Nanyang Assistant Professor and NRF-AI Fellow @ CCDS, Nanyang Technological University, Singapore. Previously a Nightingale Fellow @ Oxford Stats Department.
Joined April 2019
- Tweets33
- Following243
- Followers229
- Likes78
Our paper, “Causal Falsification of Digital Twins”, was recently published in JMLR.
jmlr.org/papers/v27/24-1752.…
We consider the problem of assessing a digital twin using observational data under minimal assumptions, and in particular without assuming all confounders are measured.
Using a novel set of longitudinal causal bounds extending Manski's classical bounds, we obtain an exact testing procedure requiring only an i.i.d. dataset of observational trajectories.
We apply this at scale to the Pulse Physiology Engine using the MIMIC-III dataset.
Joint work with @FaaizTaufiq, @ArnaudDoucet1 and @cholmesuk.
A big thanks also to the Pulse team at @Kitware for their troubleshooting and development assistance!
Great work on equivariant Clifford circuit synthesis led by @richie_yeung!
Our work on “Equivariant Reinforcement Learning for Clifford Quantum Circuit Synthesis” is finally on arXiv!
arxiv.org/abs/2605.10910
Check out our in-browser demo, vibe-coded in WebAssembly:
y-richie-y.github.io/cliffor…
TLDR: Our RL agent, equipped with a bespoke neural network which respects the symmetries of the symplectic stabiliser tableau, can effectively synthesise **optimal** Clifford quantum circuits in terms of gate count.
The model generalises across unseen tableaus of varying sizes, and on examples that are out-of-distribution.
A categorical account of the Metropolis-Hastings algorithm
arxiv.org/abs/2601.22911
Joint work with Andi Wang.
We give algebraic (i.e. measure theory-free) necessary and sufficient conditions for an abstract Metropolis-Hastings-type sampler to be reversible wrt a given target.🧵
We also study CD categories enriched over commutative monoids. This provides an expressive setting for reasoning about:
* Substochastic kernels
* Probability values
* Finiteness and sigma-finiteness
* (Pointwise) absolute continuity
* Singular measures
* Lebesgue decompositions
With this additional structure, we can formulate an abstract version of an involutive-style MH kernel and give conditions under which it is reversible. When instantiated in terms of measure theory, we recover the results of Andrieu et al. (arxiv.org/pdf/2012.14881).
I'm looking for talented and ambitious PhD students to join me at Nanyang Technological University Singapore to work on safe and robust AI systems!
Full scholarships covering tuition and a stipend are available, and are open to local and international students alike.
We develop methodology with precise guarantees under minimal assumptions, with large-scale safety-critical applications in mind.
We also develop better tools for describing and reasoning about AI systems, using category theory, programming languages theory, and proof assistants.
I still have positions available for a January 2026 intake - get in touch ASAP! Later intakes are also possible (please get in touch to discuss).
For more information, and details on how to apply, please see: jrmcornish.github.io/recruit…
Rob Cornish retweeted
We finally have the strong law of large numbers in Markov categories.
arxiv.org/abs/2503.21576
A meta-point of this paper is that category theory has utility for reasoning about current problems of interest in mainstream machine learning. The theory is predictive, not just descriptive. 🧵(1/6)
In our new paper (accepted at ICLR!), we propose the first framework for constructing equivariant diffusion models via symmetrisation
This allows us to ensure E(3)-equivariance with just highly scalable standard architectures such as Diffusion Transformers, instead of EGNNs, for molecular generation
Joint work with @kiaashour @yeewhye @rob_cornish
(1/9)
If this is of interest to you, here is a recent talk that @PaoloPMath and I gave for a class at MIT: youtube.com/watch?v=ozN4zLEU…
You can also find an extended abstract of my longer Markov categories paper here: arxiv.org/abs/2412.09469