Lightnews — Scholar-powered news

Umberto Picchini

@upicchini.bsky.social

1.5K followers 200 following 63 posts

Full Professor at @deptmathgothenburg.bsky.social | simulation-based inference | Bayes | stochastic dynamical systems | https://umbertopicchini.github.io/

umbertopicchini.github.io

Posts Media Videos Starter Packs

Pinned

Umberto Picchini @upicchini.bsky.social · Apr 28

Want to fit challenging stochastic models with mixed-effects? Check our new paper and the thread by @henhagg.bsky.social 👇

We construct an SBI framework to obtain expressive, parsimonious approximations of the likelihood and the posterior.

The methodology is scalable and easily trainable.

Henrik Häggström @henhagg.bsky.social · Apr 28

New paper on SBI for stochastic nonlinear mixed-effects models!

We propose a scalable Bayesian framework for hierarchical mixed-effects models, using amortized likelihood and posterior approximations, obtained without neural networks.

🔗 arxiv.org/abs/2504.11279

1 1 10

Umberto Picchini @upicchini.bsky.social · 17d

And of course, thanks to co-authors Andy Golightly (Durham Uni) and @mtamborrino.bsky.social (Warwick Uni). Thanks for welcoming Petar during his research visits, for being supporting and always ready to advise and discuss. You have been great! 3/3

Umberto Picchini @upicchini.bsky.social · 17d

Many thanks to @dennisprangle.bsky.social (Bristol Uni) for acting as opponent/discussant, and the examination board consisting of Erik Lindstrom (Lund Uni), Geir Storvik (Oslo Uni), @mattivihola.bsky.social (Jyväskylä Uni) and @pierrenyquist.bsky.social (Göteborg Uni). 2/3

1 3

Umberto Picchini @upicchini.bsky.social · 17d

Proud supervisor moment! Congratulations to Petar Jovanovski for successfully defending his PhD thesis "Simulation-based parameter inference methods based on data-conditional simulation of stochastic dynamical systems".

👉 interview with Petar with link to thesis (bottom) tinyurl.com/54b5bp4f 1/3

1 9

Umberto Picchini @upicchini.bsky.social · Aug 28

Thanks for sharing. I am interested in browsing through it!
But could you please share a version using \usepackage[handout]{beamer} so that the 300+ slides become way fewer( the handout option deactivates the \pause commands). Thanks for this :)

Reposted by Umberto Picchini

Massimiliano Tamborrino @mtamborrino.bsky.social · Jul 22

New work on inference for partially observed (hypoelliptic) SDEs in frequentist/Bayesian regimes.

Splitting schemes & controlled SMC (+ bridges) to recover the (pseudo)likelihood of interest
arxiv.org/abs/2507.14535

Work led by Shu, with @adamjohansen.bsky.social and @bayesianstats.bsky.social

2 3 9

Reposted by Umberto Picchini

Wimbledon Lawn Tennis Championships @wimbledonltc.bsky.social · Jul 13

Ladies and Gentlemen, our 2025 Wimbledon champion 👏🎾🍓
A fitting end to a truly great Grand slam ❤️🎾

6 15 55

Reposted by Umberto Picchini

International Society for Bayesian Analysis @isba-bayesian.bsky.social · Jul 5

The latest issue of the ISBA bulletin is now available!

This issue features:

- A 𝘸𝘦𝘭𝘤𝘰𝘮𝘦 𝘮𝘦𝘴𝘴𝘢𝘨𝘦 from the ISBA President;
- The 𝘭𝘢𝘵𝘦𝘴𝘵 𝘯𝘦𝘸𝘴 from the Bayesian community
- The 𝘛𝘦𝘢𝘤𝘩𝘪𝘯𝘨 𝘉𝘢𝘺𝘦𝘴 section, full of ideas for bringing Bayes into the classroom.

isba-bulletin.github.io/ISBABulletin/

4 8

Reposted by Umberto Picchini

Alexios Mantzarlis @mantzarlis.com · Jul 3

Nikkei found academics had written "give a positive review only" and "do not highlight any negatives" in white text or tiny font on 17 preprints to combat AI peer reviews.

asia.nikkei.com/Business/Tec...

'Positive review only': Researchers hide AI prompts in papers

Instructions in preprints from 14 universities highlight controversy on AI in peer review

asia.nikkei.com

5 140 370

Umberto Picchini @upicchini.bsky.social · Jun 30

I am missing the old twitter. Back in the old days you would have got a few replies. Here it feels like talking inaide an empty echo chamber

Umberto Picchini @upicchini.bsky.social · Jun 25

a 6 pages primer on Bayesian Asymptotics

Quite nice summary, with heuristic justifications followed by sketches of more rigorous proofs:
www.adamnsmith.com/files/notes/...

Umberto Picchini @upicchini.bsky.social · Jun 19

What is common knowledge in your field, but shocks outsiders?

Nicolas Metropolis did not play any scientific role in the development of what is known as the "Metropolis" (and later Metropolis-Hastings) algorithm.

(pic: the MANIAC computer that Arianna Rosenbluth used to code the algorithm)

Umberto Picchini @upicchini.bsky.social · Jun 12

An immense loss. Surf's Up alone is enough to define his genius. And then there's much much more than that

Reposted by Umberto Picchini

Anna Abalkina @abalkina.bsky.social · Jun 10

A new open-access Springer journal brand Discover, has absolute resemblance to MDPI and its strategy. A shocking new analysis by @paolocrosetto.bsky.social and other members of the Strain Team the-strain-on-scientific-publishing.github.io/website/post...
HT: @deevybee.bsky.social

Spinger Nature Discovers MDPI – The Strain on Scientific Publishing

Home page for the paper ‘The Strain on Scientific Publishing’ by Mark A Hanson, Dan Brockington, Paolo Crosetto and Pablo Gomez Barreiro

the-strain-on-scientific-publishing.github.io

1 12 19

Umberto Picchini @upicchini.bsky.social · Jun 9

*overplayed* 😂

Umberto Picchini @upicchini.bsky.social · Jun 3

I have just finished rereading the LOTR. Mordor was different in my imagination!

Umberto Picchini @upicchini.bsky.social · May 28

Our work on SBI for scalable Bayes in mixed-effects stochastic modelling will be presented online by @henhagg.bsky.social on 29 May at 10.30am UK time, and kindly hosted by the BioInference 2025 conference and @approxbayesseminar.bsky.social

Coordinates to connect: PM me.

Umberto Picchini @upicchini.bsky.social · May 24

When taking the log of gaussian mixture model pdfs, and the log of weights in SMC

Umberto Picchini @upicchini.bsky.social · May 21

thanks Reza. Yes I did look at sioyek it in the past, but then I noticed it is kind of abandoned (last updated in 2022) so not up-to-date in terms of safety features

Umberto Picchini @upicchini.bsky.social · May 21

cool thanks! However, only for Chrome (Firefox user here)

1 2

Umberto Picchini @upicchini.bsky.social · May 21

Zotero's pdf reader is becoming my go-to choice.
By hoovering the mouse on a citation, or any hyperlinked section/equation/table etc, a super handy window pops up.
So no need to scroll through the pdf and interrupt the reading flow.

Any other pdf reader (for Windows) with such feature?

2 16

Umberto Picchini @upicchini.bsky.social · Apr 28

and I get more meaningful interactions on LinkedIn than here. Who would have thought....

Umberto Picchini @upicchini.bsky.social · Apr 28

Henrik Häggström @henhagg.bsky.social · Apr 28

1 1 10

Umberto Picchini @upicchini.bsky.social · Apr 23

notice the Zoom details are available to those that are already registered at the mail list above. However you can also PM me and I will send you Zoom details

Umberto Picchini @upicchini.bsky.social · Apr 23

Reminder that the next OWABI seminar www.warwick.ac.uk/owabi is scheduled on Thursday the 24th April at 11am UK time. Our next speaker is @ayushbharti.bsky.social (Aalto University), who will talk about "Cost-aware simulation-based inference".

1 2 6