Archive for seminar

Bayesian decision-theory for data privacy [surfin’ the Oce’n, 30 April, INRIA Paris]

Posted in Statistics, University life with tags , , , , , , , , , , , , , , , , , on April 23, 2025 by xi'an

Abstract

The scientific and economic value of data continues to grow alongside technology advances. New hardware and software developments enable, but often require, larger and more complex datasets to function effectively. As the importance of input data to these systems becomes increasingly recognized, so too does the loss of privacy for data providers. In this context, data privacy emerges as a critical issue for fields such as statistics and machine learning, as well as for scientific and industrial endeavours that rely on sensitive data. We propose a framework for measuring privacy from a Bayesian decision-theoretic perspective. This framework enables the creation of new, purpose-driven privacy principles that are rigorously justified, while also allowing for the assessment of existing privacy definitions through decision theory. We pay particular attention to the privacy of deterministic algorithms, which are overlooked by current privacy standards, and to the privacy of N Monte Carlo samples drawn from an invariant distribution as N goes to infinity. We show that Probabilistic Differential Privacy is a special case of our framework and provide some new interpretations for Differential Privacy as a result.

gradient flow for projected Langevin dynamics

Posted in Books, Statistics, University life with tags , , , , , , , , , , , , , , on April 7, 2025 by xi'an

Daniel Lacker (Columbia U) gave a talk at the probability seminar of Paris Dauphine this week which I happened to attend by happenstance, on a recent paper, Projected Langevin dynamics and a gradient flow for entropic optimal transport, written with Giovanni Conforti and, Soumik Pal. The talk was quite progressive and I hence could follow most of it. The core idea is in studying Langevin-type diffusion dynamics that sample from an entropy-regularized optimal transport, i.e. looking for an optimal distribution (in the sense of achieving entropy minimisation problem within a Wasserstein space, with regularisation) obtained via a gradient flow equation (as eg in variational inference) that couples two SDEs that are recentred by conditional expectation terms. Expectations in the equations are estimated by a Nadaraya-Watson estimate in optimal transport problem (reminding me of SMC), with no theoretical derivation of an optimal bandwidth, and they achieve quantitive bounds on the convergence, namely for exponential convergence, energy decay and new logarithmic Sobolev inequalities. From the talk and a quick glance at the paper, it is unclear to me there are direct algorithmic consequences, since the SDEs need be discretised, while the expectation approximations are costly, being repeated at each iteration of the discretised SDE.

mostly MCMC and definitely April¹¹

Posted in pictures, Statistics, University life with tags , , , , , , , , , , , , , on April 6, 2025 by xi'an

 

The next mostly MCMC seminar is coming. It will take place at 3PM (CEST) on Friday 11th April 2025 at the PariSanté Campus (in Room 08). The talks are

  • Susan Wei (Monash University) “Dynamical versus Bayesian Transitions in a Toy Model of Superposition”
  • Emma Kopp (Université Paris Dauphine, PSL), “Contrasting modes of cultural evolution: Kra-Dai languages and weaving technologies”

     

when open publishing is not fair [PCI webinar, 20 March, 4pm CET]

Posted in Books, University life with tags , , , , , , , , , , , on March 15, 2025 by xi'an

9th seminar of the PCI webinar series

Mandatory registration using this link: https://univ-cotedazur.zoom.us/meeting/register/rz8ZKReZQeqs-3yR3S6q4g

When Open Publishing Is Not Fair

Sabina Leonelli (Technical University of Munich, TUM)

Summary: There are obvious ways in which Open Access has augmented inequity rather than mitigating it, for instance in relation to Author Publishing Costs and the differential access that researchers based in academic institutions around the world may have to publishing deals and packages. Less obvious but equally fundamental are inequities in the access to infrastructures, skills and information fostering an effective use of online resources ranging from Open Access journals to Open Data infrastructures. Most importantly, openness as a paradigm of “sharing” is predicated on a model of research practice that does not fit well with most domains and methods of research, and particularly with science done in low-resourced environments. I reflect on these issues and draw on examples and cases emerging from the PHIL_OS project (“A Philosophy of Open Science for Diverse Research Environments”; www.opensciencestudies.eu ), as well as my experiences as Open Science advocate and participant in Open Access debates over the last ten years.

Speaker’s bio: After high school in Italy, Sabina Leonelli studied History and Philosophy of Science at University College London (BSc Hons, 2000) and London School of Economics (MSc, 2001). She earned a PhD from Vrije Universiteit Amsterdam (2007), while research assistant for Hasok Chang and attending the Dutch graduate schools for STS and philosophy. After coming back to LSE to work with Mary Morgan (2006-2008), she moved to the Centre for the Study of the Life Sciences at the University of Exeter, which she directed from 2013 to 2024. She was appointed TUM Professor on 09/24. Building on philosophical, historical and social science methods and collaborations with scientists and policy-makers, Sabina Leonelli studies: (1) the role of technology, data and organisms in knowledge production, and especially how computing and digitalisation efforts are transforming research and its social dynamics and roles; and (2) the institutionalisation of Open Science as a window on the methods, epistemology and political economy of contemporary scientific inquiry, particularly in the life, biomedical and environmental sciences.

What is the PCI webinar series?
-What is it? Seminars on research practices, publication practices, evaluation, scientific integrity, meta-research.
-How does it work? Remote conferences using zoom with registration.
-For whom is it? For anyone interested in scholarly publication, all PCI users, all PCI recommenders who do preprint evaluations for PCI, authors of articles, etc.
-When is it? Once a quarter
-Why is it for? To learn about scholarly publishing, to improve our knowledge about scholarly review, to become better reviewers, to create a sense of community among PCI users.

Find details about the PCI webinar series and past seminars at https://peercommunityin.org/pci-webinar-series/

mostly [14] M[ar]C[h] seminar

Posted in Books, Statistics, University life with tags , , , , , , , , , , , , , , , , , on March 8, 2025 by xi'an