Nicola Branchini
@nicolabranchini.bsky.social
🇮🇹 ProbAI Research Fellow @warwickstats.bsky.social. Previously @ellis.eu Stats PhD @edinunimaths.bsky.social @aalto.fi. 🤔💭 about Monte Carlo, approximate inference, UQ
"We're all worried," as what it means to do research (in my field, Theoretical CS) seems to be shifting, and shifting fast. What to do? Senior researchers must lead by example, knowing that not everything will pan out. What I'm suggesting below may not work everywhere, but here's my own advice: 1/
Fantastic paper demonstrating how LLM editing is gradually distorting our writing and the way we think to write: arxiv.org/abs/2603.18161
How LLMs Distort Our Written Language
Large language models (LLMs) are used by over a billion people globally, most often to assist with writing. In this work, we demonstrate that LLMs not only alter the voice and tone of human writing, b...
arxiv.org
So much masochism in academia. We try to find reasons to reject (I do try not to). We decide to reject a paper for reason X & Y when multiple, very closely related work has been published at the same venue (or better for some, whatever that means), where X&Y could have equally applied as criticism
Federic Perlino, who is currently in the second year of a PhD working with Theo Damoulas and I has just arxived the first paper arising from it: arxiv.org/abs/2607.09645. In which he develops models for function composition over graphical structures using Gaussian processes.
Deep Gaussian Processes on Directed Acyclic Graphs
Many real-world processes can be represented as compositions of functions along a directed acyclic graph (DAG). In causal modelling, these correspond to the underlying mechanisms; in engineering, to m...
arxiv.org
I'm so happy studying and learning stuff, especially when I can momentarily disassociate from various sources of anxiety/stress. It's just like playing a good videogame.
“Springer about to hijack Statistics & Computing” xianblog.wordpress.com/2026/06/28/s...
Springer about to hijack Statistics & Computing
I recently learned that Springer Nature is about to make the reference journal Statistics and Computing, where I published close to twenty papers over the years, fully “open access”, wh…
xianblog.wordpress.com
"Topic X is well-studied [...]" true, but are people *studying* those works for the well-studied X recently? getting doubts by reading some papers...
Mehdi defends our recent preprint on rho-posteriors and their variational approximations at ISBA @Nagoya Link to the preprint: arxiv.org/abs/2601.07325
With friends at the University of Warwick (in particular, Rocco Caprio and @adriencorenflos.bsky.social), we've recently arXived some work (arxiv.org/abs/2605.30253) on a method for approximate inference known as "Coordinate Ascent Variational Inference", or "CAVI" for short. Let me explain:
Things I am not interested in, re: your research talk - how prestigious is the venue where you / the related work published - how much stuff you know / you did Things I am interested in: - the core ideas (explained with intent to teach, not impress, ideally) - how it fits within broader landscape
🧵 Preprint alert! Introducing Augmented Gaussian sum filters (AGSF), a novel class of Bayesian filtering algorithms which unifies Gaussian sum (GSF) and particle filters (PF) by interpolating continuously between them, while being robust to common failure modes. arxiv.org/abs/2605.21698
A Gaussian Sum Filter for Unifying Gaussian and Particle Filters
State-space models (SSMs) are a broad class of probabilistic models for dynamical systems with many applications in engineering and science. Bayesian filtering is analytically tractable only in the li...
arxiv.org
MCQMC 2026 program looking full of interesting stuff maths.ed.ac.uk/events/mcqmc...
Programme | MCQMC 2026 | School of Mathematics
Schedule and book of abstracts
maths.ed.ac.uk
"Accept (spotlight)" at ICML'26 😎 Our paper brings particle filters back to life: autoregressive diffusion models + posterior sampling yield optimal proposals for Bayesian filtering, scaling up to GenCast-sized systems. arxiv.org/abs/2605.20028 w/ Thomas Savary and @francois-rozet.bsky.social
Training-Free Bayesian Filtering with Generative Emulators
Bayesian filtering is a well-known problem that aims to estimate plausible states of a dynamical system from observations. Among existing approaches to solve this problem, particle filters are theoret...
arxiv.org
ICML Conference@icmlconf.bsky.social · 3mo agoCongrats again to authors of accepted #ICML2026 papers! The camera-ready deadline is 5/28. Drawing your attention to two specific features: 1. As last year, to help communicate research to a broad audience, papers will have lay summaries. Tips & details in blog 1/3
I hate reading [something] "models" p(x) or even worse versions: p(x|y), p_{θ}(x|y)
one less in the paper pipeline…. thanks paper gods
Very happy to share that our paper has been accepted at ICML @icmlconf.bsky.social ! This is joint work with Sahel Mohammad Iqbal, Simo Särkkä, and Dominik Baumann. The preprint is available below, and we will update the details once the official version is published. arxiv.org/abs/2511.04403
Online Bayesian Experimental Design for Partially Observed Dynamical Systems
Bayesian experimental design (BED) provides a principled framework for optimizing data collection by choosing experiments that are maximally informative about unknown parameters. However, existing met...
arxiv.org
Lena Zellinger, Nicola Branchini, Lennert De Smet, V\'ictor Elvira, Nikolay Malkin, Antonio Vergari: How to Approximate Inference with Subtractive Mixture Models https://arxiv.org/abs/2604.16714 https://arxiv.org/pdf/2604.16714 https://arxiv.org/html/2604.16714
Very excited to announce that the #BayesianWorkflow book by @statmodeling.bsky.social, @avehtari.bsky.social, @rmcelreath.bsky.social et al publishes in June! routledge.com/9780367490140 #RStats #DataScience #Bayesian
Regarding claims of Monte Carlo methods working or not working in high dimensions, I think I've officially read and seen everything and its opposite (sometimes in the same paper)
I’ve recently started as a Research Fellow at @warwickstats.bsky.social, working with Gareth Roberts within the Probabilistic AI hub. I’m looking forward to learning and doing some stats in a place with so much interesting work going on!
I am a bit late to the party, but I am happy to share that our latest work was accepted to #ICLR2026 🥳🥳 📜 How to Square Tensor Networks and Circuits Without Squaring Them arxiv.org/abs/2512.17090
How to Square Tensor Networks and Circuits Without Squaring Them
Squared tensor networks (TNs) and their extension as computational graphs--squared circuits--have been used as expressive distribution estimators, yet supporting closed-form marginalization. However, ...
arxiv.org
ProbML 2026 (formerly AABI) invites submissions on probabilistic ML (both Bayesian and otherwise!), July 5 in Seoul (co-located with ICML). Website: probml.cc. Tracks: proceedings (PMLR), workshop, fast track. New focus includes applications in healthcare and climate! Submit by: 20 March 2026.
4 ICLR papers 🥳 There’s an insightful story between them: If you sample LLMs multiple times, they are calibrated, even on higher levels [1], but they cannot talk about this uncertainty in a single prompt [2], so you have to help them out to gather information Bayes-optimally [3]
Postdoc in Milan on scalability for high-dimensional Bayesian learning statmodeling.stat.columbia.edu/2026/02/02/p...
Postdoc in Milan on scalability for high-dimensional Bayesian learning | Statistical Modeling, Causal Inference, and Social Science
statmodeling.stat.columbia.edu
Another exciting workshop to be held at Warwick Statistics this coming June: The ProbAI Theory of Scaling Laws Workshop from 22-24 June! Website of the workshop: warwick.ac.uk/fac/sc... Registration open until 31 March on a first-come-first-serve basis: warwick.ac.uk/fac/sc....
I want to advertise the PhD thesis of my good friend and luckily also collaborator Thomas Guilmeau theses.hal.science/tel-05474635/ I've learnt so much talking to Thomas about divergence minimization.
Divergence-minimization for variational inference, black-box global optimization, and importance sampling
Many methods across applied mathematics aim at constructing specific parametric probability distributions. Examples of these tasks include evolution strategies or simulated annealing for black-box global optimization, Monte Carlo methods based on adaptive importance sampling, and variational inference algorithms in machine learning. The construction of such distributions can often be formulated as the minimization of a statistical divergence. However, these divergence-minimization problems are challenging because of the following reasons. First, the specific geometry of the considered set of parametric probability distributions needs to be taken into account. Second, efficient evaluations of statistical divergences come with important noise. Third, divergence-minimization problems are generally non-convex. Because of these difficulties, standard methods may fail to converge to good solutions. We tackle these challenges in this thesis.First, we show that evolution strategies, which are sampling-based algorithms for black-box global optimization problems, can be analysed through the lens of divergence-minimization problems. Our approach allows to establish and quantify the improvement brought at each iteration of the algorithms. We show that existing methods fit within our framework, yielding a new approach for their analysis. We also establish improvement results for two novel algorithms, one related with mixture models, and another one using heavy-tailed parametric probability distributions.Second, we consider the minimization of a regularized Rényi divergence over an exponential family. We propose to solve this problem with a stochastic Bregman proximal-gradient algorithm, with biased gradient estimator. By leveraging the geometry of the exponential family, we prove strong convergence guarantees for our algorithm, with proof techniques that are of interest beyond the considered problem. We then extend this algorithm to propose an adaptive simulated annealing algorithm with solid theoretical understanding. We show through a rigorous benchmarking that our algorithm outperforms similar non-adaptive algorithms.Finally, we go beyond exponential families and look at variational inference problems over lambda-exponential families. Using generalized convexity tools, we give new sufficient optimality conditions for these problems, which generalize existing similar results for the exponential family. For the resolution of these problems, we propose novel proximal-like algorithms that exploit the geometry underlying the lambda-exponential family. These results are especially useful for heavy-tailed distributions. We then leverage our results to propose an adaptive importance sampling algorithm to cover these cases. We show that our algorithm is able to learn Student distributions that capture the location, scale, and tail behaviour of target distributions, both in heavy-tailed and light-tailed cases.
theses.hal.science
I disagree with the view that peer review isn't problematic just because your papers usually get accepted. Big difference between being accepted and being accepted for the right reasons (nevermind having proper feedback)
Bayesian Workflow by Andrew Gelman, Aki Vehtari, @rmcelreath.bsky.social with @danpsimpson.bsky.social, @charlesm993.bsky.social, @yulingy.bsky.social, Lauren Kennedy, Jonah Gabry, @paulbuerkner.com, @modrakm.bsky.social, @vianeylb.bsky.social (in production, estimated copy-editing time 6 weeks)
I hate when people refrain from giving me blunt feedback on my work out of politeness. I really want to know if you don't see the point 😄 I promise you can't hurt my feelings. This doesn't happen often, but more so at conferences than anywhere else.
