Postdoc position in Singular Learning Theory Machine Learning Models

Overview

Are you a PhD in Machine Learning, Computer Science, Mathematics, Statistics, Physics or a closely related field and want to join the mission of unlocking the “geometry of artificial intelligence”? Then come join us!

Join us
We are looking for a highly motivated postdoctoral researcher to join the mission of unlocking the geometry of artificial intelligence in the group of Dr. Patrick Forré. If you want to join the mission of unlocking the geometry of artificial intelligence then please apply!

Singular Learning Theory (SLT) is a mathematical framework for analysing statistical models that do not follow classical regularity assumptions. Such singular models include many models from statistical physics as well as almost all modern (“overparameterized”) machine learning models, like probabilistic graphical models, deep neural networks, diffusion models, transformers, e.g. large language models, etc. SLT is based on the geometrical understanding of the parameter space in relation to the statistical model.

One of the main goals of SLT is to quantify the complexity of such models with respect to the data generating process (and some prior probability distribution). SLT and the estimation of such quantities has recently led to many applications, ranging from model selection and uncertainty quantification, over detecting phase transitions in machine learning models during training, the finding of interpretable substructures to the explainability of general learning behaviour of such machine learning models, etc.

In this project, we want to build upon the recent developments in the field and either push the boundaries of SLT on the mathematical foundational theory side, extend SLT to new learning frameworks (e.g. variational inference or reinforcement learning, etc.) and/or apply it to modern machine learning models like large language models or diffusion models, etc.

You are expected to:

• take an active role in the research project either in the development of the mathematical theory of SLT and/or through novel applications of SLT in the machine learning domain.
• publish and present your findings in academic peer-reviewed journals, international workshops and/or conferences.
• support the teaching activities at the faculty (up to 10% of the time).

• A PhD in Machine Learning, Computer Science, Mathematics, Statistics, Physics or a closely related field. (In case you have finished a PhD thesis, and you are just awaiting your PhD defence, please also apply.)
• Enthusiasm for the scientific process: formulating and investigating hypotheses, either mathematically or by conducting experiments, disseminating findings via writing and oral presentations, etc.
• Experience in publishing academic papers in peer-reviewed journals and/or conferences.
• Professional command of English, both written and spoken.
• Ability both to work independently as well as to cooperate and work effectively within a (possibly interdisciplinary) team of researchers.
• Experience in programming and software development. Familiarity with Python and statistical computing libraries, like PyTorch or JAX, etc., would be preferred.
• You are a motivational teacher, with an encouraging teaching style.

We offer a temporary employment contract for 38 hours per week for a period of 18 months. The preferred starting date is as soon as possible and to be discussed.

The gross monthly salary, based on 38 hours per week and dependent on relevant experience, ranges between €3546 and €5538 (scale 10). This does not yet include the 8% holiday allowance and 8,3% year-end allowance, which will come on top. The UFO profile Researcher/Onderzoeker 4 is applicable. A favourable tax agreement, the 30% ruling, may apply to non-Dutch applicants. The Collective Labour Agreement of Universities of the Netherlands is applicable.

You will work here

The Faculty of Science (FNWI) has a student body of around 8,000, as well as 1,800 members of staff working in education, research or support services. Researchers and students at the Faculty of Science are fascinated by every aspect of how the world works, be it elementary particles, the…