|
Activity Number:
|
139
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #308375 |
|
Title:
|
Consistent Learning Methods Are Approximately Local
|
|
Author(s):
|
Yaacov Ritov*+ and Alon Zakai
|
|
Companies:
|
The Hebrew University of Jerusalem and The Hebrew University of Jerusalem
|
|
Address:
|
Department of Statistics, Jerusalem, 91905, Israel
|
|
Keywords:
|
Nonparametric ; Classification ; SVM
|
|
Abstract:
|
We investigate the following learning-theoretical statement: We call a learning method local if, given a training set, it produces an estimate for a particular point that depends only on close-by points from the training set. We formally define this concept in several ways and show several theoretical results relating to those definitions, in particular that any consistent method is very close to being local.
|
