I am a Ph.D. candidate at Tufts University researching how to use machine learning theory to create new deep learning methods with theoretic guarantees. My work spans traditional and deep computer vision, supervised learning, semi-supervised learning, generalized category discovery, and novelty detection.
I am currently exploring theoretically motivated methods for leveraging unlabeled data to improve supervised learning. This includes semi-supervised learning, which assumes unlabeled data matches the supervised task, and robust semi-supervised learning, which can use less refined unlabeled data by filtering out data that does not match the task. Generalized category discovery, which aims to cluster the data robust semi-supervised learning would filter out, is also of interest.