Sharon Yixuan Li
Title: Towards Reliable Open-world Machine Learning
Abstract: The real world is open and full of unknowns, presenting significant challenges for machine learning (ML) models that must reliably handle diverse, and sometimes anomalous inputs. Out-of-distribution (OOD) uncertainty arises when a machine learning model sees a test-time input that differs from its training data, and thus should not be predicted by the model. As ML is used for more safety-critical domains, the abilities to handle out-of-distribution data are central in building open-world learning systems. In this talk, I will talk about methods, challenges, and opportunities towards building ROWL (Reliable Open-World Learning).To tackle these challenges, I will talk about recent advancements in energy-based OOD detection, a novel framework that improves uncertainty estimation with theoretical guarantees. We show that energy score is less susceptible to softmax’soverconfidence issue, and leads to superior performance on common OOD detection benchmarks. Going beyond, I will discuss new directions and progress towards building scalable, robust, and computationally efficient OOD detection algorithms.
Bio: Sharon Yixuan Li is an Assistant Professor in the Department of Computer Sciences at the University of Wisconsin-Madison. Previously, she was a postdoctoral researcher at Stanford AI Lab (SAIL). She obtained her Ph.D. from Cornell University in 2017, advised by John E. Hopcroft, Kilian Q. Weinberger, and Thorsten Joachims. She has served as Program Chair and a founding organizer of ICML Workshop on Robustness and Uncertainty in Deep Learning (UDL) in 2019 and 2020, area chair for NeurIPS, ICLR, and ICML. She has previously spent time at Google AI and Facebook AI. She was named 30 Under 30 Rising Stars in AI in 2019, and Forbes 30 Under 30 in Science in 2020. Website: http://pages.cs.wisc.edu/~sharonli/