|
Activity Number:
|
15
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #307406 |
|
Title:
|
Improving Classification When a Class Hierarchy Is Available Using a Hierarchy-Based Prior
|
|
Author(s):
|
Babak Shahbaba*+ and Radford Neal
|
|
Companies:
|
University of Toronto and University of Toronto
|
|
Address:
|
55 Ellerslie Ave., Toronto, ON, M2N 1X9, Canada
|
|
Keywords:
|
Bayesian models ; hierarchical classification ; multinomial logistic regression
|
|
Abstract:
|
We introduce a new method for building classification models when we have prior knowledge of how the classes can be arranged in a hierarchy, based on how easily they can be distinguished. The new method uses a Bayesian form of the multinomial logit (MNL) model, with a prior that introduces correlations between the parameters for classes nearby in the tree. We compare the performance of simulated data on the new method, the ordinary MNL model, and a model that uses the hierarchy in a different way. We also test the new method on page layout analysis and document classification problems and find it performs better than the other methods.
|
