In this talk, I will present our recent research on a new learning paradigm of deep learning by AUC maximization (including both AUROC and AUPRC). I will present a new surrogate loss for AUROC and non-convex min-max optimization algorithms for solving deep AUROC maximization problem. Then, I will present our recent work on non-convex optimization algorithms for solving deep AUPRC maximization problem. I will also talk about their applications in medical image classification and molecular property prediction for drug discovery. In particular, our deep AUC maximization is currently ranked at the 1st place at Stanford CheXpert competition and also the 1st place at MIT AICURES challenge.
Dr. Yang is an associate Professor of Computer Science at the University of Iowa. His research interests center round optimization, machine learning and AI. He received the best student paper award at COLT in 2012, NSF Career Award in 2019, and was named Dean’s Excellence in Research Scholar. Recently, his group has achieved the 1st place at Stanford CheXpert competition, and the 1st Place at MIT AICURES Challenge with collaborators. He has published more than 100 papers at top tier conferences such as ICML, NeurIPS. He serves as an associate editor of Neurocomputing, senior PC of AAAI and IJCAI, and AC of ICML and NeurIPS.