Tycho van der Ouderaa

Tycho is a postgraduate researcher (PhD) in probabilistic machine learning at Imperial College London and a visiting researcher at the University of Oxford, supervised by Dr. Mark van der Wilk.

He is currently working on allowing machine learning models to automatically learn inductive bias from data. He is particularly interested in Bayesian statistics, Generative Modeling, Probabilistic Deep Learning, Information Theory, Multivariate Statistics, Dynamical Systems and Reinforcement Learning. The goal is to create tools that help solve real-world problems. University of Amsterdam alumnus. Interned at Qualcomm Inc. and Microsoft Research.

Publications

Pyramid Vector Quantization for LLMs Tycho F. A. van der Ouderaa, Maximilian L. Croci, Agrin Hilmkil, James Hensman In submission. Pre-print online.

Noether's razor: Learning Conserved Quantities Tycho F. A. van der Ouderaa, Mark van der Wilk, Pim de Haan In NeurIPS 2024

Variational Inference Failures Under Model Symmetries: Permutation Invariant Posteriors for Bayesian Neural Networks Yoav Gelberg, Tycho F. A. van der Ouderaa, Mark van der Wilk, Yarin Gal In ICML 2024, GRaM Workshop (Awarded with Best Paper Award)

The LLM Surgeon Tycho F.A. van der Ouderaa, Markus Nagel, Mart van Baalen, Yuki M. Asano, Tijmen Blankevoort In ICLR 2024

Learning Layer-wise Equivariances Automatically using Gradients Tycho F.A. van der Ouderaa, Alexander Immer, Mark van der Wilk In NeurIPS 2023 (Awarded with spotlight)

Stochastic Marginal Likelihood Gradients using Neural Tangent Kernels Alexander Immer, Tycho F.A. van der Ouderaa, Mark van der Wilk, Gunnar Ratsch, Bernhard Schölkopf In ICML 2023

Sparse Convolutions on Lie Groups Tycho F.A. van der Ouderaa, Mark van der Wilk In NeurIPS 2022 (Workshop on Symmetry and Geometry in Neural Representations)

Relaxing Equivariance Constraints with Non-stationary Continuous Filters Tycho F.A. van der Ouderaa, David W. Romero, Mark van der Wilk In NeurIPS 2022

Invariance Learning in Deep Neural Networks with Differentiable Laplace Approximations Alexander Immer*, Tycho F.A. van der Ouderaa*, Vincent Fortuin, Gunnar Rätsch, Mark van der Wilk In NeurIPS 2022

Learning Invariant Weights in Neural Networks Tycho F.A. van der Ouderaa, Mark van der Wilk In UAI 2022 (Awarded with Oral) In ICML 2021 (UDL Workshop)

Deep Group-wise Variational Diffeomorphic Image Registration Tycho F.A. van der Ouderaa, Ivana Isgum, Wouter B. Veldhuis, Bob D. de Vos In MICCAI 2020 (Extended Oral at TIA Workshop)

Chest CT Super-resolution and Domain-adaptation using Memory-efficient 3D Reversible GANs Tycho F.A. van der Ouderaa, Daniel E. Worrall, Bram van Ginneken In MIDL 2019

Reversible GANs for Memory-efficient Image-to-Image Translation Tycho F.A. van der Ouderaa, Daniel E. Worrall In CVPR 2019

*: equal contribution.

Talks

Learning Inductive Bias in Neural Networks RainML Research Lab at University of Oxford, UK (29 Mar 2025). Invited talk. Slides	Bayesian Neural Model Selection for Symmetry Learning Joint Statistical Meetings 2024 in Portland, US (8 Aug 2024). Invited talk. Slides
The LLM Surgeon Microsoft Research AI & Pizza Event, Cambridge, UK (14 Mar 2024). Slides / ICLR 2024 paper
The LLM Surgeon Qualcomm AI Research Internship presentation, Amsterdam, NL (4 Nov 2023). Slides / ICLR 2024 paper
Learning Layer-wise Equivariances Automatically using Gradients Spotlight talk at NeurIPS@Cambridge event, Cambridge, UK (4 Dec 2023). Slides / NeurIPS 2023 paper (awarded with spotlight)
Learning Invariant Weights from Neural Networks Oral presentation at UAI 2022, Eindhoven, NL (2 Aug 2022). Slides / UAI 2022 paper

Theses

Reversible Networks for Memory-efficient Image-to-Image Translation in 3D Medical Imaging This master thesis proposes a novel image-to-image translation model that adapts reversible neural networks to perform memory-efficient image-to-image translation on both paired and unpaired data. We show that using the model we are able to perform several medical imaging tasks in 3D that would otherwise have not been possible due to GPU memory constraints. The thesis was assesed by Prof. Dr. Max Welling (UvA) and supervised by Prof. Dr. Bram van Ginneken (Radboud UMC) and Daniel E.Worrall (UvA). The code is available online at Github as well as presentation slides.

Deep Reinforcement Learning in Pac-man This bachelor thesis explores how a computer is able to learn how to play the game of Pac-man, solely by exploring different actions and observing the patterns in the observed states using deep neural networks. The thesis showed that a technique known as Deep Q-Learning can effectively be used for this problem. The thesis was supervised by Dr. Efstratios Gavves (UvA) and Matthias Reisser (UvA). The code is available online on Github and often used by other people for their projects.