Deep Robust & Explainable AI Lab
Deep-REAL is a research lab affiliated with the Department of Computer and Information Sciences and Data Science Institute, at the University of Delaware. We work in the area of Deep Learning, Machine Learning, and Computer Vision. Our mission is to develop flexible, reliable, and explainable machine learning models, upon which cross-disciplinary research (biomechanics, geoscience, bioinformatics) can advance synergistically. Specially, we consider three fundamental data and learning challenges:
How to learn comprehensive ML models (e.g. DNNs) given extreme data limitations (e.g. few-shot without labels)?
How to make trustworthy decisions on evolving domains (e.g. out-of-domain distributions)?
How to quantify the risk of uncertainty as well as potential cause-effects in the above scenarios?