A PDE Approach to a Prediction Problem Involving Randomized Strategies

Monday, October 8, 2018 - 1:25pm - 2:25pm
Lind 305
Nadejda Drenska (University of Minnesota, Twin Cities)
This work investigates a classical problem from online machine learning using methods from optimal control theory. The problem is a discrete time iterative process involving decision making at every step; the goal for mathematical analysis is to understand the optimal strategy and its consequences over a long period of time. The solution is analyzed through its continuous limit--an appropriately defined value function, which solves a PDE in the viscosity sense. The PDE is then used to determine the optimal strategies.

Nadejda went to Brown University as an undergraduate. She graduated with her Ph.D. in mathematics from the Courant Institute (NYU) with an adviser Robert Kohn. Nadejda is now a MCFAM postdoc at the University of Minnesota.