Skip to main content

Publications

I like developing simple and robust methods from first principles, and I believe good ideas should always be easy to explain, so I have included a short TL;DR for my main papers below.

* denotes equal contribution. See also my Google Scholar profile.

2026

Khoo, S.*, Liu, S.* & Beaumont, M. (2026). Local Manifold Identification with Latent Linear Models and OT Flows. Advances in Neural Information Processing Systems (NeurIPS 2026). Poster
(Also presented as a spotlight poster at the ICML 2026 Workshop on Foundations of Deep Generative Models.)
TL;DR We combine a pretrained generative model with a simple linear model, fitted in its latent space, to perform local manifold estimation.

Khoo, S., Prangle, D., Liu, S. & Beaumont, M. (2026). Minimum Distance Summaries for Robust Neural Posterior Estimation. International Conference on Machine Learning (ICML 2026). Poster [arXiv]
TL;DR We make a pretrained neural posterior estimator robust to model misspecification without retraining, by choosing the summary statistics that best match the observed data.

2025

Khoo, S., Wang, Y., Liu, S. & Beaumont, M. (2025). Direct Fisher Score Estimation for Likelihood Maximization. Advances in Neural Information Processing Systems (NeurIPS 2025), 38, 87571–87604. Spotlight, top 3% [arXiv]
TL;DR We perform maximum likelihood estimation for simulator models by gradient ascent, learning the Fisher score with a local score-matching objective that has a fast closed-form solution for linear score models.

2024

Wang, Y., Khoo, S. & Liu, S. (2024). Lightspeed Black-Box Bayesian Optimisation via Local Score Matching. NeurIPS 2024 Workshop on Bayesian Decision-Making and Uncertainty. [OpenReview]

Pacchiardi, L., Khoo, S. & Dutta, R. (2024). Generalized Bayesian Likelihood-Free Inference. Electronic Journal of Statistics, 18(2), 3628–3686. [Paper]

Preprints

Dutta, R., Xu, Y., Khoo, S., Basini, F. & Futschik, A. (2025). Signature-Informed Selection Detection: A Novel Method for Multi-Locus Temporal Population Genetic Model with Recombination. arXiv preprint. [arXiv]

Talks & posters

* presenting author

Khoo, S.*, Wang, Y., Liu, S. & Beaumont, M. Approximate Maximum Likelihood Estimation with Local Score Matching. Poster, BayesComp 2025, National University of Singapore, 20 June 2025. Best Poster Award

Khoo, S.*, Wang, Y., Liu, S. & Beaumont, M. Direct Fisher Score Estimation for Likelihood Maximization. Talk, Simulation-Based Inference for Galaxy Evolution Workshop, University of Bristol, 29 May 2025.

Research in econometrics

During my undergraduate degree, I carried out applied econometrics research, published in Rationale, the LSE Economics Society’s journal.

Kim, M., Johansson, M. & Khoo, S. (2020). The Impact of Immigration on Average Income in Sweden. Rationale, 2, 35–68. [Paper]

Dann, C., Grohmann, J., Ho, A., Khoo, S., Kim, A. & Lal, S. (2020). The Hedonic House Price Model in a Development Context: Evidence from the National Capital Region of the Philippines, 2000–2010. Rationale, 1, 23–58. [Paper]