Mark van der Wilk

Together with my research group, I work on three central questions: - How do we find general patterns that allow generalization beyond the training set, without humans manually encoding them? (Equivariance, causality, continual learning…) - How can we create neurons that automatically assemble their connectivity structure (architecture), while minimising the computational costs of the network as a whole? (Generalisation bounds, Bayesian model selection, MDL, meta-learning) - How do we interact with the environment, while avoiding risk but learning as quickly as possible? (Bayesian optimisation, foundation models for industrial applications e.g. chemistry) The hope is that answering these questions will lead to ML/AI methods that make better predictions with less technical input from humans. I collaborate with practitioners in science and industry to apply these methods in science and engineering applications (e.g. in bioengineering and chemistry). Longer term, I believe that new training algorithms may open doors to more energy-efficient methods, that are more suitable for future hardware architectures, which often favour local communication patterns. I am also exploring ways in which AI systems can make it easier for humans to share information, and build consensus, to make collaboration easier.

(2025). A Meta-Learning Approach to Bayesian Causal Discovery. The Thirteenth International Conference on Learning Representations (ICLR). (2025). Adjusting Model Size in Continual Gaussian Processes: How Big is Big Enough?. Forty-second International Conference on Machine Learning (ICML). (2025). Continuous Bayesian Model Selection for Multivariate Causal Discovery. Forty-second International Conference on Machine Learning (ICML). (2025). Estimating Interventional Distributions with Uncertain Causal Graphs through Meta-Learning. ICML 2025 Workshop on Scaling Up Intervention Models. (2025). Rethinking Aleatoric and Epistemic Uncertainty. Forty-second International Conference on Machine Learning (ICML).