The Sample Compression Framework in Machine Learning

March 12, 2009
Mohak Shah | McGill University

This talk will focus on the Sample Compression learning framework emphasizing some of its advantages over more conventional frameworks such as the VC learning paradigm. Moreover, unlike traditional VC and Rademacher based learning paradigms, we will show how practical realizable guarantees on the generalization performance of the learning algorithms can be obtained in this framework. We will also study examples of learning algorithms where such risk bounds can practically guide the model selection process yielding competitive results with the state-of-the-art learning algorithms as well as conventional re-sampling techniques such as the k-fold cross validation. Finally, we will show how some of the well known algorithms such as decision trees and SVM can be characterized within this framework, followed with a discussion on some open questions.

Speaker Details

Mohak Shah is a Postdoctoral Fellow at the Centre for Intelligent Machines, McGill University. His research interests span both theoretical machine learning and machine learning applications to bioinformatics and medical imaging. His current research, in collaboration with the Montreal Neurological Institute and Hospital, focuses on studying Multiple Sclerosis lesions in the human brain MRIs. Mohak obtained his PhD in Computer Science from the University of Ottawa in 2006. He was a Postdoctoral Fellow at the CHUL Genomics Research Centre in Quebec prior to joining McGill in 2008.