Developing New Optimization Algorithms and Applying them to Difficult Multimodal Environmental Problems
Listen to presentation here for Q&A
Speaker: Christine Shoemaker, Winner of 2023 Hotelling Medal from INFORMS
Bio: Prof. Shoemaker received her PhD in (pure) Mathematics supervised by Richard Bellman at University of Southern California. Her goals are a) to develop efficient general purpose optimization algorithms with a current focus on fast global optimization of expensive objective functions , b) to calibrate parameters in complex environmental models to improve predictive capability and c) to optimize decisions and costs related to environmental management and water resources. She has been Chaired Professor at Cornell University and at National University of Singapore. She is a joint author (with Eriksson and Bindel) of PYSOT (surrogate global optimization toolbox) that has been downloaded over 140,000 times. She is a member of U.S. National Academy of Engineering and elected Fellow in INFORMS, SIAM, and ASCE. She has won the 2023 SIAM Career Prize in Geoscience as well as the 2023 INFORMS Hotelling Medal.
Abstract: Many environmental problems are difficult to analyze because the a) models describing them may have many parameters to calibrate or decisions to make and the resulting functions are not convex and b) the objective function is often a simulation model that is expensive (minutes or hours) for each evaluation. This seminar will focus on Prof. Shoemaker’s recent research using RBF surrogate global optimization of expensive multimodal functions and its application to environmental model calibration or decision making. The algorithms (serial or parallel) are general purpose and can be used on expensive, multimodal objective functions (including for simulation optimization unrelated to the environment). Her earlier research (on groundwater, acid rain, and pesticide reduction) and her international organizational efforts to protect groundwater from pollution will also be discussed briefly.