[Lisa_teatalk] Tea Talk Tomorrow!

Aaron Courville aaron.courville at gmail.com
Wed Nov 3 12:53:47 EDT 2010


Hi All,

We have a tea talk tomorrow by Ian and Yann on their AISTAT submission.

Time/Date: 14h30, Thursday Nov. 4th.
Location: LISA Lab (AA3256)

A Probabilistic Framework for Multi-Instance Learning

We present a probabilistic framework for multi-instance learning. Our
framework, PMIL, allows a wide variety of supervised learning models
to be easily adapted to the multi-instance setting. We demonstrate the
effectiveness of our approach on real-world benchmark datasets,
achieving state of the art peformance on several TREC tasks. We use
controlled experiments with artificially constructed multi-instance
problems in order to demonstrate the ability of our approach to learn
a good model of individual instances and to empirically analyze its
behavior under conditions that violate the assumptions justifying our
framework.

Cheers,
Aaron

-- 
Aaron C. Courville
Département d’Informatique et
de recherche opérationnelle
Université de Montréal
email:Aaron.Courville at gmail.com


More information about the Lisa_teatalk mailing list