CAST Seminar: Sebastian Jaimungal

Join us at the CANSSI Ontario STatistics Seminars (CAST) with

Sebastian Jaimungal

Professor and Chair
Department of Statistical Sciences
University of Toronto
Associate Member and Visiting Academic
Oxford-Man Institute for Quantitative Finance
University of Oxford

Free Event | Registration Required

Talk Title

Projections and Barycentres of Stochastic Processes

Abstract

This talk presents approaches for combining probabilistic models of stochastic processes through the projection of models onto constraint sets. We investigate the problem of an agent seeking a probability measure that best reconciles multiple expert models while satisfying prescribed constraints. The optimal model is obtained by minimizing an appropriate divergence subject to these constraints, and we provide explicit characterizations of the resulting optimal dynamics. The framework naturally accommodates a wide range of applications. For example, in climate modeling, multiple stochastic models produced by different research groups or Earth system components can be aggregated while enforcing regulatory or policy-driven constraints, such as requiring that the expected global mean temperature increase to be 3 °C over a 50-year horizon, or that temperature variance remains within prescribed bounds. These constraints encode physical, regulatory, or societal objectives and are imposed directly on the admissible probability measures. I will present how to aggregate expert diffusions via weighted Kullback–Leibler divergence, and discuss optimal risk sharing under model ambiguity using chi-squared divergence. Time permitting, I will discuss deep learning algorithms that efficiently approximate the optimal constrained dynamics and demonstrate applications spanning finance, insurance, and climate-risk modeling, highlighting how principled model combination can reconcile expert disagreement while respecting externally imposed constraints.

Speaker Profile

Dr. Jaimungal is a Full Professor and the current Chair of the Department of Statistical Sciences at the University of Toronto. He is also an Associate Member and Visiting Academic at the Oxford-Man Institute for Quantitative Finance, University of Oxford, and a Fellow of the Fields Institute for Research in Mathematical Sciences. He previously served as Chair of the SIAM Activity Group on Financial Mathematics and currently sits on the editorial boards of the ASA Data Science in Science, SIAM Journal on Financial Mathematics, Quantitative Finance, and the Journal of Dynamics and Games.

Professor Jaimungal’s research lies at the intersection of probability theory, optimization, and data-driven decision making. His work spans reinforcement learning, stochastic optimal control, mean-field games, the physics of AI and their applications to risk management, clean energy markets, and market microstructure, among others. He is a coauthor of the graduate textbook “Algorithmic and High Frequency Trading” published by Cambridge University Press which introduces stochastic control theory and its application to trading.

A unifying theme across his research is the development of mathematically principled models that bridge theory and practice, particularly in environments characterized by uncertainty, strategic interaction, and feedback effects. His work has appeared in leading journals across applied mathematics, finance, and machine learning, and he is a frequent invited speaker at international conferences and workshops.