- Random - Mailing list — Dipartimento di Matematica, UniPI

Mini-course by Dario Trevisan at Tor Vergata, September 24-25
by Michele Salvi 16 Sep '26

16 Sep '26

Dear all, We are glad to announce that next week RoMaDS <https://www.mat.uniroma2.it/~rds/events.php> will host at the University of Rome Tor Vergata the following mini-course: September 24-25 (10h00-12h00 and 13h30-15h30 on both days) Department of Mathematics, Aula Dal Passo Dario Trevisan (Università di Pisa) From Gaussian Limits to Feature Learning in Wide Neural Networks Abstract: Wide neural networks can be seen as random functions whose behavior simplifies in several different large-width regimes. At the scale of typical fluctuations, central-limit phenomena lead to Gaussian-process limits with recursively defined covariance kernels, while the parameter-space tangent geometry converges to the neural tangent kernel and yields a linearized description of training. These Gaussian and kernel limits provide a remarkably tractable theory of random initialization and learning, but they capture only part of the finite-width behavior. The course develops a probabilistic view of random wide neural networks across these different scales. We will study Gaussian and quantitative finite-width approximations, neural tangent kernel limits, large deviations of outputs and empirical layer kernels, and regimes in which feature learning survives at large width. Large-deviation theory provides a complementary description of rare, non-Gaussian behavior: atypical outputs can be realized through atypical internal kernels, leading to variational representations of the output rate function. Under suitable Bayesian scalings, the same viewpoint yields posterior variational problems over both predictors and kernels, providing one probabilistic mechanism for data-dependent feature learning. We will compare this with dynamical feature-learning limits such as mean-field and maximal-update regimes, emphasizing throughout how different scalings reveal different aspects of neural networks. We encourage in-person partecipation. Should you be unable to come, here is the link to the Teams streaming: Mini-corso Trevisan | Join meeting in Teams | Microsoft Teams <https://teams.microsoft.com/meet/35005768873253?p=IK9DCp8bgD0QL9P7PK> See you soon in Tor Vergata! This events are part of the Excellence Project MatMod@TOV.

1 0

Workshop Malliavin Calculus and Applications
by Stefania Ugolini 16 Sep '26

16 Sep '26

Dear Colleagues, It is our great pleasure to invite you to the following workshop. Best regards, Stefania Ugolini on behalf of the Organizing Committee Workshop: Malliavin Calculus and its Applications Thursday 17 September 2026 Milano, Aula 208, Via Celoria 20. Schedule: 10h-11h Giulia Di Nunno (Oslo University), Malliavin calculus in Sandwich Volterra Volatility models 11h15-12h15 Vlad Bally (Université Gustave Eiffel), Integration by parts and convergence in distribution norms in the CLT. 14h30-15h30 Sonia Mazzucchi (Università di Trento), Feynman path integrals for Schrödinger equation with magnetic field: a Malliavin Calculus approach. The workshop is organized by Francesco De Vecchi (Università di Pavia), Frédéric Patras (CNRS-Université Cote d’Azur) e Stefania Ugolini (Università di Milano) as the first scientific event of the Intensive Research Trimester titled: "Algebraic & Geometric Structures in Stochastic Calculus" of the Center for Training and Research, in the framework of the Excellence Project 2023-2027 of the Department of Mathematics " “F. Enriques”, Milano. Siamo nati per chiederci perché. Il tuo 5xmille all'Università degli Studi di Milano Codice fiscale 80012650158 [https://sites.unimi.it/loghi/5x1000/5x1000.png]

1 1

SPASS seminar: Lorenzo Marino - Tue 15 Sept at 15.00 CET
by Andrea DI PRIMIO 14 Sep '26

14 Sep '26

Dear Colleagues, We are pleased to invite you to the following SPASS seminar https://sites.google.com/unipi.it/spass/seminars?authuser=0 <https://sites.google.com/unipi.it/spass/seminars?authuser=0%3E> (jointly organised by UniPi, SNS, UniFi and UniSi): Speaker: Lorenzo Marino (Scuola Normale Superiore, Pisa) Title: Stochastic Transport by Heavy-Tailed Jump Processes Date and time: Tuesday, September 15th at 15:00 CET Place: Aula Mancini, Palazzo della Carovana, Scuola Normale Superiore, Pisa Abstract: In the study of turbulent fluid dynamics, modeling the velocity field as a superposition of divergence-free vector fields with stochastic weights provides crucial insights into the effect of small-scale turbulence on large-scale dynamics. Indeed, under suitable conditions, such transport noise can produce an additional dissipation effect on dynamics via a certain scaling limit. While previous results have mainly focused on the diffusive scaling regime (i.e. when the random weights are described by Gaussian or other light-tailed noises), I will present some recent results, exploring the advection of a passive scalar by a turbulent fluid when the driving noise exhibits heavy-tailed jump statistics. On behalf of the organisers, Andrea Di Primio

1 0

AI and Mathematics
by Faggionato, Alessandra 14 Sep '26

14 Sep '26

Buongiorno, I’m sharing a link where you can sign (if you're interested) a document entitled *“A Severe Misalignment of AI in Mathematics,”* which has already been signed by several Fields Medal winners: https://mathandai.org/ I apologize for any repetitions that I may have overlooked. Grazie dell'attenzione Saluti Alessandra -- ****************************************************************************** Prof. Alessandra Faggionato https://sites.google.com/mat.uniroma1.it/alessandrafaggionato/home Department of Mathematics University "La Sapienza" Piazzale Aldo Moro, 5 00185 - Rome Office 123, first floor ******************************************************************************

1 0

SPASS seminar: Francesco Russo - Tue 15 Sept at 14.00
by Lorenzo Marino 11 Sep '26

11 Sep '26

Dear Colleagues, We are pleased to invite you to the following SPASS seminar <https://sites.google.com/unipi.it/spass/seminars?authuser=0> (jointly organised by UniPi, SNS, UniFi and UniSi): _Speaker_: *Francesco Russo* (ENSTA Paris, Institut Polytechnique de Paris) _Title_: *Exponential twist of probability measures: drift correction in term of a generalized gradient* _Date and time_: *Tue, September 15th* at *14:00 CET* _Place_: Aula Mancini, Palazzo della Carovana, Scuola Normale Superiore, Pisa On behalf of the organisers, Lorenzo Marino ********************* _Abstract_: \This talk will concern the exponential twist, i.e. a path-integral exponential change of measure, of a Markovian reference probability measure P. This type of transformation naturally appears in variational representation formulae originating from the theory of large deviations and can be interpreted in some cases, as the solution of a specific stochastic control problem. Under a very general Markovian assumption on P, we fully characterize the exponential twist probability measure as the solution of a martingale problem and prove that it inherits the Markov property of the reference measure. The ”generator” of the martingale problem shows a drift depending on a generalized gradient of some suitable value function v. Applications of this work refer to an entropy minimization algorithm, see e.g. [1] and [2].\ \This work is based on a collaboration with Th. Bourdais (ENSTA and Mazars), and N. Oudjane (EDF). It is the object of [3].\ *References* [1] T. Bourdais, N. Oudjane, and F. Russo. An entropy penalized approach for stochastic control problem. Complete Version. Preprint HAL-04193113 v3, 2025. [2] T. Bourdais, N. Oudjane, and F. Russo. An entropy penalized approach for stochastic control problem. SIAM Journal on Control and Optimization (SICON), 64 (1):363–386, 2026. [3] T. Bourdais, N. Oudjane, and F. Russo. Exponential twist of probability measures: drift correction in term of a generalized gradient. Stochastic Processes Appl., 201:28, 2026. Id/No 105064.

1 0

Recent events and the Mathathon
by Ottobre, Michela 11 Sep '26

11 Sep '26

Dear colleagues, many of you will have seen this post from OpenAI, which appeared on September 8th https://openai.com/index/navier-stokes-solution/ The letter of Tristan Buckmaster, Clay Award for his work on Navier Stokes, https://cims.nyu.edu/~tristanb/statement.pdf tells the unsettling story of how OpenAI came to "solve" the millennium problem (if that is indeed the case). I prefer not to give a short version of the story, the letter speaks for itself. I would also like to flag up below the link to a petition against the Mathathon. Context (from the form below): On October 30th, 2026, Caltech will begin hosting the Caltech Mathathon<https://mathathonchallenge.com/>, a math hackathon in which participants are encouraged to solve open research problems by prompting LLMs. Two prominent AI companies, Anthropic and OpenAI, will supply participants with 2 million dollars in AI credits. This event is likely to have destructive impacts for the mathematical community. AI companies see research mathematics as an advertising opportunity. Over the past few months, they have advanced a campaign of scientific misinformation about the goals of mathematical research. Major AI companies are engaged in an arms race to declare themselves the first to prove prominent open conjectures. New announcements of AI-generated results appear on social media on a weekly basis. I would add, destructive impact on the environment as well. Link to the petition below. This is not just about whether you like using AI or not, and it is not even "just about" the impact that big companies are trying to have on our profession. This is about the impact on science, on the values we choose as foundations for advancing human knowledge and progress. And about being respectful of Planet Earth. https://docs.google.com/forms/d/e/1FAIpQLSf6dBeDwjRP4UkchGSpLwLQjiG8qLX5_wW… Another useful link is here https://www.ahmath.org/<https://urldefense.com/v3/__https://www.ahmath.org/__;!!DZ3fjg!91iL4UmnI4yy…> Best Regards, Michela -------------------------------------- Michela Ottobre Professor of Mathematics Heriot-Watt University https://www.macs.hw.ac.uk/~mo3/ My working hours are flexible. I do not expect a reply outside of your normal working hours. The School of Mathematical and Computer Sciences (MACS) at Heriot-Watt University is currently accepting applications to our PhD programmes, see www.findaphd.com<https://www.findaphd.com/phds/program/phd-opportunities-in-the-school-of-ma…>. I have a postdoc position available in my group and I am interested in hiring Phd students. Info at https://www.macs.hw.ac.uk/~mo3/ ________________________________ Founded in 1821, Heriot-Watt is a leader in ideas and solutions. With campuses and students across the entire globe we span the world, delivering innovation and educational excellence in business, engineering, design and the physical, social and life sciences. This email is generated from the Heriot-Watt University Group, which includes: 1. Heriot-Watt University, a Scottish charity registered under number SC000278 2. Heriot- Watt Services Limited (Oriam), Scotland's national performance centre for sport. Heriot-Watt Services Limited is a private limited company registered is Scotland with registered number SC271030 and registered office at Research & Enterprise Services Heriot-Watt University, Riccarton, Edinburgh, EH14 4AS. The contents (including any attachments) are confidential. If you are not the intended recipient of this e-mail, any disclosure, copying, distribution or use of its contents is strictly prohibited, and you should please notify the sender immediately and then delete it (including any attachments) from your system.

1 0

Seminario di Claudio Landim alla Sapienza
by Faggionato, Alessandra 11 Sep '26

11 Sep '26

Buongiorno, Claudio Landim (IMPA, Rio de Janeiro) farà un seminario Martedi' 15 Settembre alle ore 14.00 in Sala di Consiglio del Dipartimento di Matematica G. Castelnuovo alla Sapienza. Sotto titolo e abstract Saluti Alessandra Titolo: Non-equilibrium fluctuations of gradient exclusion processes in dimension [image: d\leq 3] Abstract: We consider a gradient, speed-change, symmetric exclusion process on the discrete torus [image: \mathbb{T}^d_n], [image: d\le 3], whose empirical measure evolves, on the diffusive scale, according to a non-linear parabolic equation. We prove that, starting from a sequence of initial states whose relative entropy with respect to the inhomogeneous product measure associated with the initial profile is [image: o(n^{d/2})], and whose density fluctuation field converges, the density fluctuation field converges, in [image: L^1(0,T; H_{-r})], to the solution of a linear stochastic partial differential equation with time-dependent coefficients driven by a conservative space-time white noise. -- ****************************************************************************** Prof. Alessandra Faggionato https://sites.google.com/mat.uniroma1.it/alessandrafaggionato/home Department of Mathematics University "La Sapienza" Piazzale Aldo Moro, 5 00185 - Rome Office 123, first floor ******************************************************************************

1 0

JOB: 3-years PhD position in probabilistic machine learning and statistics, University of Oslo (Norway)
by Valeria Vitelli 10 Sep '26

10 Sep '26

Dear all, We are recruiting for a fully funded, three-year PhD position in probabilistic machine learning and statistics at the Oslo Centre for Biostatistics and Epidemiology (OCBE) <https://www.med.uio.no/imb/english/research/centres/ocbe/>, University of Oslo, Norway. We are once more looking for motivated candidates to enlarge our research team focusing on Bayesian unsupervised learning for structured data, and to work at the intersection of machine learning, probabilistic modelling, and Bayesian inference. The project aims at developing new methods for analysing complex biomedical data with latent dependencies, combining machine learning models with efficient Bayesian inference techniques, such as variational approximations. Candidates with solid background in Statistics, Data Science, Machine Learning, Mathematics, or Computer Science are highly encouraged to apply. The successful candidate will join OCBE’s lively and supportive research community, which includes around 15 PhD students. The position offers a competitive salary and the opportunity to live in Oslo, with easy access to stunning nature (forests, the fjord, and ski slopes), and very favourable welfare conditions particularly with strong support for families. We would be grateful if you could share this opportunity with interested master’s students and recent graduates from your network. Interested candidates are very welcome to contact us directly to discuss the project before applying. *Application deadline: *4th October 2026 *Start date: *Flexible *More info and how-to-apply: * https://www.jobbnorge.no/en/available-jobs/job/307053/3-years-phd-position-… *Contacts:* • William Denault: williade(a)uio.no • Valeria Vitelli: valeria.vitelli(a)medisin.uio.no. Thank you for helping us share this opportunity. Best regards, William and Valeria -- Valeria Vitelli Oslo Centre for Biostatistics and Epidemiology, Department of Biostatistics, University of Oslo, Norway group page: https://www.med.uio.no/imb/english/research/groups/statistics-high-dimensio… <https://www.med.uio.no/imb/english/research/groups/statistics-high-dimensio…> mailto: valeria.vitelli(a)medisin.uio.no

1 0

[STAR Seminars] Thursday September 17 - Josep Vives
by David Vencato 10 Sep '26

10 Sep '26

Dear all, Join us for our upcoming talk in the STAR seminar<https://www.mn.uio.no/math/english/research/groups/risk-stochastics/star-se…> series on Thursday, Septermber 17th, from 11:00 to 12:00 (Oslo time). As always, the talk will take place in a hybrid format: participants in Oslo can attend the talk in Room 723 in Niels Henrik Abels hus, whereas the international audience will be able to follow the talk via Zoom<https://uio.zoom.us/j/64362132069>. The speaker is Josep Vives (Universitat de Barcelona) with the talk: Title: Option price asymptotics under a stochastic volatility Lévy model with infinite activity jumps Abstract: We apply techniques of Malliavin-Skorokhod calculus for Lévy processes, including a suitable Hull-White type formula, to study the short-time asymptotics of the vanilla option price in the at-the-money (ATM), in-the-money (ITM) and out-of-the-money (OTM) scenarios, under a very general Lévy stochastic volatility model with infinite activity jumps. We are looking forward to seeing you! Best regards, Giulia, Salvador, Leonardo, Pere, and David

1 0

Seminario Fabio Centofanti, KU Leuven | Polimi - 11.09.2026 - 14:00.
by Simone Vantini 10 Sep '26

10 Sep '26

Cari Colleghi, A seguire annuncio di un seminario organizzato presso il Laboratorio MOX - Dipartimento di Matematica, Politecnico di Milano Fabio Centofanti, KU Leuven, Belgium 11.09.26 ore 14:00 – Aula Saleri, sesto piano Dipartimento di Matematica (Edificio 14) Link Zoom: https://polimi-it.zoom.us/j/97446037096 Titolo: Robust Multivariate Analysis under Cellwise and Casewise Contamination Abstract: Robust statistics has traditionally focused on casewise outliers, meaning observations that behave differently from the bulk of the data. In modern datasets, however, contamination can occur in the form of cellwise outliers, which are individual contaminated entries in the data matrix. Even when the proportion of contaminated cells is relatively small, many observations may still contain at least one outlying value, which can make both classical and casewise robust methods unreliable. This talk presents recent developments for handling situations in which casewise and cellwise outliers appear simultaneously. We begin by introducing cellPCA, a novel robust principal component analysis method that estimates the principal subspace while simultaneously accommodating cellwise and casewise outliers, as well as missing values. Building on cellPCA, we then introduce cellRCov, which is a robust covariance estimator for high dimensional settings. The second part of the talk focuses on multivariate linear regression. The cellMR estimator combines robust covariance estimation with regularization to handle outliers in both the predictor and response matrices. We then introduce cellBoot, a bootstrap procedure based on indirect inference that complements robust estimation with valid confidence intervals for the regression coefficients. We conclude with a perspective on more complex data structures and contamination patterns, including tensor data, time series, functional data, and blockwise contamination. Un caro saluto a tutti, dal Gruppo di Statistica del MOX ____________ Simone Vantini Professor of Statistics MOX Laboratory for Modeling and Scientific Computing Dept. of Mathematics - Politecnico di Milano Piazza Leonardo da Vinci 32, 20133, Milano, Italy

1 0