Jonathan Wenger

Graduate Student in Computer Science

University of Tübingen

About me

I am a doctoral student in computer science at the University of Tübingen and an IMPRS-IS fellow advised by Philipp Hennig. My research focuses on numerical methods for machine learning, in particular probabilistic numerics which interprets numerical methods as learning procedures. I am also closely collaborating with John P. Cunningham at Columbia University’s Zuckerman Institute.

In my free time I love to do sports, in particular any kind of cycling.

Interests

Probabilistic Numerics
Numerical Analysis
Gaussian Processes

Education

Doctoral Degree in Computer Science
University of Tübingen, IMPRS-IS fellow
MSc in Engineering Physics, 2019
KTH Royal Institute of Technology
MSc in Mathematics, 2019
TU München

Publications

Recent Preprints and Papers

Physics-Informed Gaussian Process Regression Generalizes Linear PDE Solvers

Linear partial differential equations (PDEs) are an important, widely applied class of mechanistic models, describing physical …

Marvin Pförtner, Ingo Steinwart, Philipp Hennig, Jonathan Wenger (2022). Physics-Informed Gaussian Process Regression Generalizes Linear PDE Solvers.

Posterior and Computational Uncertainty in Gaussian Processes

Gaussian processes scale prohibitively with the size of the dataset. In response, many approximation methods have been developed, which …

Jonathan Wenger, Geoff Pleiss, Marvin Pförtner, Philipp Hennig, John P. Cunningham (2022). Posterior and Computational Uncertainty in Gaussian Processes. NeurIPS.

PDF Code Video arXiv

Preconditioning for Scalable Gaussian Process Hyperparameter Optimization

Gaussian process hyperparameter optimization requires linear solves with, and log-determinants of, large kernel matrices. Iterative …

Jonathan Wenger, Geoff Pleiss, Philipp Hennig, John P. Cunningham, Jacob R. Gardner (2022). Preconditioning for Scalable Gaussian Process Hyperparameter Optimization. ICML (oral, 2.1%).

PDF Code Poster Slides Video arXiv Proceedings

ProbNum: Probabilistic Numerics in Python

Probabilistic numerical methods (PNMs) solve numerical problems via probabilistic inference. They have been developed for linear …

Jonathan Wenger, Nicholas Krämer, Marvin Pförtner, Jonathan Schmidt, Nathanael Bosch, Nina Effenberger, Johannes Zenn, Alexandra Gessner, Toni Karvonen, François-Xavier Briol, Maren Mahsereci, Philipp Hennig (2021). ProbNum: Probabilistic Numerics in Python. arXiv preprint.

PDF Code arXiv Documentation

Probabilistic Linear Solvers for Machine Learning

Linear systems are the bedrock of virtually all numerical computation. Machine learning poses specific challenges for the solution of …

Jonathan Wenger, Philipp Hennig (2020). Probabilistic Linear Solvers for Machine Learning. NeurIPS.

PDF Code Poster Slides Video arXiv Proceedings

See all publications

Talks

Recent and Upcoming

Preconditioning for Scalable Gaussian Process Hyperparameter Optimization

Mar 2, 2023 16:45 Amsterdam, The Netherlands

Preconditioning for Scalable Gaussian Process Hyperparameter Optimization

Jul 19, 2022 14:00 Baltimore, MD, USA

ProbNum: Probabilistic Numerics in Python

Oct 25, 2021 10:00 Schloss Dagstuhl, Wadern, Germany

Teaching

Classes and Tutorials

Lecture: Numerics of Machine Learning WS2022/23

Tutorial: Data Literacy WS2021/22

Tutorial: Probabilistic Machine Learning SS2021

Projects

Software, Tech Reports, etc.

ProbNum: Probabilistic Numerics in Python

Probabilistic Numerics (PN) interprets classic numerical routines as inference procedures by taking a probabilistic viewpoint. This …

Jonathan Wenger

Code Documentation

On Probability Calibration and Overconfidence in Image Classification

Accurate representation of uncertainty in classification problems can be as critical as high prediction accuracy in computer vision, …

Jonathan Wenger

Numerical Methods for Time-optimal Control of Linear Systems

In this project we analyzed time-optimal control problems with linear dynamics and numerical methods for solving them. When …

Jonathan Wenger