Addison Hu

About

I am a Visiting Scientist at Latitude AI. My work draws on concepts from off-policy evaluation and causal inference to better guide the development of the self-driving stack.

In November 2023, I received a PhD in Statistics and Machine Learning from Carnegie Mellon University, where I was lucky to be advised by Ryan Tibshirani. I spent the final year and a half of graduate school as a visiting researcher at the UC Berkeley Department of Statistics. Prior to all this, I received a BS in Statistics from Yale University and spent some time working at Facebook.

Interests

My work touches on statistical methodology, nonparametric theory, and optimization. My main focus at this time is understanding the use of total variation penalties in the scattered data, d-dimensional setting, which can be thought of as a multivariate generalization of locally adaptive regression splines or trend filtering. My collaborators and I have provided a treatment of the zeroth-order case, and we are currently working on the first-order case. In each of these projects, I consider the estimation problem from different angles: statistical rates, efficient computation, practical usage.

I also have an applied interest in computational epidemiology. At the beginning of the Covid-19 pandemic, I joined CMU's Delphi group on an emergency basis to help produce real-time Covid-19 indicators and forecasts. I found the work compelling enough to continue working on related problems, including forecasting influenza for CDC FluSight.

Papers

Ordered by time of completion.

Mathis et al. Evaluation of FluSight influenza forecasting in the 2021–22 and 2022–23 seasons with a new target laboratory-confirmed influenza hospitalizations. Submitted, 2023.
Addison Hu, Alden Green, and Ryan Tibshirani. The Voronoigram: Minimax estimation of bounded variation functions from scattered data. Submitted, 2022.
Veeranjaneyulu Sadhanala, Yu-Xiang Wang, Addison Hu, and Ryan Tibshirani. Multivariate trend filtering for lattice data. Submitted, 2021.
Daniel McDonald, Jacob Bien*, Alden Green*, Addison Hu*, et al. Can auxiliary indicators improve COVID-19 forecasting and hotspot prediction? Proceedings of the National Academy of Sciences, 2021.
Reinhart et al. An open repository of real-time COVID-19 indicators. Proceedings of the National Academy of Sciences, 2021.
Cramer et al. Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the United States. Proceedings of the National Academy of Sciences, 2022.
Addison Hu, Mikael Kuusela, Ann Lee, Donata Giglio, and Kimberly Wood. Spatio-temporal methods for estimating subsurface ocean thermal response to tropical cyclones. Submitted, 2020.
Addison Hu and Sahand Negahban. Minimax estimation of bandable precision matrices. Advances in Neural Information Processing Systems, 2017.

* denotes equal contribution.

Awards

I am grateful to have been supported by an NSF GRFP award in Mathematical Statistics.

Service

I have served as a referee/reviewer for the Annals of Statistics; Journal of Machine Learning Research; Journal of Computational and Graphical Statistics; and Neural Information Processing Systems.