I am advised by the following unordered pair: {Geoff Gordon, André Platzer}. Currents interests are:
- complementarity problems
- convex optimization and analysis
- planning and MDPs
- learning theory and machine learning
- probabilistic inference
I am currently working on methods for incorporating
function approximation into solution methods for a broad class of
linear complementarity problems.
These complementarity problems are strongly connected to the Karush-Kuhn-Tucker conditions for linear and quadratic programs, but are more general.
Applications include approximating Markov decision processes and approximating support vector machines.
Approximate solution methods include projected-gradient descent, proximal methods, and interior point methods.
I am a member of the Logical Systems Lab and the SELECT Lab.
I have also been helping a CMU/JHU APL team out with verifying the FAA's new Airborne Collision Avoidance System (ACAS X), which is based on an MDP.
