|
Activity Number:
|
124
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #308117 |
|
Title:
|
The Lasso with Attribute Partition Search
|
|
Author(s):
|
Suhrid Balakrishnan*+ and David Madigan
|
|
Companies:
|
Rutgers University and Rutgers University
|
|
Address:
|
c/o DIMACS, Rutgers Univ., Piscataway, NJ, 08854,
|
|
Keywords:
|
Group Lasso ; Partial Exchangeability ; Time Series ; Fused Lasso
|
|
Abstract:
|
Regression and classification problems involving ordered attributes (for example where some input patterns are a set of samples from time series variables) arise in application domains like finance and epidemiology. In such cases, identifying and building models involving predictive runs of the attributes leads to highly interpretable models that may also be very accurate. We present an approach to build such models using a variant of the Group Lasso (Yuan and Lin, 2006).
|
