Online Conformal Prediction, Multi-Level Quantile Tracking, and Gradient Equilibrium

Abstract

This talk reviews uncertainty quantification tools in time series prediction. The overarching goal is to provide easy-to-use algorithms with formal guarantees. The algorithms we present build upon ideas from conformal prediction and control theory, are able to prospectively model conformal scores in an online setting, and adapt to the presence of systematic errors due to seasonality, trends, and general distribution shifts. We will then discuss an extension of these ideas to the setting of probabilistic forecasting, which is essentially a generalization of the framework to handle vector-valued predictions, i.e., predictions which take the form of a set of ordered quantile forecasts at different probability levels. Finally, we will generalize this even further to discuss an abstract property in online learning called gradient equilibrium, which encapsulates these settings, and more.

Bio

Ryan Tibshirani is a Professor in the Department of Statistics at UC Berkeley and a Principal Investigator in the Delphi group. From 2011-2022, Ryan was a faculty member in Statistics and Machine Learning at Carnegie Mellon University. He did his Ph.D. in Statistics at Stanford University (2011), with Jonathan Taylor as his thesis advisor. His research interests lie broadly in statistics, machine learning, and optimization, and he likes to think about problems from different angles: applied, computational, theoretical. More specifically, his interests include high-dimensional statistics, nonparametric estimation, distribution-free inference, probabilistic forecasting, convex optimization, and numerical methods. He also has an applied interest in computational epidemiology, specifically tracking and forecasting epidemics.