|
Activity Number:
|
185
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #306199 |
|
Title:
|
Graph-Based Classifiers in Semi-Supervised Learning
|
|
Author(s):
|
George Michailidis*+
|
|
Companies:
|
University of Michigan
|
|
Address:
|
Department of Statistics, Ann Arbor, MI, 48109-1107,
|
|
Keywords:
|
semi-supervised learning ; sequential predictions ; smoother matrices
|
|
Abstract:
|
A graph-based, nearest-neighbor classifier is proposed for semi-supervised learning, which utilizes the labeled data and topology of the graph for training and produces proper probability class estimates. In addition, a sequential procedure is developed for classifying unlabeled nodes. The procedure is iterative in nature and uses the topology of the graph and the cotraining information available in labeled and unlabeled data. The performance of the proposed classifier is assessed on several synthetic and real data. Extensions to settings where the available information comes from several domains modeled by graphs also are considered.
|
- 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 2006 program |
