|
Activity Number:
|
423
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Learning and Data Mining
|
| Abstract - #304853 |
|
Title:
|
Multicategory Composite Least Squares Classifier
|
|
Author(s):
|
Seo Young Park*+ and Yufeng Liu
|
|
Companies:
|
The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
|
|
Address:
|
B04 Hanes Hall, Chapel Hill , NC, 27514,
|
|
Keywords:
|
Fisher Consistency ; Multicategory Classification ; SVM ; Regression
|
|
Abstract:
|
The Support Vector Machine (SVM) has been very successful in many applications. However, extension of the binary SVM to the multicategory classification problem is not trivial. Although there have been a number of versions multicategory SVM proposed, computation can be difficult for large scale problems. In this talk, we propose a new efficient multicategory composite least squares classifier (CLS classifier), which utilizes a composite squared loss as the loss function. We show Fisher consistency of our proposed method, and suggest a sensible method for class probability estimation. Our numerical examples demonstrate great performance of the proposed method.
|
