UTIAS Seminar Series



Loading Events

« All Events

  • This event has passed.

Bayesian Sequential Optimal Experimental Design

June 8 @ 1:10 pm - 2:10 pm

Speaker: Xun Huan

Date & time: Thursday, June 8th, 1pm

Location: UTIAS Lecture Hall

Title: Bayesian Sequential Optimal Experimental Design


Experiments are crucial for developing and refining models in engineering and science. When experiments are expensive, a careful design of these limited data-acquisition opportunities can be immensely beneficial. Optimal experimental design (OED) thus leverages the predictive power of simulation models to systematically quantify and maximize the value of experiments.

We first introduce OED for a batch of experiment under a Bayesian setting using the expected information gain (EIG) (uncertainty reduction) objective. We then formulate OED for a sequence of experiments via a Markov decision process (MDP), where an optimal design rule (policy) can (a) adapt to newly collected data along the way (feedback) and (b) anticipate future consequences (lookahead). We solve the sequential OED problem with policy gradient techniques from reinforcement learning together with an efficient lower bound EIG estimator. This is achieved numerically by directly parameterizing the policy, value function, and variational posteriors using neural networks and improving them via gradient estimates produced from simulated design sequences. We demonstrate our method on several examples, including an optimal sensor movement application for source inversion in a convection-diffusion field.


Xun Huan is an Assistant Professor of Mechanical Engineering at the University of Michigan, where he leads the Uncertainty Quantification and Scientific Machine Learning Group. He is affiliated with the Michigan Institutes for Computational Discovery and Engineering (MICDE) and for Data Science (MIDAS). Xun received a B.A.Sc. (Engineering Science–Aerospace) from the University of Toronto, a S.M (Aerospace Engineering) and Ph.D. (Computational Science and Engineering) from MIT, and was a postdoctoral researcher at MIT and Sandia National Laboratories. His research interests include optimal experimental design, Bayesian analysis, reinforcement learning, and physics-aware data-driven modeling.


June 8
1:10 pm - 2:10 pm


Lecture Hall, University of Toronto Institute for Aerospace Studies