I am a PhD candidate at the UCL Centre for Artificial Intelligence, supervised by Carlo Ciliberto, and I also collaborate with Emtiyaz Khan’s group at RIKEN AIP. My research asks how learning unfolds during training and what properties emerge once a model has been trained. Beyond research, I am interested in institution-building and in developing a vibrant scientific ecosystem in Singapore.
Research
How does learning occur in neural networks?
I study how optimization algorithms, training-data geometry, and model architecture interact during training, which properties emerge at different stages, and which mechanisms persist across models and scales.
How can we design better learning algorithms?
I use principles from Bayesian inference, optimization, and information geometry to explain why training updates work and to derive new ones systematically. Posterior correction and its connection to variance reduction is one concrete example.
Read the research agenda →Selected Publications
- SVRG and Beyond via Posterior CorrectionN. Daheim, T. Möllenhoff, M. L. Ang, M. E. KhanICML 2026. Oral.
- Explanation in an Emerging Science of Large Language ModelsJ. M. L. AngICML 2026 Workshop, Philosophy Meets Machine Learning
Writing
A science of machine learning
An argument for building a more unified science of machine learning using insights from philosophy and the history of science.
Perhaps you were born for such a time as this
A reflection on institution-building and my role to help Singapore navigate a changing international order.
Read the writing →About
My path into machine learning began before university and has passed through National Service, applied mathematics, research, teaching, and two startups. The longer account explains how those experiences shaped my interests in science, business, history, and ulitimately serving people.
Read more about me →How to Collaborate
If you are an undergraduate or master’s student, researcher, scientist, or founder with an idea that overlaps with my work, I would be glad to hear from you. Send a short note about what you are working on and how you think we might collaborate to ming.ang.23@ucl.ac.uk.