PAC-Bayesian Computations are learning and certification strategies based on PAC-Bayes bounds. Have been used for neural network classifiers, showing promising results. They are interesting, among other reasons, because they ensure generalisation by design. (Restrictions apply.)

👋 Meet the ELLIS Board! This ep features Samuel Kaski, Director of the ELLIS Institute Finland 🇫🇮 & Prof at @AaltoUniversity 🇫🇮 & @OfficialUoM 🇬🇧. He answers rapid-fire questions about AI and recommends this paper: https://t.co/LgI4jx5vau Watch now 👉

Don’t use the ICML template unless your paper is published at ICML. Maybe I’m old fashioned…

🏹 Job alert: Postdocs/research fellows in probabilistic machine learning at ELLIS Institute Finland & Manchester Centre for AI Fundamentals 📍 Helsinki 🇫🇮 & Manchester 🏴󠁧󠁢󠁥󠁮󠁧󠁿 📅 Apply by Dec 1st 🔗

🔥 WANTED: Student Researcher to join me,@ValentinDeBort1,@thjashin,@liwenliang,@ArthurGretton in DeepMind London. You'll be working on Multimodal Diffusions for science. Apply here

📣 #ICML tutorials: We want to know what *you* would like to learn. This year, Adam White and I are calling for nominations of topics and/or presenters. Until December 7th, you can send us your suggestions, and we will use them to shape the program. https://t.co/IwWgHtOGFn

📣 Heads-up 📣 As in previous years, TMLR will pause new submissions over the upcoming holiday period from December 2 2025 to January 5 2026 (midnight AOE on both dates). We will resume accepting new submissions on January 6, 2026. Happy Holidays!

Here's a shot 📸 of Sean Meyn's talk:

Also, had a fun blackboard discussion with the great Sean Meyn, I share a few pictures 📷 playing with the blackboard signs 😄 Was very good to see Michal Valko again! @misovalko

Those 2 pictures were borrowed from this 𝕏 post: https://t.co/YWhzY3ONOp

Was great to meet many excellent colleagues from around the UK and overseas, presenting their research on problems at this fascinating intersection ✨

✍ Earlier this week I was at the Isaac Newton Institute for Mathematical Sciences in Cambridge, attending the workshop "Bridging Stochastic Control and Reinforcement Learning: Theories and Applications" @NewtonInstitute 🔗

What makes PhD students happy? Good supervision Supervisors who invest in positive mentoring relationships with their PhD candidates also reap the benefits for their own research. https://t.co/qPKKYHm0i8

Find the differences 😉

🧠 The 'Bridging Stochastic Control and Reinforcement Learning' workshop is well underway! Check out the fantastic group of productive participants gathered here this week, collaborating on cutting-edge research. #IsaacNewtonInstitute #TheAlanTuringInstitute #NewtonSCL2025

This 👇🏻

#Halloween2025 Really scary.

“The bad review will come from your list of suggested reviewers”

biggest problem in multimodality: each modality is a better condition for itself than the others. diffusion will pay attention to the partially denoised image and ignore your prompt. vlm will pay attention to the first half of the sentence and ignore your image.

Booked my trip to Cambridge for this one

Intriguing results of a PhD student experience survey across a number of countries. Brazil and Australia scored high on "satisfaction" and UK didn't fall too far behind. Mentoring reported as very important feature, and so was quality of human connections https://t.co/SPh3hzJpqY