This is a project which is currently making use of HPC facilities at Newcastle University. It is active.
For further information about this project, please contact:
MCMC remains a cornerstone of Bayesian computation, yet its practical performance is limited by two fundamental challenges: the difficulty of tuning transition kernels, and the inefficiency of raw sample paths produced by these kernels. This thesis addresses both challenges by developing a unified framework for post-processing and online improvement of MCMC, combining tools from Stein’s method, optimal quantisation, and modern Reinforcement Learning.
Wang, C., Chen, Y., Kanagawa, H., & Oates, C. J. (2024). Stein Π-Importance Sampling [Selected for spotlight presentation]. Advances in Neural Information Processing Systems (NeurIPS), 36.
Wang, C., Chen, W. Y., Kanagawa, H., & Oates, C. J. (2025). Reinforcement Learning for Adaptive MCMC. Proceedings of The 28th International Conference on Artificial Intelligence and Statistics (AISTATS), 258, 640–648.
Wang, C., Fisher, M. A., Kanagawa, H., Chen, W., & Oates, C. J. (2025). Harnessing the Power of Reinforcement Learning for Adaptive MCMC. https://arxiv.org/abs/2507.00671
Python-based scientific computing and machine-learning stack, including NumPy/SciPy for numerical linear algebra and statistics, Matplotlib for data visualisation, and PyTorch, JAX and Flax for GPU-accelerated deep learning models and high-dimensional MCMC experiments.