Professor in Machine Learning, Gatsby Computational Neuroscience Unit Research Scientist, Google DeepMind https://xitter.cf/t.co/JJNqmn7jak
London, England
Joined September 2009
- Tweets1.1K
- Following804
- Followers11.8K
- Likes3.1K
Pinned Tweet
Proxy Variables for Causal Effect Estimation with Hidden Confounding
Keynote at @Conf_CLeaR 2026, video now online!
youtube.com/watch?v=ilspIe6Z…
Slides (with bonus content and fewer bugs):
gatsby.ucl.ac.uk/~gretton/co…
(1/2)
PhD applications at @GatsbyUCL are open!
Full funding available regardless of nationality.
Includes ✨new✨ Jump Trading studentships in AI with potential application to finance, eligible for employment opportunities with Jump Trading during PhD.
Deadline: 🚨November 9th🚨
Interested in a #PhD in machine learning & theoretical neuroscience?
Applications for 2027 entry to the Gatsby Unit PhD programme now open!
💰 Fully-funded 4-year programme
🤝 Close links to Sainsbury Wellcome Centre & the ELLIS Unit @ UCL (@uclcsml)
ℹ️ ucl.ac.uk/life-sciences/gats…
Congratulations @DimitriMeunier1 on winning a G-Research PhD Prize! 🎉🎉
gresearch.com/news/g-researc…
Congratulations Dr @agalashov on an outstanding thesis defence!!
...and warm thanks to Sinho Chewi and Brooks Paige for their diligent thesis examining!
Congratulations Dr @DimitriMeunier1 on a fantastic thesis defence!!
...and warm thanks to @songandyliu and Sivaraman Balakrishnan for their diligent thesis examining!
Exciting talk by @ValentinDeBort1 next Wednesday Oct 14! The first in this year's CSML seminar series.
The UCL CSML Seminar is back! For our first talk, Valentin de Bortoli (GDM) @ValentinDeBort1 will present simplex diffusion models! The talk will be 12:30 (BST) Wed 14 of Oct at @SWC_Neuro @GatsbyUCL , and streamed at ucl.zoom.us/j/99748820264.
Arthur Gretton retweeted
The submission portal for CLeaR 2027 is open!
Portal: openreview.net/group?id=ccle…
Website: cclear.cc/2027
Gradient flows on the Maximum Mean Discrepancy (and friends)!
Talk on Monday 05 October, 9am EST.
Register here for the livestream!
edpmworkshop.github.io/
Arthur Gretton retweeted
⚡️Few-step diffusion without a teacher, self-distillation or JVPs.
iDDM: a single model trained from scratch in one stage. ImageNet-256: 4.48 FID at 4 steps and 2.38 at 50, from the same checkpoint
🧵 (1/6)
Arthur Gretton retweeted
🔥Welcome 🔺Simplex Diffusion Models🔻!
Standard discrete diffusion models suffer from "information collapse" because they discard uncertainty at intermediate steps by sampling categorical tokens. (1/5)
Arthur Gretton retweeted
We are thrilled to announce the 6th Conference on Causal Learning and Reasoning (CLeaR), taking place in Paris, France, from May 3-5, 2027!
📢 The Call for Papers for CLeaR 2027 is now open!
📅 Submission deadline: Nov. 16, 2026
For more details: cclear.cc/2027
Applying to #UCL for a PhD in AI? Then be sure to apply for ELLIS as well! A great opportunity to collaborate with leading European AI labs.
🎓 Ready for a PhD in AI/ML? The #ELLISPhD Program connects top students with leading researchers across Europe.
The application portal opens on 1 Oct.
Register now to join the virtual info session on 30 Sep.
All info at bit.ly/3TrH9CJ
Congratulations Dr @Hudson19990518 for an outstanding thesis defence!!
...and a big thank you to the thesis examiners @lqiang67 and Samuel Livingstone!
Thrilled to share that I passed my PhD viva on "MiniMMD: Minimization of MMD"! 🎓
A huge thank you to my examiners, @lqiang67 and Samuel Livingstone, for the thoughtful discussion, and to my supervisors, @ArthurGretton and @fx_briol , for the guidance and support throughout!
Bonferroni? Aggravate. 😠
Minimax with exact level? Aggregate! 😄
Upgrade any Bonferroni-based permutation test to SB:
* exact finite-sample level control,
* uniform power dominance,
* inheritance of all power guarantees, minimax optimality included!
New Aggregation Methods!
-SingleBatch SB
-TwoBatch TB
-DataDriven SB
-Sequential SB
Applications
-Multiple Hypothesis Testing
-Conformal Prediction
MMDAgg-SB HSICAgg-SB
-Exact level
-Higher power
-Admit p-value
-Minimax optimal
Schrab* Shah Gretton Kim*
arxiv.org/abs/2607.15823
Statistical Tests of Dependence and Conditional Dependence!
Why does your conditional dependence test get so many false positives? What can you do about it?
Talk at the #ICML2026 workshop on hypothesis testing
testing.ml/
Saturday 11 July, 11:20-12:05
Room 318
Variational Learning for Insertion-based Generation
✨Spotlight poster✨ #ICML2026
HALL A #2402
Tue, Jul 7, 2026 • 10:30 AM – 12:15 PM
Zhang, @zhexwang , Ying, @david_van_dijk Titsias, @thjashin
Dr MMD is at #ICML2026!
(De)-regularized Maximum Mean Discrepancy Gradient Flow
HALL A #4302
Wed, Jul 8, 2026 • 2:30 PM – 4:15 PM
@Hudson19990518 Mustafi @PierreGlaser @Korba_Anna Sriperumbudur
Interventional Processes for Causal Uncertainty Quantification, at #ICML2026 !
HALL A #4302
Wed, Jul 8, 2026 • 5:00 PM – 6:45 PM
@DanceHugh Orbanz
Outcome Aware Spectral Feature Learning for Instrumental Variable Regression, at #ICML2026
HALL A #4413
Tue, Jul 7, 2026 • 10:30 AM – 12:15 PM
@DimitriMeunier1 Wornbard @vkostic30 @antoine_mln Fröhlich Lounici @MPontil
Arthur Gretton retweeted
After a few years of paper rejections and visa issues, I finally made it to ICML.
If you’re in Seoul, feel free to come say hi—I’d be happy to chat and share our work.
Posters: July 8, 14:30 (S4) & 17:00 (S5)
#icml
Covers:
Outcome bridge proxy methods (kernel, NN)
arxiv.org/abs/2105.04544
arxiv.org/abs/2106.03907
Proxy Methods for Domain Adaptation
arxiv.org/abs/2403.07442
Treatment bridge and doubly robust proxy methods:
arxiv.org/abs/2503.08371
arxiv.org/abs/2505.19807