|
Activity Number:
|
210
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
ASA
|
| Abstract - #308148 |
|
Title:
|
Regularization Methods in Statistical Model-Building: Statisticians, Computer Scientists, Classification, and Machine Learning
|
|
Author(s):
|
Grace Wahba*+
|
|
Companies:
|
University of Wisconsin-Madison
|
|
Address:
|
Statistics Department, Madison, WI, 53705,
|
|
Keywords:
|
regularization ; cross validation ; model building ; optimization
|
|
Abstract:
|
We survey members of a broad class of statistical model building tools that are popular in nonparametric regression and classification, which have the common the feature that they involve an optimization problem with an explicit tradeoff between fit to the data and complexity of the model. We examine some relationships between Bayes estimates, penalized likelihood nonparametric regression methods, and the classification method known as a Support Vector Machine, in the context they share as regularization methods. Cross validation based methods for choosing the tradeoff (tuning) parameters will be examined, along with problems in selecting important variables and variable clusters. Interplay between Statisticians and Computer Scientists in extending this rich class of methods will be shown to be valuable to both.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2007 program |
