Geoff Pleiss
CIFAR AI Chair, Vector Institute
geoff.pleiss <at> stat.ubc.ca
I am an assistant professor in the Department of Statistics at the University of British Columbia, where I am an inaugural member of CAIDA's AIM-SI (AI Methods for Scientific Impact) cluster. I am also a Canada CIFAR AI Chair and a faculty member at the Vector Institute.
My research interests intersect deep learning and probablistic modeling. More specifically, I'm interested in heuristic and approximate notions of uncertainty from machine learning models, and how they can inform reliable and optimal downstream decisions within the contexts of experimental design and scientific discovery. Major focuses of my work include:
- neural network uncertainty quantification,
- Bayesian optimization,
- Gaussian processes, and
- ensemble methods.
I am also an active open source contributior. Most notably, I co-created and maintain the GPyTorch Gaussian process library with Jake Gardner.
Previously, I was a postdoc at Columbia University with John P. Cunningham. I received my Ph.D. from the CS department at Cornell University in 2020 where I was advised by Kilian Weinberger and also worked closely with Andrew Gordon Wilson.
Current Students and Interns
Donney Fan
PhD StudentUBC Computer Science(co-sup. by Mark Schmidt)
Tim G. Zhou
MSc StudentUBC Computer Science(co-sup. by Evan Shelhamer)
Zachary Lau
MSc StudentUBC StatisticsNathan Cantafio
BSc StudentUBC StatisticsTristan Cinquin
Research InternVector Institute
Interested in joining my lab? See the page on joining my lab for information on how to apply/contact me.
Recent and Selected Publications
For a full list of publications, please see my CV or my Google Scholar page.
NEWAsymmetric Duos: Sidekicks Improve Uncertainty
- Tim G. Zhou
- Evan Shelhamer
- Geoff Pleiss
— Under SubmissionNEWTheoretical Limitations of Ensembles in the Age of Overparameterization[Spotlight presentation]
- Niclas Dern
- John P. Cunningham
- Geoff Pleiss
— In International Conference on Machine Learning, 2025Approximation-Aware Bayesian Optimization[Spotlight presentation]
- Natalie Maus
- Kyurae Kim
- Geoff Pleiss
- David Eriksson
- John P. Cunningham
- Jacob R. Gardner
— In Neural Information Processing Systems, 2024Deep Ensembles Work, But Are They Necessary?
- Taiga Abe*
- E. Kelly Buchanan*
- Geoff Pleiss
- Richard Zemel
- John P. Cunningham
— In Neural Information Processing Systems, 2022Fast Matrix Square Roots with Applications to Gaussian Processes and Bayesian Optimization
- Geoff Pleiss
- Martin Jankowiak
- David Eriksson
- Anil Damle
- Jacob R. Gardner
— In Neural Information Processing Systems, 2020GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration.[Spotlight presentation]
- Jacob R. Gardner*
- Geoff Pleiss*
- David Bindel
- Kilian Q. Weinberger
- Andrew Gordon Wilson
— In Neural Information Processing Systems, 2018On Fairness and Calibration
- Geoff Pleiss*
- Manish Raghavan*
- Felix Wu
- Jon Kleinberg
- Kilian Q. Weinberger
— In Neural Information Processing Systems, 2017On Calibration of Modern Neural Networks
- Chuan Guo*
- Geoff Pleiss*
- Yu Sun*
- Kilian Q. Weinberger
— In International Conference on Machine Learning, 2017
Recent and Selected Talks
Oct. 2024Ensembles in the Age of Overparameterization: Promises and Pathologies
Recent empirical and theoretical work characterizing ensembles of neural networks.Dec. 2023Troubling Trajectories for Uncertainty Quantification and Decision Making with Neural Networks
A discussion of uncertainty quantification and my recent work on neural network ensembles.Spring 2022Bridging The Gap Between Deep Learning and Probabilistic Modeling
A talk connecting my Gaussian process and neural network research.
Selected Open Source
For a full list of respositories I actively contribute to, please see my Github page.
CoLA (Compositional Linear Algebra)Beta Release
A library for structured linear algebra operations in JaX and PyTorch.Coauthors: Andres Potapczynski, Marc Anton Finzi