I'm a PhD student studying under Dr. Ross Beveridge .
My work focuses on understanding convolutional neural network (CNN) feature spaces. Most recently, we have learned that face recognition models converge to the same outputs after rotation, even when those models don't share training data or CNN architecture. This research has been applied in the DARPA Active Interpretation of Disparate Alternatives (AIDA) program.
Previously, I contributed to the DARPA Communicating with Computers (CwC) program, which focused on DIANA, an intelligent avatar described here .