Using Group Information to Improve Generalization

Speaker:	Professor Vladimir Cherkassky
		University of Minnesota
		USA

Title:		"Using Group Information to Improve Generalization"

Date:		Friday, 30 May 2008

Time:		4:00pm - 5:00pm

Venue:		Room 3408 (via lifts 17/18), HKUST)

Abstract:

This talk describes several new learning settings that are appropriate for
estimating predictive models with sparse heterogeneous data. The general
approach follows methodological framework of VC-theory (Vapnik 2006) which
emphasizes the role of a priori knowledge about available data, as opposed
to knowledge about good (or true) model used in traditional machine
learning approaches. In this talk, I will discuss a particular form of a
priori knowledge when the training data is known to belong to several
disjoint groups. This includes 'structured data' in (Vapnik, 2006) and
several Multi-Task Learning approaches in Machine learning. Incorporating
such a group information (about the data) into the learning process leads
to several new learning settings such as Learning with Structured Data or
SVM+ (Vapnik 2006) and SVM+ Multi-Task Learning (Liang and Cherkassky
2008).

This talk will introduce SVM+ and SVM+MTL settings and show several
empirical comparisons illustrating:

1. Different ways of incorporating group information into a learning
formulation

2. Empirical comparisons illustrating advantages and limitations of
different learning formulations utilizing the group information in
different ways. Comparisons are presented using both synthetic and
real-life data (fMRI data and Heart data set).


*******************
Biography:

Vladimir Cherkassky is a Professor of Electrical and Computer Engineering
at the University of Minnesota, Twin Cities .  He received Ph.D. in
Electrical Engineering from University of Texas at Austin in 1985. His
current research is on methods for predictive learning from data, and he
has co-authored a monograph Learning From Data published by Wiley in 1998
and in 2007 (second revised edition). Prof. Cherkassky has served on the
Governing Board of INNS. He has served on editorial boards of IEEE
Transactions on Neural Networks, the Neural Networks Journal, the Natural
Computing Journal and the Neural Processing Letters. He served on the
program committee of all major international conferences on Artificial
Neural Networks. He was Director of NATO Advanced Study Institute (ASI)
From Statistics to Neural Networks: Theory and Pattern Recognition
Applications held in France, in 1993. He presented numerous invited talks
and tutorials on neural network and statistical methods for learning from
data. He was elected in 2007 as Fellow of IEEE for 'contribution and
leadership in statistical learning and neural networks'.