Congratulations to Dr. Dhruva Patil for defending his Ph.D. dissertation and to Jason Garcia for defending his Master’s thesis!
Dhruva’s dissertation is entitled Something Is Fishy! How ambiguous language affects generalization of video action recognition networks. In this research, he analyses weaknesses in the label set of the popular Something-Something video action recognition dataset, and discusses how the one-hot encoding of class labels in computer vision tasks may effectively be hamstringing deep neural networks in vision tasks. Dhruva will be starting a research position at Amazon in June.
Jason’s thesis is entitled Applications of Topological Data Analysis to Natural Language Processing and Computer Vision. Jason applies techniques of topological data analysis, a mathematical technique for exploring the “shape” of data, to problems in NLP and computer vision, including a novel analysis of the topological properties of the encoder layers in BERT. Jason will be continuing at CSU for a Ph.D. in Math.
Congratulations to Dhruva and Jason! It’s been a pleasure to work with you and to see you acheive your degrees!
(X-posted on signallab.ai)