Optimal Learning and Real-time Decisions under Imperfect Information


                Department of Systems Engineering and Engineering Management
                             The Chinese University of Hong Kong

Date: 11:00am - 12:30pm, December 15 (Tuesday), 2015

Title: Optimal Learning and Real-time Decisions under Imperfect Information

Speaker: Dr. Wang Jue, Queen’s School of Business


Sensors and streaming data are revolutionizing many industries, and competitive advantage is increasingly tied to the capabilities of monitoring, control, and optimization in real time. However, most streaming data is not perfect and involves uncertainty. Online learning and decision making under such imperfect information becomes a pressing challenge, however the existing decision theory is generally not scalable to handle large and complex problems.

As an example, if we observe customer behaviour over time, customers need to be assigned to segments in real time. The optimal solution method, known as Wald’s sequential probability ratio test (SPRT), is applicable to only two segments. For multiple segments, the optimal solution is widely believed intractable due to the “curse of dimensionality” in dynamic programming.

We develop a new solution method called “belief-state reconstruction” that can circumvent the curse of dimensionality in many practical cases, making the optimal solution scalable to large problems. We show that, rather surprisingly, some well-known "high-dimensional" problems are in fact low dimensional in nature. We describe how to uncover the intrinsic dimensionality of the problem on three representative applications in customer segmentation, natural resource exploration, and postmarketing drug surveillance.


Jue Wang is a postdoctoral fellow at Queen’s School of Business, where he conducts both methodological research at the intersection of stochastic models, statistics and optimal control theory, as well as applied research in the area of revenue management, quality and reliability engineering. He has involved in various research projects with GE Healthcare, Walmart, Syncrude Canada and China Southern Power Grid, and taught courses in stochastic processes and Markov decision processes at the graduate level. He received his Ph.D from the University of Toronto, with a thesis on Bayesian control of partially observable stochastic processes.


Everyone is welcome to attend the talk!

Venue: Room 513,
      William M.W. Mong Engineering Building (ERB),
      (Engineering Building Complex Phase 2)
      The Chinese University of Hong Kong.

The talk will be hosted by:
Prof. Helen Meng,
Department of Systems Engineering and Engineering Management,
The Chinese University of Hong Kong,

SEEM-5201 Website:

Tuesday, December 15, 2015 - 03:00 to 04:30