Omar Rivasplata

Booked my trip to Cambridge for this one

Are these the happiest PhD students in the world? Brazil, Australia and Italy have the highest satisfaction scores in Nature’s global 2025 PhD survey — but are these nations really the best places to do a doctorate?

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. Quality of mentoring reported as very important feature, and so was quality of human connections
www.nature.com/articles/d41...

Reminder of the preferred approach when sharing arXiv preprint links

PAC-Bayesian Reinforcement Learning Trains Generalizable Policies We derive a novel PAC-Bayesian generalization bound for reinforcement learning that explicitly accounts for Markov dependencies in the data, through the chain's mixing time. This contributes to overco...

📢 New preprint alert - This work with collaborators is on Reinforcement Learning + PAC-Bayes for Markov Chain data. Had been in the oven for some months and received feedback from NeurIPS'25. We uploaded the preprint in arXiv for visibility and comments would be welcome!
arxiv.org/abs/2510.10544

Meet the candidates for the 2025 RSS Council elections

If you haven't already, there's still time for members to vote in our Council elections and have your say in our future! 🗣️

Meet this year's candidates 👉 rss.org.uk/news-publica...

✉️ Check your email for your personalised voting link.

Voting closes midnight on Sunday 12 October.

IMA The Institute of Mathematics and its Applications is the UK's chartered professional body for mathematicians and one of the UK's learned societies for mathematics.

Mathematics Today dropped through the letterbox this morning. Contains a cute piece on quantum computing illustrated with maths topics taught in A-levels (UK school system). Recommendable for those who teach maths to connect with the trendy topic!
ima.org.uk

Reference for the Gibbs variational principle/dual characterization of KL Let $P,Q$ be two probability distributions. Then, one has the following dual characterization of their Kullback-Leibler divergence (relative entropy): $$ D(P \,\|\,Q) = \sup_f ( \mathbb{E}_P[f(X)]...

"So, it apparently is not proven in the Donsker—Varadhan paper from 1983, but instead in Lemma 2.1 from their 1975 paper (first in the series, not the fourth)" -- C. Canonne
@ccanonne.github.io
math.stackexchange.com/questions/34...

Asymptotic evaluation of certain markov process expectations for large time. IV Click on the article title to read more.

TIL The reference for Donsker-Varadhan variational formula is Donsker and Varadhan's "Asymptotic evaluation of certain Markov process expectations for large time-IV" (1983), not the previous parts I, II, III from years 1975 and 1976 that are commonly cited
onlinelibrary.wiley.com/doi/abs/10.1...

Good Morning!

"All models are wrong, but some are useful.". --George E. P. Box
x.com/probnstat/st...

a man wearing suspenders and a watch looks down in front of a machine Alt: Alarn Turing looks down in front of a machine

In 1950, Alan Turing - while at @manchester.ac.uk - proposed a simple but profound experiment: can a machine convincingly imitate human thought?

We now call it the Turing Test. 🧠💻

Greetings from Stockport

I have experience with Entropy MDPI
I'm convinced this is a predatory publisher

In case everybody didn't notice, the Royal Statistical Society's Computational Statistics and Machine Learning (CSML) section now has a bluesky account: @rss-csml.bsky.social

Newcastle Non-Reversible Sampling Workshop Topic: This workshop will bring together researchers at the forefront of modern sampling techniques and related disciplines, to discuss recent advances in the field. This event is funded by the ProbA...

Earlier this week attended the Non-Reversible Sampling workshop at Newcastle University. A very nice event!

Thanks to the organisers Sam Power (Bristol), Giorgos Vasdekis (Newcastle) and Andi Wang (Warwick)
sites.google.com/view/probai-...

@aifunmcr.bsky.social

We also will continue running our Research Forum which is a very good venue for PhD students, postdocs and others to present in a friendly setting

A roster of high-profile invited speakers is shaping up, stay tuned and look out for exciting talks coming up on exciting topics, mainly around Fundamentals of Applied AI, from theory and methodology to applications of impact

✨ We kicked off a new season of talks, the AI FUN with ELLIS Invited Speaker Lecture series, brought to you by the Centre for AI FUNdamentals and the #ELLIS Unit at Manchester 🐝

The 1️⃣st talk was given by our guest speaker Rafael Izbicki (@rafael-izbicki.bsky.social) from UFSCar in Brazil 🇧🇷

Our 2025 Council Elections are now open!

Fellows can vote until 12 Oct to help shape the future of the Society.

🔗 Check your email for your voting link

📄 Meet the candidates: rss.org.uk/news-publica...

Just finished delivering a course on 'Robust and scalable simulation-based inference (SBI)' at Greek Stochastics. This covered an introduction to SBI, open challenges, and some recent contributions from my own group.

The slides are now available here: fxbriol.github.io/pdfs/slides-....

The institute's opening is part of the AI Summit, with distinguished speakers from across European business and government, as well as top researchers and leadership from the @ellis.eu network. An exciting two-day program that marks a milestone for AI & ML research. More info at aisummit.fi

The Fast and Curious 2: MCMC in Action September 25-27, Toronto, Canada Aims: Recent strides in Bayesian inference applied to intricate complex-structured datasets offer a unique opportunity to bring together disparate research communit…

Are you curious about what makes some Markov chain Monte Carlo samplers run fast? Then perhaps you will consider joining us in Toronto, Sept 25-27 for the "Fast and Curious 2: MCMC in action" conference. Registration information is available at:
raducraiu.com/the-fast-and...

I want to advertise some relatively recent work which I really like, and have been fortunate to play a small role in.

The paper is titled "A New Proof of Sub-Gaussian Norm Concentration Inequality" (arxiv.org/abs/2503.14347), led by Zishun Liu and Yongxin Chen at Georgia Tech.

The past, present, and future of computer intelligence | The University of Manchester Research. Teaching and learning. Social responsibility. Discover more about The University of Manchester here.

ICYMI The University of Manchester's 2025 Magazine is out. There is plenty of interesting content in there. Don't miss taking a look at the opinion piece article on the past, present, and future of computer intelligence

👉 www.manchester.ac.uk/about/magazi...

The image shows an illustrated pair of hands holding open a colorful book filled with abstract, dreamlike visuals such as a moon in space, musical notes with glowing orbs, and the Earth.

The 2025 Magazine is out now!

Featuring:
- @duncanivison.bsky.social on Manchester’s bold future
- Simon Johnson on the journey to his Nobel Prize
- Dr Leo Geyer revives music composed more than 80 years ago in Auschwitz

Read it now! 👇
www.manchester.ac.uk/about/magazi...

ICYDK what clouds are made of

I am pleased to announce that together with some friends, we are organising a workshop on Non-Reversible MCMC Sampling, taking place at Newcastle University from 8–10 September 2025.

Details on the programme and registration can be found at the workshop website (sites.google.com/view/probai-...).

OCEAN OCEAN is an ERC Synergy grant gathering researchers to lay the fundations of the next-generation machine learning algorithms, involving decentralized inference and predictive tasks in the presence …

If anyone is, or knows anyone who is, looking for a postdoc position to work on the foundations of scalable inference and such... applications for a postdoctoral position at Warwick as a part of the OCEAN project (oceanerc.com) are open until 20th August at warwick-careers.tal.net/vx/appcentre....

BitVI on 1D Gaussian mixture models.

Remember that computers use bitstrings to represent numbers? We exploit this in our recent @auai.org paper and introduce #BitVI.

#BitVI directly learns an approximation in the space of bitstring representations, thus, capturing complex distributions under varying numerical precision regimes.

🚨 A few days left to apply for Research Associate and Research Fellow positions in probabilistic machine learning with the Manchester Centre for AI Fundamentals.

The deadline is 27.July.2025⌛️