About the Project Quasi-Monte Carlo (QMC) methods and machine learning (ML) are two important areas of computational mathematics, with growing
opportunities for interaction. QMC methods improve upon conventional random sampling by constructing point sets that provide more uniform coverage of a domain, often leading to greater accuracy and computational efficiency. Meanwhile, advances in machine learning and optimisation provide new tools for designing and improving sampling methods. This PhD project will explore the intersection of these fields, investigating both how QMC methods can improve machine learning algorithms and how machine learning techniques can advance the theory and construction of QMC methods. One research direction will investigate the use of QMC and low-discrepancy sampling within machine learning, aiming to improve accuracy, efficiency and robustness. Potential applications include stochastic optimisation, kernel-based methods, high-dimensional integration and sampling algorithms. Particular attention will be given to understanding when QMC methods offer advantages over conventional random sampling and how these benefits depend on the structure of the underlying problem. A complementary direction will explore the use of machine learning and optimisation to develop new QMC methods. This may include optimising point configurations, learning effective sampling distributions, developing adaptive sampling strategies and constructing low-discrepancy point sets for complex or high-dimensional domains. The project will combine mathematical theory, algorithm development and computational experimentation. Research may involve discrepancy theory, numerical analysis, statistical learning and optimisation, with opportunities to establish theoretical guarantees and develop practical algorithms. The precise research questions will be shaped by the interests and expertise of the successful candidate. The project is particularly suited to students with a strong background in mathematics, statistics or computational science, and an interest in developing new methods at the interface of numerical computation and machine learning. How to apply To make a formal application, complete the online form at https://www.st-andrews.ac.uk/study/postgraduate-research/apply/ Although there is no fixed deadline, you are strongly encouraged to make your application and contact the potential supervisors as early as possible, ideally by December 2026. Further details More details about the PhD programme can be found in this document . Funding Notes Fully funded scholarship places (fees, plus stipend currently £20,467 for AY 2026/27 with an uplift to be determined for AY 2027/28) are typically available for well-qualified students. UK, EU and other overseas students are all encouraged to apply. Further details of the application and selection procedure are available at View Website and View Website . References Rusch, T. K., Kirk, N., Bronstein, M. M., Lemieux, C., & Rus, D. (2024). Message-Passing Monte Carlo: Generating Low-Discrepancy Point Sets via Graph Neural Networks. Proceedings of the National Academy of Sciences, 121(40), e2409913121. https://doi.org/10.1073/pnas.2409913121 Van Huffel, M. E., Kirk, N., Chahine, M., Rus, D., & Rusch, T. K. (2026). Neural Low-Discrepancy Sequences. Proceedings of the 43rd International Conference on Machine Learning, 306, 123416–123433. https://proceedings.mlr.press/v306/van-huffel26a.html Mishra, S., & Rusch, T. K. (2021). Enhancing Accuracy of Deep Learning Algorithms by Training with Low-Discrepancy Sequences SIAM Journal on Numerical Analysis. https://doi.org/10.1137/20M1344883 Buchholz, A., Wenzel, F., & Mandt, S. (2018). Quasi-Monte Carlo Variational Inference. Proceedings of the 35th International Conference on Machine Learning, 80, 668–677. https://proceedings.mlr.press/v80/buchholz18a.html


