Antonin Schrab

@AntoninSchrab

Doctor in Foundational AI @ai_ucl & @GatsbyUCL. Kernel methods, hypothesis testing, generative models.

London, England

antoninschrab.github.io

Joined February 2011

475 Following

555 Followers

Replies Reposts Media

Pinned
Antonin Schrab
@AntoninSchrab
Jul 23
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
Aggregation of Statistical Evidence under Exchangeability
We study aggregation of statistical evidence under unknown and potentially complex dependence using group-invariance. Building on permutation-based constructions that treat transformed datasets as...
2
Antonin Schrab
@AntoninSchrab
Jul 27
How to verify if a model has "unlearned" something? 🤔 Exciting to see our AISTATS 2026 paper Regularized f-Divergence Kernel Tests featured in Google Research blog post on Auditing Machine Unlearning! Joint work with Mónica Ribero & @ArthurGretton 👇
research.google
New framework for auditing machine unlearning
1
Antonin Schrab
@AntoninSchrab
Jun 17
Check out the power of MMDAgg and MMDFuse to "Catch, Adapt, and Operate: Monitoring ML Models Under Drift" #ICLR26 Workshop 🦾🚀
Arthur Gretton
@ArthurGretton
Jun 12
Optimized MMD for Detecting Distribution Shift! Video: slideslive.com/39064367/optim… and slides gatsby.ucl.ac.uk/~gretton/cours… from the #ICLR26 workshop sites.google.com/view/iclr-2026… Covers: jmlr.org/papers/v24/21-… and proceedings.neurips.cc/paper_files/pa…
Antonin Schrab
@AntoninSchrab
May 3
Check out our latest work at #AISTATS26 using any f-divergence for hypothesis testing!
Arthur Gretton
@ArthurGretton
May 2
Regularized f-Divergence Kernel Tests! At #AISTATS26, poster 65 Monday May 4th at 3 p.m. –f-div sensitive to localized differences –Automatic kernel choice Applications💻 –Physics Outlier Detection –Audit DP –Unlearning Monica Ribero @AntoninSchrab
Antonin Schrab
@AntoninSchrab
Feb 2
🥁Two-Sample Testing with ANY f-Divergence using Kernels🤯 –f-div sensitive to localized differences –Power separation f-div rates –Agg+FUSE over kernels+reg Applications💻 –Physics Outlier Detection –Audit DP –Machine Unlearning M Ribero+@ArthurGretton
arxiv.org
Regularized $f$-Divergence Kernel Tests
We propose a framework to construct practical kernel-based two-sample tests from the family of $f$-divergences. The test statistic is computed from the witness function of a regularized...

Antonin Schrab

Antonin Schrab’s posts