Scholar
Matthew Regehr
Google Scholar ID: Jy0SwIcAAAAJ
University of Waterloo
Citations & Impact
Citations
64
H-index
5
i10-index
3
Publications
10
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- - Query-Efficient Locally Private Hypothesis Selection via the Scheffé Graph (2025), to appear in NeurIPS 2025
- - Avoiding Pitfalls for Privacy Accounting of Subsampled Mechanisms under Composition (2024), published in SaTML 2025
- - A Bias-Accuracy-Privacy Trilemma for Statistical Estimation (2023), published in Journal of the American Statistical Association
- - An Elementary Proof that Q-learning Converges Almost Surely (2021)
- - Automatic Right Atrial Segmentation from Magnetic Resonance Imaging (2019), won best poster prize at the University of Alberta Faculty of Medicine & Dentistry 52nd Annual Summer Students' Research Day
Research Experience
- Involved in multiple research projects related to differential privacy and machine learning, such as privacy budgeting theory, locally private hypothesis selection, etc.
Education
- PhD, Cheriton School of Computer Science, University of Waterloo, advised by Gautam Kamath.
Background
- Matthew Regehr is a PhD student in the Cheriton School of Computer Science at the University of Waterloo. His research interests broadly include machine learning theory and differential privacy. His current research projects focus on the theory of privacy budgeting, locally private hypothesis selection, and fairness in machine learning.
Miscellany
- Personal interests and other writings:
- - University of Waterloo CS 798 Project: A Brief Survey of Private and Online Learning
- - University of Waterloo CS 860 Project: A Strange Square Complex and c3-Locally Testable Codes
Co-authors