A Talk On: Data-driven algorithms in control and their numerical stability

#Dynamic #Mode #Decomposition #(DMD) #Projection #onto #latent #Structures #(SIMPLS) #Empirical #Models #Data-Driven
Share

For systems of high complexity and dimensionality, building models empirically rather than analytically is much more practical, especially when high definition data is readily available. Therefore, advancing the theoretical foundation for data-driven methods in dynamical systems theory and control has become increasingly important. Dynamic Mode Decomposition (DMD) and Projection onto latent Structures (SIMPLS) are some of the popular algorithms which learn dynamic behavior from large volumes of data. Approaches to improve the numerical stability and robustness of these algorithms against noise and missing data will be discussed. The application of these methods in traffic flow predictions in Georgia’s diverging diamond interchanges will also be presented. The speaker will also briefly discuss how to build a research program in a primarily undergraduate institution, since this might be of interest to graduate students who are considering faculty positions after graduation.



  Date and Time

  Location

  Hosts

  Registration



  • Date: 04 Dec 2020
  • Time: 01:00 PM to 02:00 PM
  • All times are (GMT-05:00) US/Eastern
  • Add_To_Calendar_icon Add Event to Calendar
If you are not a robot, please complete the ReCAPTCHA to display virtual attendance info.
  • Burlington, Vermont
  • United States

  • Contact Event Host


  Speakers

Dr. Makhin Thitsa Dr. Makhin Thitsa of Mercer University

Topic:

Data-driven algorithms in control and their numerical stability

For systems of high complexity and dimensionality, building models empirically rather than analytically is much more practical, especially when high definition data is readily available. Therefore, advancing the theoretical foundation for data-driven methods in dynamical systems theory and control has become increasingly important. Dynamic Mode Decomposition (DMD) and Projection onto latent Structures (SIMPLS) are some of the popular algorithms which learn dynamic behavior from large volumes of data. Approaches to improve the numerical stability and robustness of these algorithms against noise and missing data will be discussed. The application of these methods in traffic flow predictions in Georgia’s diverging diamond interchanges will also be presented. The speaker will also briefly discuss how to build a research program in a primarily undergraduate institution, since this might be of interest to graduate students who are considering faculty positions after graduation.

Biography:

Dr. Makhin Thitsa is an Associate Professor in the Mercer University School of Engineering. She received her B.S. in Electrical Engineering in 2005, her M.S in Electrical Engineering in 2008, and her Ph.D. in Electrical and Computer Engineering in 2011, all from Old Dominion University. She was then a Research Assistant Professor in the Department of Electrical and Computer Engineering at Old Dominion University until 2013, when she joined the faculty of Mercer University as an Assistant Professor in the School of Engineering. Her research interests include nonlinear systems and control theory, model-free control, and data-driven control strategies. She has successfully applied control methods to photonic devices, unmanned aerial vehicles, and traffic flow networks and her research has attracted over $200,000 in external funding. In addition to her teaching and scholarship at the university, she also holds a research affiliate position at the Georgia Institute of Technology. As the director of the Cyber-physical Systems and Control Laboratory at Mercer University School of Engineering, she has mentored a large number of undergraduate students, including four who have been selected to receive a prestigious Barry M. Goldwater Scholarship. For her research efforts and service to Mercer University community Dr. Thitsa has earned the 2020-21 All-Southern Conference Faculty Member Award for Mercer University.