Philipp Hennig

Philipp Hennig is a machine learning researcher at the University of Tübingen who helped establish probabilistic numerics, which treats numerical computation as a form of learning. After studying physics in Heidelberg and at Imperial College London, he earned a Cambridge PhD in 2011 and joined the Max Planck Institute for Intelligent Systems in Tübingen. He published Entropy Search in 2012, began leading a research group of his own in 2015, and became professor of Methods of Machine Learning at Tübingen in 2018. In 2021 he coauthored Laplace Redux, which introduced a PyTorch library for Bayesian deep learning, and in 2022 Cambridge University Press published his textbook Probabilistic Numerics. He is a director of the Tübingen AI Center, which he leads with Matthias Bethge and Co-Director Bernhard Schölkopf.

Now at: University of Tübingen, Tübingen AI Center and Max Planck Institute for Intelligent Systems
Works: 229
Citations: 4,286
h-index: 33

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

By 2011

Physics and a Cambridge PhD

He studies physics in Heidelberg and at Imperial College London. In 2011, he earns a PhD at the University of Cambridge on approximate inference in graphical models, in the group of Sir David J.C. MacKay.
2011

Max Planck Institute in Tübingen

He joins the Max Planck Institute for Intelligent Systems in Tübingen in 2011, working first in the Empirical Inference department of Bernhard Schölkopf.
2012

The Entropy Search algorithm

In 2012, he and Christian J. Schuler publish Entropy Search for Information-Efficient Global Optimization in the Journal of Machine Learning Research. The paper presents an algorithm for probabilistic global optimization that explicitly addresses the decision problem of maximizing the information gained from each evaluation.
2015

A call for probabilistic numerics

In 2015, he begins leading a research group of his own, with which he establishes the concept of probabilistic numerics, a mathematical theory of uncertainty in computer calculations. He, Michael A. Osborne and Mark Girolami post Probabilistic Numerics and Uncertainty in Computations on arXiv in June 2015, a call for numerical algorithms that return uncertainties in their calculations.
2018

Professor at Tübingen

In May 2018, he becomes professor of Methods of Machine Learning in the computer science department of the University of Tübingen. His project PANAMA, funded by the European Research Council and running from 2018 to 2023, develops a software framework for building numerical methods for AI in an increasingly automated way.
2021

Laplace Redux and a PyTorch library

In 2021, he coauthors Laplace Redux with Erik Daxberger, Agustinus Kristiadi, Alexander Immer, Runa Eschenhagen and Matthias Bauer. The NeurIPS 2021 paper introduces laplace, an easy-to-use PyTorch library for Laplace approximations in Bayesian deep learning.
2022

The Probabilistic Numerics textbook

Cambridge University Press publishes Probabilistic Numerics: Computation as Machine Learning in 2022, which he writes with Michael A. Osborne and Hans P. Kersting. The book shows that computational routines can be treated as learning machines that use Bayesian inference to build more flexible and efficient algorithms.
As of October 2026

Director of the Tübingen AI Center

He is a director of the Tübingen AI Center, which he leads with Matthias Bethge and Co-Director Bernhard Schölkopf. He is also a founding Fellow and Co-Director of the ELLIS Program on Theory, Algorithms and Computations of Modern Learning Systems, with Francis Bach and Lorenzo Rosasco.