BEGIN:VCALENDAR PRODID:-//eluceo/ical//2.0/EN VERSION:2.0 CALSCALE:GREGORIAN BEGIN:VEVENT UID:cdab43a58a41744ef6ff78fcdd67fe50 DTSTAMP:20260923T212727Z SUMMARY:Applied Statistics Workshop DESCRIPTION:02/10/2026 - 14:30 - ISBA C115 - \; \;Consulting Showca se: Revisiting a project from the 2025/26 CourseResults from the collabora tion between the Consulting Class of 2025/26 with the Belgian Foreign Serv iceWith participation by Mohamed Achraf Saidi\, Maxime Dejean\, Lucie Rasc ol\, Gauss Dountsop\, Christian Ritter (UCLouvain) and Lev Lhommeau (SPF A ffaires Etrangères) \;In person in room C-115\, 20 voie du Roman Pays \, UCLouvain\, Louvain-la-Neuve."Generating Narratives for Development Pro jects"This seminar will consist of two parts.Part 1: 14h30-15h30In part 1\ , we will discuss the problem of creating\, checking and delivering narrat ive texts for international transparency of development projects. Lev Lhom meau from the Belgian Foreign Service will explain the context of the tran sparency documentation needs. We will then define and discuss approaches t o handle this need by automatic text generation processes using available tabular and textual information. Key points in this process are text acqui sition using web scraping and targeted post processing\, uses of classical NLP to assemble relevant elements in the acquired texts and tabular infor mation and the generation of prototype narratives. In a third stage\, we w ill discuss uses of generative AI to improve these prototypes. This involv es a discussion of security and authenticity. Security (avoiding egress of internal information to generate texts) is handled using local implementa tions of a small LLM. Authenticity is handled by carefully crafting contex t grounded prompts and by checking against common patterns of confabulatio n (such as adding commonly known information about international events.Th e work has yielded a prototype application on which the work group at the Foreign Service can continue building and adapting.Part2: 16h-17hAfter the coffee break\, we shall demonstrate how versions of these approaches can work in practice.Participants will have ample opportunity to ask questions and give suggestions. \; URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261002T143000 DTEND;TZID=Europe/Brussels:20261002T170000 LOCATION:ISBA - C115 (1st Floor) 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:1f6b2138b6a55c7676a48e40e24a51a5 DTSTAMP:20260923T212727Z SUMMARY:LIDAM Statistics Seminar by Kayvan Sadeghi DESCRIPTION:25/09/2026 - 14:30 - ISBA C115 - \; \;Kayvan Sadeghi &n bsp\;(University College London) \;Will give a presentation on : \ ;Characterising \;and Identifying Graphical Causal ModelsAbstract:&nbs p\;Using a characterization of faithfulness\, we describe the foundational graph orientation rule in constraint-based causal structure learning and the assumptions under which it recovers the “correct” graph. We presen t the theory for the class of directed acyclic graphs (DAGs) and then gene ralize it to the broader class of mixed graphs\, which can encode independ ence structures arising from feedback\, latent variables\, and selection m echanisms. We also introduce a general and computationally efficient struc ture learning algorithm for mixed graphs. URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20260925T143000 DTEND;TZID=Europe/Brussels:20260925T153000 LOCATION:ISBA - C115 (1st Floor) 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:1e7bf9cc4e69e7d3926ecda2b8c26511 DTSTAMP:20260923T212727Z SUMMARY:Applied Statistics Workshop by Marina Vives DESCRIPTION:More details coming soon URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261106T143000 DTEND;TZID=Europe/Brussels:20261106T153000 LOCATION:ISBA - C115 (1st Floor) 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:8f2c11ec2274dc0623f12137fd39ea6d DTSTAMP:20260923T212727Z SUMMARY:Applied Statistics Workshop by Aurelie Bertrand and Laura Symul DESCRIPTION:More details coming soon URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261127T143000 DTEND;TZID=Europe/Brussels:20261127T153000 LOCATION:ISBA - C115 (1st Floor) 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:d6e34e1de5d3cbb233e35f007a952e56 DTSTAMP:20260923T212727Z SUMMARY:LIDAM Statistics Seminar by Alessandra Luati DESCRIPTION:11/12/2026 - 14:30 - ISBA C115 - \; \;Alessandra Luati  \;(Imperial College London) \;More details coming soon URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261211T143000 DTEND;TZID=Europe/Brussels:20261211T153000 LOCATION:Voie du Roman Pays\, 20 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:6dc33f1c5d1ea47470ec3cc512e35f9c DTSTAMP:20260923T212727Z SUMMARY:LIDAM Statistics Seminar by Ivan Kojadinovic DESCRIPTION:27/11/2026 - 11:00 - ISBA C115 - \; \;Ivan Kojadinovic  \;(Université de Pau et des Pays de l'Adour) \;Will give a prese ntation on : \;An algorithmic procedure for solving the generalized mi nimum information checkerboard copula problemAbstract: \;The minimum i nformation copula principle (see Meeuwissen and Bedford\, 1997) is a maxim um entropy-like approach for finding the least informative copula that sat isfies a certain number of expectation constraints specified either from e xpert knowledge or the available limited data. In this presentation\, we f irst propose a generalization of this principle allowing the inclusion of additional constraints fixing certain higher-order margins of the copula. We next prove that the associated optimization problem has a unique soluti on under a natural condition. As the latter problem is intractable in gene ral\, following the existing literature\, we consider its version with all the probability measures involved in its formulation replaced by checkerb oard approximations. This amounts to attempting to solve aso-called discre te I-projection linear problem. We then use the seminal results of Csiszar (1975) to derive an IPFP-like procedure for solving the latter and provid e theoretical guarantees for its convergence. We next report numerical exp eriments in dimensions up to four with substantially finer discretizations than those encountered in the literature. We finally discuss the computat ional cost of the procedure and hint at memory saving implementations that should allow to tackle higher dimensional problems under certain conditio ns. URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261127T110000 DTEND;TZID=Europe/Brussels:20261127T120000 LOCATION:Voie du Roman Pays\, 20 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:2e66442457dcb050853b3fb7152a12b1 DTSTAMP:20260923T212727Z SUMMARY:LIDAM Statistics Seminar by Jing Zhou DESCRIPTION:09/10/2026 - 14:30 - ISBA C115 - \; \;Jing Zhou(Univers ity of Manchester) \;Will give a presentation on : \;Dense High-Di mensional Huber Regression under HeteroscedasticityAbstract: \;We stud y Huber regression in a heteroscedastic linear model when the number of pr edictors is comparable to the sample size. We show that the first-order ce nter of the Huber estimator may contain an additional component along the direction governing the conditional scale\, and we derive the resulting as ymptotic center and normalized risk. A conditional Huber-centering conditi on is shown to be sufficient for this nuisance-direction component to vani sh. The analysis uses matrix approximate message passing to track the regr ession and scale directions simultaneously. The talk will introduce the st atistical mechanism behind the result\, explain the connection with classi cal sandwich variance formulas\, and outline how state evolution yields an exact high-dimensional characterization. No prior knowledge of approximat e message passing is required. URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261009T143000 DTEND;TZID=Europe/Brussels:20261009T153000 LOCATION:ISBA - C115 (1st Floor) 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:0780ea2869ac0f37237e8accf160be28 DTSTAMP:20260923T212727Z SUMMARY:LIDAM Statistics Seminar by Kellie Archer DESCRIPTION:13/11/2026 - 14:30 - ISBA C.115 - \; \;Kellie Archer&nb sp\;(The Ohio State University) \;Will give a presentation on : \; High-Dimensional Mixture Cure Models for Time-to-Event Data \;Abstract : \;Medical breakthroughs in recent decades have led to cures for many diseases including cancer.  \;For example\, various groups have shown that advances in therapy for leukemia and myelodysplastic syndrome have i ncreased overall survival rates and some groups have identified factors re lated to long-term survival\, defined as survival exceeding three years\, which included that patients treated on newer treatment regimens were more likely to be long-term survivors. In fact\, some argue that these improve d outcomes indicate that some AML patients can be considered “potentiall y cured.” The mixture cure model (MCM) is a time-to-event model that is used when a cured fraction exists. MCMs assume the population consists of two subgroups\, those cured will not experience the event of interest and those susceptible to the event of interest. Therefore\, “cured” can be considered synonymous with attaining long-term relapse-free survival. Thu s\, there are two regression components in MCMs which permit identificatio n of features associated with cure and/or latency of susceptible patients. Many researchers have sought to identify a prognostic model for a time-to -event outcome. When the covariate space is high-dimensional\, as in the c ase with gene expression data\, typically penalized Cox proportional hazar ds (PH) models are fit. However\, when a dataset includes a cured fraction \, MCMs are more appropriate than Cox PH models. In these scenarios\, Cox models often yield inaccurate hazard and survival estimates due to violati ons of the proportional hazards assumption. Despite their utility\, robust methods for fitting MCMs to high-dimensional data remain scarce. In this talk\, I will present our recent work extending MCMs to effectively handle high-dimensional covariate spaces\, offering a more accurate prognostic t ool for modern high-dimensional datasets. URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261113T143000 DTEND;TZID=Europe/Brussels:20261113T153000 LOCATION:Voie du Roman Pays\, 20 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:1a261391b028812c6a09250ad706684e DTSTAMP:20260923T212727Z SUMMARY:LIDAM Statistics Seminar by Thomas Jaki DESCRIPTION:16/10/2026 - 14:30 - ISBA C115 - \; \;Thomas Jaki   \;(Cambridge University) \;Will give a presentation on : \;Control ling false decision errors in platform trialsAbstract: \;Platform tria ls are a new class of clinical study design \;that allow treatment arm s to enter and leave the trial \;over time \;and have proven to be particularly popular during the COVID-19 pandemic where no fewer than 58 trials have been registered as a platform trial. A common feature of these designs is the desire to answer several research questions within a singl e protocol. \;One of the questions arising from this feature that has generated a lot of discussion in the literature is the need for (or lack o f need to) \;control error rates as well as the most appropriate \ ;type of error control.In this talk I will begin by reflecting on the disc ussion around error rates and introduce different possible testing strateg ies that could be considered for platform trials. \;Throughout this pr esentation I will highlight areas where further research is necessary to e nable platform trials to unleash their full potential. URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261016T143000 DTEND;TZID=Europe/Brussels:20261016T153000 LOCATION:Voie du Roman Pays\, 20 1348 Louvain-la-Neuve END:VEVENT BEGIN:VEVENT UID:666a5638979c0ed70f164dc916105e09 DTSTAMP:20260923T212727Z SUMMARY: EDT short course by Margaux Zaffran DESCRIPTION:29-30/10/2026 - 09:30 -  \; \;Margaux Zaffran \;(La boratoire de Mathématiques d’Orsay) \;The course will be held at IS BA\, UCLouvain (20 Voie du Roman Pays\, Louvain-la-Neuve) on:Thursday Octo ber\, 29th\, 2026\, 9h30-12h30 and 13h30-15h (Room C115)\;Friday October\, 30th\, 2026\, 9h30-12h30 and 13h30-15h (Room C115).To better organize the event\, may we kindly ask you to register (before October 19th at 17h\; t he sooner the better\, so that we can reserve a larger room in case needed and the catering) using the following link:Workshop EDT on Conformal Pred iction – Remplir le formulaireFor those travelling by car\, if you need parking please let us know at least one week in advance (it takes at least a couple of days to get a QR code from the central administration). For t hose in need of a certificate of attendance\, please let us know as well i n advance.More details and information regarding the event will follow.Nev ertheless\, you are all warmly invited to already check out some slides th e speaker prepared for a similar event at SFdS (on which our short course is based) here:https://github.com/mzaffran/ECAS_SFdS_ConformalPredictionKi nd regards\,EDT in Statistics\, Biostatistics and Actuarial Sciences \ ; \; \; URL:https://uclouvain.be/en/calendar/isba DTSTART;TZID=Europe/Brussels:20261029T093000 DTEND;TZID=Europe/Brussels:20261030T150000 LOCATION:Voie du Roman Pays\, 20 1348 Louvain-la-Neuve END:VEVENT END:VCALENDAR