A Series of Statistically Probable Events

Jorn W.T. Peters

About Me

Ardness, 2020 wfh, 2021

Welcome to the personal website of Jorn Peters. Presently, I am a Data Scientist at Xomnia in Amsterdam. Before that I was a PhD Student with Max Welling at AMLab at the Unversity of Amsterdam—where I also obtained my MSc in Artificial Intelligence. Through these positions I have developed an interest in the tooling (i.e., libraries and frameworks) that help machine learning researchers and practitioners, such as myself, to be productive.

I consider myself a team player; I thrive when I am able to bounce ideas off other people and I enjoy being part of a joint creative and stategic process. Moreover, I take pleasure in supporting and enabling the people around me, especially if this is achieved through the software I write. My research interests include: generative modeling, deep learning (e.g., probabilistic and information theoretic DL), and frameworks/libraries that support machine learning research and applications. In terms of applied machine learning I am particularly interested in fields such as: health, medicine, and environmental sciences.

Tubus

Currently I am working on a library to support reproducible machine learning experiments called Tubus (to be released soon). The main goals of Tubus are:

  1. Modularity: Tubus should help the user write modular and reusable code while at the same time make it easy to combine modules into larger experiments.
  2. Time travel: it should always be possible to go back to (or compare) the exact state of an old experiment and rerun it—even if the code was not properly checked into version control.
  3. Opt in: Tubus is designed as a library and not a framework. That is, Tubus will not force a particular way of working on its users, but instead provides tools to ease the explorative machine learning experience.

I am working hard to release a first (open source) version of Tubus in Q2 of 2021.

Papers

Peters Self Normalizing Flows. T Keller, J. Peters, P. Jaini, E. Hoogeboom, P. Forré, M. Welling. ICML 2021.

Integer Discrete Flows and Lossless Compression. Emiel Hoogeboom, Jorn W.T. Peters, Rianne van den Berg, Max Welling. NeurIPS 2019.

Probabilistic Binary Neural Networks. Jorn W.T. Peters, Tim Genewein, Max Welling. arXiv 2018.

HexaConv. Emiel Hoogeboom, Jorn W.T. Peters, Taco S. Cohen, Max Welling. ICLR 2018.