Research, Open Source and Personal Projects

Projects

Code I've written and maintain, for research projects I list the papers as well.

Pythonscikit-learnMMD
reg-mmd-scikit
A scikit-learn compatible implementation of the regMMD estimation and regression procedure.
Estimation based on the MMD criterion allows for robust estimation. This implementation follows the standard scikit-learn format, allowing for easy integration into any Data Science project. Based on the work done by Piere Alquier, and ...
PythonPyTorchOnline optimisation
meta-grad-pytorch
A PyTorch implementation of the MetaGrad algorithm.
MetaGrad adapts its learning rate by keeping track of multiple learning rates and estimating the best one. Packaging it as a standard torch.optim.Optimizer makes it usable as a drop-in replacement in an existing training loop. Based on the work done by Tim van Erven, Wouter Koolen and Dirk van der Hoeven
PythonReproducibility
sample-efficient-learning-of-concepts
Experiment code for 'Sample-efficient Learning of Concepts with Theoretical Guarantees' (NeurIPS 2025). [Paper]
Repository containing all the code needed to reproduce the experiments. Additionally contains a scikit-learn style implementation of the estimator we propose.
PythonReproducibility
consequences-of-recourse
Experiment code for 'Risks of Recourse in Binary Classification' (AISTATS 2024). [Paper]
Repository containing all the code needed to reproduce the experiments.
PythonReproducibility
recourse-robust-explanations-impossible
Experiment code for 'Attribution-based Explanations that Provide Recourse Cannot be Robust' (JMLR 2023). [Paper]
Repository containing all the code needed to reproduce the experiments.
AstroTypeScriptDocker
This website
The site you're looking at, written in Astro, static output, zero javascript dependencies.
This was a React app first, but I rewrote it to have a new design and a new framework. The idea was to simplify the site, and reduce what is served to basically zero dependencies.