Research
I am broadly interested in uncertainty quantification and decision-making under uncertainty. Bayesian statistics provides a principled framework for updating our beliefs about probabilistic models in light of observed data, while Bayesian decision theory uses these updated beliefs to guide decisions under uncertainty.
A central theme of my research is sequential data acquisition, including active learning and Bayesian optimisation, where we iteratively decide what data to collect next to achieve a particular learning or decision-making objective. These methods rely on quantifying uncertainty about the systems we wish to learn or optimise. My work in this area spans applications in oceanography [1, 2] and engineering [3].
Gaussian processes are a versatile class of probabilistic models for quantifying uncertainty and are widely used in sequential data acquisition. Their computational and statistical properties therefore play an important role in the efficiency of these methods. Some of my research focuses on developing scalable methods for Gaussian process inference [4].
More fundamentally, I am also exploring questions about uncertainty quantification itself, particularly from Bayesian and post-Bayesian perspectives. I am interested in understanding how different approaches to inference quantify uncertainty, how reliable these uncertainties are under model misspecification, and what implications they have for subsequent decision-making.
Publication
Preprints
Diaz-Aylwin, G., Neighbor, J., Talwar, A. M., Zhang, R. Y., & Moss, H. B. (2026). Example-driven Parametrisations for Bayesian Shape Optimisation. ArXiv. [arxiv]
Iguchi, Y., Livingstone, S., Vasdekis, G., & Zhang, R. Y. (2026). Pathwise skew-symmetric discretisation for SDEs with superlinear drift. ArXiv. [arxiv] [code]
Zhang, R. Y., Astfalck, L. C., Cripps, E. J., and Leslie, D. S., Moss, H. B. (2026). Dynamic Gaussian Processes and the Vanilla-SPDE Exchange. ArXiv. [arxiv]
Publications and Workshop Papers
Zhang, R. Y., Astfalck, L. C., Cripps, E. J., and Leslie, D. S., Moss, H. B. (2026). BALLAST: Bayesian Active Learning with Look-ahead Amendment for Sea-drifter Trajectories under Spatio-Temporal Vector Fields. International Conference on Machine Learning (ICML). [arxiv] [paper] [code] [talk] [poster]
Iguchi, Y., Livingstone, S., Nüsken, N., Vasdekis, G., & Zhang, R. Y. (2026). Skew-symmetric schemes for stochastic differential equations with non-Lipschitz drift: an unadjusted Barker algorithm. IMA Journal of Numeical Analysis. [arxiv] [paper] [code]
Zhang, R. Y., Moss, H. B., Astfalck, L. C., Cripps, E. J., and Leslie, D. S. (2024). BALLAST: Bayesian Active Learning with Look-ahead Amendment for Sea-drifter Trajectories. NeurIPS 2024 Workshop on Bayesian Decision-making and Uncertainty. [OpenReview] [slides] [talk]