Past Event! Note: this event has already taken place.
Learning through the Lens of Compression
November 26, 2020 at 1:00 PM to 2:00 PM
Data Science Distinguished Speaker Seminar Series
Speaker: Dr. Hassan Ashtiani
Abstract
Characterizing the sample complexity of different machine learning tasks is one of the central questions in statistical learning theory. The classic Vapnik-Chervonenkis theory — which characterizes the sample complexity of binary classification — is arguably the most important result in learning theory. Despite this early progress, the sample complexity of many important learning tasks — including density estimation and adversarial learning — are not yet resolved. In this talk, we review the less conventional approach of using compression schemes for proving sample complexity bounds, proving new results for adversarial learning and density estimation. In particular, we will go through the journey of settling the sample complexity of learning Gaussian Mixture Models.
About the Speaker
Hassan Ashtiani is an Assistant Professor in the Department of Computing and Software at McMaster University. He obtained his Ph.D. in Computer Science from University of Waterloo in 2018. Before that, he received his master’s degree in AI and Robotics, and his bachelor’s degree in Computer Engineering, both from University of Tehran. He is interested in a broad range of problems in the intersection of theoretical and applied machine learning. In recent years, he has particularly focused on building sample efficient learning methods for unsupervised learning problems including distribution learning and clustering that are robust to (i) model mis-specificaiton, (ii) adversarial attacks, and/or (iii) distribution shift. He is a recipient of a NeurIPS’18 best paper award.
Seminar Moderator:
Dr. Tracey P. Lauriault, Associate Professor, Critical Media and Big Data, School of Journalism and Communication, Carleton University
Zoom webinar information for the Data Science Distinguished Speaker Seminar Series is sent to our mailing list (sign up here). If you are not on our mailing list and would like to attend this virtual seminar, please e-mail cuids@carleton.ca.