Jay H. Lee
Professor
Contact Information
Building: Ford ES&T
Office:
2202-2203
Phone: 404.385.2148
Fax: 404.894.2866
email
Mailing Address
Georgia Institute of Technology
School of Chemical &
Biomolecular Engineering
311 Ferst Drive, N.W.
Atlanta, GA 30332-0100
Links
LIDCUS (Laboratory for Information and Decisions for Complex and Uncertain Systems)
Jay H. Lee
Education
B.S. 1986, University of Washington
Ph.D. 1991, California Institute of Technology
Bio
Dr. Lee received the National Science Foundation's Young Investigator Award and a number of other research and teaching awards. He is also a co-author of the forthcoming book "Model Predictive Control." He is a member of AIChE, IEEE, and ASEE, has given a number of keynote talks at international conferences and participated in organizing several others.
Research Interests
Dr. Lee is currently Director of the LIDCUS (Laboratory for Information and Decisions for Complex and Uncertain Systems). His group is working on ways to use powerful computers, numerical optimization methods, information processing techniques, and novel sensors to improve the safety and efficiency of chemical and biological processes.
Currently, his group's research is focused on two topics.
Optimal control decisions in a Complex, Dynamic, and Uncertain Environment
Multi-stage stochastic decision-making represents an important class of problems as many interesting real-world control decision problems can be formulated within this framework. We study optimal decision-making for complex systems, particularly in the presence of quantifiable uncertainty.
Dr. Lee's group is developing algorithms that can evolve substantially suboptimal decision policies into optimal ones through interaction with the stochastic system, either in simulation or in real world. These algorithms are inspired by the way human brains learn to make decisions and are rooted in the classical concept of dynamic programming as well as the modern artificial intelligence (AI) concept of reinforcement learning.
Application areas include advanced control, real-time optimization, planning and scheduling, supply chain management, infrastructure investment, and other high-level strategic decision problems. In particular, in collaboration with Weyerhaeuser NR, he has recently begun working on optimal infrastructure design / operation of bio-refinery networks. In line with this, his group also investigates the 'value-of-information' to obtain systematic methods to obtain, store, and use information to combat against uncertainty.
Learning from High-Dimensional Data
With the advent of high-throughput sensing devices and sophisticated information systems, there is a plethora of data to be analyzed everywhere. This is particularly true in the area of bio-science and engineering where tremendous amounts of genomic, proteomic, and metabolomic data are being gathered through high-dimensional array devices. However, due to the typically very high dimension of these data, analysis of and extraction of knowledge from these data are non-trivial.
They are developing various machine-learning algorithms to mine data from high-dimensional, high-volume data. Applications include learning of critical and interacting residues in biocatalysts, identification of gene expression networks, and critical feature extraction from image data with particular applications to biopolymers.