Events Calendar

TK-MARS: An Efficient Approach for Deterministic and Stochastic Black-Box Optimization

Abstract
Simulators can be used to study, and potentially optimize many real-world complex systems that may either be computer simulators such as air quality, green building design, simulation model of a hospital emergency room, investment portfolio for financial planning or actual experiments like vehicle crash testing, and agricultural crop ideotype simulation models that are called black-box systems. 

Since these simulators are often computationally too expensive to be embedded directly within a global optimization method, a more practical approach would be to employ less computationally expensive surrogate models. Surrogate optimization for black-box systems requires careful selection of simulator runs, so as to simultaneously improve the surrogate model and gain more information on the optimum location in each iteration. Historically, surrogate optimization methods assume the complex system involves no uncertainty in performance and the set of important decision variables is known apriori. In the real world, both of these assumptions are violated. To this end, my research has introduced a new black-box optimization paradigm, to address uncertainty in system performance and feature selection for important decision variables. 

My research objective is to generalize the propelled discovery techniques to any large-scale black-box systems. More importantly, my examination in black-box optimization has the potential to contribute to the challenging problem of hyper-parameter tuning as fundamental progress towards democratizing machine learning.

Biography
Hadis Anahideh is a research scientist at American Airlines’ Advanced Analytics and Operations Research department. She completed her Ph.D. program at the University of Texas at Arlington (UTA) in 2018 where she worked in COSMOS with Dr. Jay Rosenberger and Dr. Victoria Chen. She is the recipient of Dissertation Fellowship Award at UTA. She holds a Master of Science degree in Industrial & Systems Engineering and Bachelor of Science degree in Applied Mathematics. Hadis has also done two internships with American Airlines Advanced Analytics and Operations Research department in 2015 and 2016. Her research interests include mathematical modeling, black-box optimization, predictive modeling, design of experiments, statistical and regression analysis and applied machine learning.

Monday, May 6 at 9:00am to 10:00am


Rogers Hall, 204
2000 SW Monroe Avenue, Corvallis, OR 97331

Event Type

Lecture or Presentation

Organization
College of Engineering, Mechanical, Industrial, and Manufacturing Engineering
Contact Name

Cristina Olson

Contact Email

cristina.olson@oregonstate.edu

Subscribe
Google Calendar iCal Outlook

Recent Activity