🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Conformal prediction provides valid prediction regions for any machine learning model, any data distribution and and any dataset size. But what determines the size of prediction sets and can this itself size be predicted? I turns out that the answer is yes 🔥🔥🔥🔥 as shown in

We are recruiting a Postdoctoral Research Assistant working with Prof Patrick Rebeschini. Grade 7: £35,308-£43,155 per annum, fixed term for 24 months. stats.ox.ac.uk/jobs 📸Vivacious Mel Photography

Congratulations to Dr Neil Laws, who has won a 2023 MPLSOxford Teaching Award for his pivotal role in numerous teaching innovations in recent years. A very well deserved award. stats.ox.ac.uk/news/neil-laws…

The Statistics and Learning Theory Summer School 2023 started yesterday. Amazing mini-courses taught by Evgenii Chzhen, Maxim Panov, Patrick Rebeschini and Martin Takac. Gorgeous place and highly motivated participants.

I was so proud to do this speech today. Solidarity ✊

Excited that our work ‘From Denoising Diffusions to Denoising Markov Models’ has been accepted to JRSS-B for discussion! In it, we generalise diffusion models to arbitrary spaces. arxiv.org/abs/2211.03595 w/ Yuyang Shi Valentin De Bortoli George Deligiannidis Arnaud Doucet 🧵1/n

This is my meal for tonight because all I can afford is donated potatoes and a tin of beans since University of Brighton stole 100% of my wages for not marking 3 dissertations. This is how lecturers are treated here, which UCU Brighton UCU are trying to change. Help Jack Monroe

Very pleased that our work obtaining the first linear convergence bounds for diffusion models has been awarded a spotlight at ICLR! Valentin De Bortoli George Deligiannidis Arnaud Doucet

📰Our work on error bounds for flow matching methods has just been accepted to TMLR! We derive the first error bounds that apply in the case of fully deterministic sampling and data distributions without full support. w/ Arnaud Doucet George Deligiannidis arxiv.org/abs/2305.16860

Today we are thrilled to host George Deligiannidis for a fantastic joint Statistics/CIA seminar on the Quantitative Stability of the Iterative Fitting Procedure!

Is optimal transport always optimal? By learning ground costs (ICNNs+flows) one can structure OT maps to use prior information Brilliant PhD student, Samuel Howard, will be presenting Cost-Parameterized Monge Maps at the Differentiable AE workshop differentiable.xyz/papers-2024/pa…

What if the next medical breakthrough is hidden in plain text? Causal estimates drives progress but data is limited & RCTs slow. Introducing NATURAL: a pipeline for causal estimation from text data in hours, not years. Paper: tinyurl.com/ppr29 Site: tinyurl.com/web98

Congratulations to George Deligiannidis & Julien Berestycki on the award of Full Professor title! 🎉 They have been promoted through the University of Oxford Recognition of Distinction 2024 round. The title acknowledges their achievements in research, teaching & good citizenship.

Convergence of Diffusion Models Under the Manifold Hypothesis in High-Dimensions arxiv.org/abs/2409.18804… Jointly with George Deligiannidis and Judith Rousseau, we show that diffusion models achieve rates independent of the ambient dimension for score learning and sampling! (1/4)

We would like to invite new staff and students of the Department of Statistics to our department welcome drinks! 🍸 📆Date: 15.10.2024 ⌚️Time: 4.30pm 📍Location: Ground Floor Social Area We look forward to welcoming our new members to the department 🥳 #MT24

With Peter Potaptchik and George Deligiannidis we show the first realistic bound on the iteration complexity of diffusion models! Our work explains why sampling from ImageNet needs only ~100 (intrinsic dim) steps instead of ~150k (extrinsic dim). arxiv.org/abs/2410.09046

We have just arranged our next Distinguished Lecture with Arnaud Doucet more information can be found here: talks.cam.ac.uk/talk/index/228…

Attending ALT 2025 in Milan Come say hi! Presenting “Generalisation under gradient descent via deterministic PAC-Bayes” on Thu w/ Eugenio Clerico George Deligiannidis Benjamin Guedj Arnaud Doucet proceedings.mlr.press/v272/clerico25…