Siu Lun Chau

Postdoctoral Researcher

CISPA Helmholtz Center for Information Security

Biography

Hallo! I am Siu Lun Chau and my friends call me Alan. I am currently a postdoctoral researcher at the CISPA Helmholtz Center for Information Security working with Dr. Krikamol Muandet to develop trustworthy machine learning algorithms drawing insights from economics. From 2018 to 2022, I did my DPhil (PhD) in Statistical Machine Learning at the University of Oxford. I was supervised by Prof. Dino Sejdinovic, Prof. Mihai Cucuringu and Prof. Xiaowen Dong. I hold a Masters and Bachelor in Mathematics and Statistics at the University of Oxford as well. I have perviously interned at Amazon London and the Max Planck Institute for Intelligent Systems.

Download my resumé here.

Interests

Economics and Experimental design
Kernel methods and Gaussian Processes
Ranking and Preference learning
Causal Inference and Econometrics
Explainability and Uncertainty modelling
Graph machine learning

Education

DPhil in Statistical Machine Learning, 2019 - 2023

University of Oxford
MMATH in Mathematics and Statistics (First Class Honours), 2017 - 2018

University of Oxford
BA in Mathematics and Statistics (First Class Honours), 2014 - 2017

University of Oxford

Publications

Explaining Preferences with Shapley Values

We propose Pref-SHAP, a Shapley value-based model explanation framework, to explain and interpret preference models.

Published in NeuRIPS 2022

Siu Lun Chau, Robert Hu, Jaime Ferrando Huertas, Dino Sejdinovic

Explaining Preferences with Shapley Values

RKHS-SHAP: Shapley values for kernel methods

We propose RKHS-SHAP, a SHAP algorithm designed for kernel methods. It can efficiently compute both interventional and observational shapley values using kernel mean embedding of distributions.

Published in NeuRIPS 2022

Siu Lun Chau, Robert Hu, Javier Gonzalez, Dino Sejdinovic

RKHS-SHAP: Shapley values for kernel methods

Giga-scale Kernel Matrix-Vector Multiplication on GPU

Building on top of the Fast Multipole Method, we propose Faster-Fast and Free Memory Method (F3M) to run Kernel Matrix vector-Multiplication on tall (n~10^9) and skinny (D~7) data using a single GPU.

Published in NeuRIPS 2022

Robert Hu, Siu Lun Chau, Dino Sejdinovic, Joan Alexis Glaunes

Giga-scale Kernel Matrix-Vector Multiplication on GPU

Spectral Ranking with Covariates

We consider spectral approaches to the problem of ranking n players given their incomplete and noisy pairwise comparisons, but revisit this classical problem in light of player covariate information.

Published in ECML PKDD 2022

Siu Lun Chau, Mihai Cucuringu, Dino Sejdinovic

Learning Inconsistent Preferences with Gaussian processes

We challenge the assumption of rankability of duelling data and derive algorithms to model inconsistent preferences.

Published in AISTATS 2022

Siu Lun Chau, Javier Gonzalez, Dino Sejdinovic

Learning Inconsistent Preferences with Gaussian processes

BayesIMP: Uncertainty Quantification for Causal Data Fusion

We propose a Bayesian Conditional Mean Embedding to embed interventional distribution into the reproducing kernel Hilbert space while able to quantify uncertainty.

Published in NeuRIPS 2021

Siu Lun Chau, Jean Francois Ton, Javier Gonzalez, Yee Whye Teh, Dino Sejdinovic

Deconditional Downscaling with Gaussian Processes

We propose a Bayesian solution for statistical downscaling which handles unmatched multi-resolution data through the Deconditional Mean Operator.

Published in NeuRIPS 2021

Siu Lun Chau, Shahine Bouabid, Dino Sejdinovic

Kernel-based Graph Learning From Smooth Signals: A Functional Viewpoint

We proposed a graph learning algorithm to recover topological structure from observed graph signals.

Published in IEEE Transactions on Signal and Information Processing over Networks

Xingyue Pu, Siu Lun Chau, Xiaowen Dong, Dino Sejdinovic

Kernel-based Graph Learning From Smooth Signals: A Functional Viewpoint

Experience

Postdoctoral Researcher

CISPA Helmholtz Center for Information Security

Mar 2023 – Mar 2025 Saarbrucken, Germany

Applied Scientist Intern

Amazon

Jun 2022 – Dec 2022 London, UK

Data Science Consultant

Ravio

May 2022 – Present London, UK

Research Intern

Empirical Inference, Max Planck Institute of Intelligent Systems

Oct 2021 – Present Tubingen, Germany

Supervised by Dr. Krikamol Muandet, working on the interface of machine learning and econometric modelling.

Machine Learning Content Developer

Cambridge Spark

Aug 2017 – Present London, UK

Designed data science and machine learning projects and delivered training courses to upskill students and corporates.

Machine Learning Engineering

gini

Oct 2020 – Jan 2021 Hong Kong

Developed a Gaussian processes based explainable time series model for giniPredict, a forecasting tool built for use in Google spreadsheets for decision-makers.