. Conditions under which the resulting classiÿcation rules are strongly Bayes consistent are also studied. The results are nonparametric and continue to hold regardless of the VC dimension of the corresponding class of classiÿers.