Pascal PoupartResearch interests

Professor Poupart's research interests are in artificial intelligence and more precisely in the areas of reasoning under uncertainty and machine learning, with application to health informatics and natural language processing. Professor Poupart's research focuses on the development of intelligent systems for sequential decision making under uncertainty. In particular, he developed new scalable algorithms for partially observable Markov decision processes (POMDPs), a mathematical framework to optimize non-trivial sequences of decisions while learning and adapting to an uncertain environment. In collaboration with Intel, his team is developing algorithms to optimize sequential decision processes and automatically explain them to users in the context of preventive maintenance. In collaboration with the UW-Schlegel Research Institute for Aging, the Village of Winston Park (retirement community in Kitchener, Ontario) and several colleagues at the University of Waterloo, he is designing intelligent walkers to monitor and assist users. In this project, Dr. Poupart's team focuses on the development of algorithms for feature extraction, pattern recognition, information fusion, activity recognition and computer vision. In another project Dr. Poupart's team is also developing a wearable sensor system to monitor and assess the symptoms of Alzheimer's disease. The goal is to develop composite behavioural markers of Alzheimer's disease based on voice, location and motion data collected during the day and at night. In another line of research, Dr. Poupart is working with Google to automatically detect and label the topic of web documents using statistical and natural language processing techniques.

Degrees and awards

BSc (McGill), MSc (British Columbia), PhD (Toronto)

Best Paper Award Runner Up, International Conference on Uncertainty in Artificial Intelligence (2008); Google Research Award (2008); Early Researcher Award, Government of Ontario (2008-2011); Google Research Award (2007); Best Paper Award, International Conference on Computer Vision Systems (2007); Best Paper Award, Conference on Privacy, Security and Trust (2005)

Industrial and sabbatical experience

In the summers of 1996 and 1997, Professor Poupart held internships at Cogni-CASE in Montreal. As part of the research and development team, he developed software to correct legacy code suffering from the year 2000 bug. In the summer of 1998, Professor Poupart also did an internship in the research department at Lockheed Martin in Montreal. He helped develop and test a simulation platform for multi-agent systems.

Representative publications

F. Omar, M. Sinn, J. Truszkowski, P. Poupart, J. Tung and A. Caine. Comparative Analysis of Probabilistic Models for Activity Recognition with an Instrumented Walker. Uncertainty in Artificial Intelligence (UAI), Catalina, CA, 2010

J. Hoey, P. Poupart, A. von Bertoldi, T. Craig, C. Boutilier and A. Mihailidis. Automated Handwashing Assistance for Persons with Dementia Using Video and a Partially Observable Markov Decision Process. Computer Vision and Image Understanding (CVIU), Volume 114, Issue 5, Pages 503-519, May 2010

O.Z. Khan, P. Poupart and J. Black. Minimal Sufficient Explanations for Factored Markov Decision Processes. International Conference on Automated Planning and Scheduling (ICAPS), Thessaloniki, Greece, 2009.

M. Toussaint, L. Charlin and P. Poupart. Hierarchical POMDP Controller Optimization by Likelihood Optimization. Proceedings of the 24th Conference on Uncertainty in Artificial Intelligence (UAI), 2008. Best Paper Award Runner-Up.

J. Hoey, A. von Bertoldi, P. Poupart, and A. Mihailidis. Assisting Persons with Dementia during Handwashing Using a Partially Observable Markov Decision Process. Proceedings of the International Conference on Vision Systems (ICVS), 2007. Best Paper Award.

J. M. Porta, N. Vlassis, M.T.J. Spaan, and P. Poupart. Point-Based Value Iteration for Continuous POMDPs. Journal of Machine Learning Research (JMLR), 7:2329-2367, 2006.

P. Poupart, N. Vlassis, J. Hoey, and K. Regan. An Analytic Solution to Discrete Bayesian Reinforcement Learning. Proceedings of the 23rd International Conference on Machine Learning (ICML), pp. 697-704, 2006.

P. Poupart and C. Boutilier. VDCBPI: an Approximate Scalable Algorithm for Large Scale POMDPs. Advances in Neural Information Processing Systems 17 (NIPS), pp. 1081-1088, 2004.

University of Waterloo
Contact information: 

Profiles by type