Thang D. Bui
I'm looking for PhD/MPhil students and postdoctoral fellows. Please see my research interests below, my publications and read through this page.
News
5/2026 TMLR paper: Beyond ReinMax: Low-Variance Gradient Estimators for Discrete Latent Variables, led by Daniel Wang
3/2026 Will be an area chair for NeurIPS 2026.
3/2026 Preprint: Beyond ReinMax: Low-Variance Gradient Estimators for Discrete Latent Variables, led by Daniel Wang
3/2026 Preprint: Active Flow Matching, led by Yashvir Grewal, with Daniel Steinberg, Cheng Soon Ong, and Edwin Bonilla
10/2025 Awarded an ARC Discovery Project grant on diffusion models (with Sumeetpal Singh, Sahani Pathiraja, and Pierre Del Moral)
10/2025 TMLR paper received a J2C certification.
9/2025 Two papers accepted to NeurIPS!
7/2025 Preprint: Sparse GPs: Structured approximations and Power-EP revisited with Michalis Titsias
6/2025 Preprint: Rao-Blackwellised Reparam Gradients, led by Kevin Lam, with George Deligiannidis and Yee Whye Teh
5/2025 TMLR paper with Matt Ashman and Rich Turner.
Bio
I'm a senior lecturer (equivalent to tenure-track Associate Professor) in Machine Learning at the School of Computing, Australian National University.
Before that, I was a lecturer at the University of Sydney from 2018 to 2022 and spent two years (2019-2020) at Uber AI.
I completed my doctoral training at the Cambridge Machine Learning group, supervised by Richard Turner and advised by Carl Rasmussen.
My current research interests include probabilistic modelling and inference, Monte Carlo and approximate inference methods, distributed, active and continual learning, and model-based reinforcement learning. My group currently focus on
how to obtain uncertainty estimates (for example, developing efficient approximate inference methods for Gaussian processes and neural networks)
how to update models and uncertainty in changing environments (for example, developing continual learning or unlearning methods for streaming data)
how to interpret models and algorithms (for example, understanding behaviours in transformers or neural network training)
Contact
Hanna Neumann Building 145, Office 2.22
School of Computing, College of Engineering and Computer Science
The Australian National University, Australia
thang.bui at at at at anu.edu.au
|