Room 02.124, Building 5, St Priest campus Machine Learning in Montpellier, Theory & Practice - François Petit (Inserm) We developed a mathematical setup inspired by Buyse's generalized pairwise comparisons to define the notion of an optimal individualized treatment rule (ITR) in the presence of a prioritized outcomes in a randomized controlled trial, terming such an ITR pairwise optimal. We present two approaches to estimate pairwise optimal ITRs. The first is a variant of the k-nearest neighbors algorithm. The second is a meta-learner based on a randomized bagging scheme, allowing the use of any classification algorithm for constructing an ITR. We study the behavior of these estimation schemes from a theoretical standpoint and through Monte Carlo simulations and illustrate their use on trials data. Machine Learning in Montpellier, Theory & Practice

Online Machine Learning in Montpellier, Theory & Practice - Shubhankar Mohapatra (University of Waterloo) Many critical AI applications, such as personalized health assistants and social network recommender systems, rely on learning from private data. Differential privacy has become the gold standard for training models on sensitive data, gaining widespread adoption in industry and government. This growing adoption has fueled research in differentially private learning, yet significant challenges remain in its deployment. In this talk, I will discuss several challenges that arise when integrating differential privacy into an end-to-end learning pipeline, from data collection to model training. I will particularly focus on the difficulties caused by inconsistencies in private datasets, such as typos and missing values. Correcting these errors is especially challenging when direct access to the raw data is restricted due to privacy constraints. I will present two recent works addressing this issue. First, I will show how leveraging correlations and dependencies in the data lets us privately quantify inconsistencies, helping estimate data quality and the effort needed for data repair. Second, I will examine the impact of missing values in private learning and introduce simple yet effective techniques for generating differentially private synthetic data to mitigate these effects. Machine Learning in Montpellier, Theory & Practice

Amphithéâtre A (Bat. A) au campus de l’UFR STAPS de Montpellier Zaineb AJRA (EuroMov Digital Health in Motion) - Thèse AXIAUM (financement ANR) Sous la direction de Jacky MONTMAIN et Stéphane PERREY et l’encadrement de Binbin XU IA&Recherche, LabéliséHallesIA

Room 02.124, Building 5, St Priest campus Machine Learning in Montpellier, Theory & Practice - Batiste Le Bars (Inria) Conformal Prediction has recently been considered as one of the state-of-art technique to construct distribution-free prediction sets satisfying probabilistic coverage guarantees. In this talk, we study the question of volume optimality in split conformal regression. Using the fact that the calibration step can be seen as an empirical volume minimization problem, we first derive a finite-sample upper-bound on the excess volume loss of the interval returned by the classical split method. This quantity measures the difference in length between the interval obtained with the split method and the shortest oracle prediction interval. Then, we introduce EffOrt, a methodology that modifies the learning step so that the base prediction function minimizes the length of the returned intervals. In particular, our theoretical analysis of the excess volume loss of the prediction sets produced by EffOrt reveals the links between the learning and calibration steps, and notably the impact of the function class of the base predictor. We also introduce Ad-EffOrt, an extension of the previous method, which produces intervals whose size adapts to the value of the covariate. Machine Learning in Montpellier, Theory & Practice

Textmine Le groupe de travail Textmine vous invite à une journée exceptionnelle de rencontres et d’échanges entre acteurs du public et du privé autour des IA appliquées aux textes. Cet événement se tiendra le 02 juin 2025 à Montpellier, dans la salle du Conseil de l’Hôtel de Métropole (50 place Zeus – 34000 Montpellier), à seulement 10 minutes à pied de la gare. Vous êtes chercheur, entreprise, startup ou acteur public ? Venez présenter vos travaux (courtes interventions ou démos) ou simplement assister aux échanges ! Dans les deux cas, l’inscription est requise.

Campus St Priest (860 Rue Saint Priest 34095 Montpellier Cedex 5), bat. 5, Room: 02.124 Machine Learning in Montpellier, Theory & Practice - Olivier Cappé (CNRS / ENS / Université PSL) Performative learning addresses the increasingly pervasive situations in which algorithmic decisions may induce changes in the data distribution as a consequence of their public deployment. We propose a novel view in which these performative effects are modelled as push-forward measures. This general framework encompasses existing models and enables novel performative gradient estimation methods, leading to more efficient and scalable learning strategies. For distribution shifts, unlike previous models which require full specification of the data distribution, we only assume knowledge of the shift operator that represents the performative changes. Focusing on classification with a linear-in-parameters performative effect, we prove the convexity of the performative risk under a new set of assumptions. We also establish a connection with adversarially robust classification by reformulating the minimization of the performative risk as a min-max variational problem. IA&Experts

Room 02.124, Building 5, St Priest campus Machine Learning in Montpellier, Theory & Practice - Olivier Cappé () Performative learning addresses the increasingly pervasive situations in which algorithmic decisions may induce changes in the data distribution as a consequence of their public deployment. We propose a novel view in which these performative effects are modelled as push-forward measures. This general framework encompasses existing models and enables novel performative gradient estimation methods, leading to more efficient and scalable learning strategies. For distribution shifts, unlike previous models which require full specification of the data distribution, we only assume knowledge of the shift operator that represents the performative changes. Focusing on classification with a linear-in-parameters performative effect, we prove the convexity of the performative risk under a new set of assumptions. We also establish a connection with adversarially robust classification by reformulating the minimization of the performative risk as a min-max variational problem. Machine Learning in Montpellier, Theory & Practice

Online Machine Learning in Montpellier, Theory & Practice - Julie Alberge (Inria) When dealing with right-censored data, where some outcomes are missing due to a limited observation period, survival analysis —known as time-to-event analysis — focuses on predicting the time until an event of interest occurs. Multiple classes of outcomes lead to a classification variant: predicting the most likely event, a less explored area known as competing risks . Classic competing risks models couple architecture and loss, limiting scalability. To address these issues, we design a strictly proper censoring-adjusted separable scoring rule, allowing optimization on a subset of the data because the evaluation is conducted independently for each observation. The loss estimates outcome probabilities and enables stochastic optimization for competing risks, which we use for efficient gradient boosting trees. SurvivalBoost not only outperforms 12 state-of-the-art models across several metrics on 4 real-life datasets, both in competing risks and survival settings, but also provides great calibration, the ability to predict across any time horizon, and faster computation times compared to existing methods. Machine Learning in Montpellier, Theory & Practice

Room 02.124, Building 5, St Priest campus Machine Learning in Montpellier, Theory & Practice - Ilyass Moummad (Inria) Biodiversity monitoring has become increasingly popular as a way to better understand and assess the health of surrounding ecosystems. Sound, in particular, plays a crucial role in biodiversity monitoring, as many species rely on vocalizations to communicate and navigate their environment. Recent advancements in deep learning have improved our ability to automatically detect and classify sounds. However deep learning often require large annotated datasets, which can be costly and time-consuming for manual labelling, and difficult to obtain for rare or endangered species. To address these challenges, we explore the potential of transfer learning, where a feature extractor trained on one task is adapted to a new task. Specifically, we investigate invariant learning as a pre-training strategy designed to produce representations that are invariant to predefined transformations. In the unlabelled setting, this involves mapping a data example and its transformed counterpart to similar latent representations (transformation invariance), whereas in the labelled setting, examples sharing the same annotations are mapped to similar representations (label invariance). We then apply these learned representations to downstream tasks, focusing on the detection and classification of animal vocalizations in few-shot scenarios. Machine Learning in Montpellier, Theory & Practice

Online Machine Learning in Montpellier, Theory & Practice - Jim Smit (Erasmus Medical Center, Rotterdam) Background Despite several randomised controlled trials (RCTs) on the use of adjuvant treatment with corticosteroids in patients with community-acquired pneumonia (CAP), the effect of this intervention on mortality remains controversial. We aimed to evaluate heterogeneity of treatment effect (HTE) of adjuvant treatment with corticosteroids on 30-day mortality in patients with CAP. Methods In this individual patient data meta-analysis, we included RCTs published before July 1, 2024, comparing adjuvant treatment with corticosteroids versus placebo in patients hospitalised with CAP. The primary endpoint was 30-day all-cause mortality, collected across all trials, and analyses followed the intention-to-treat principle. We analysed HTE using risk and effect modelling. For risk modelling, patients were classified as having less severe or severe CAP based on the pneumonia severity index (PSI), comparing PSI class I–III versus class IV–V. For effect modelling, we trained a corticosteroid-effect model on six trials and externally validated it using data from two trials, received after model preregistration. This model classified patients into two groups: no predicted benefit and predicted benefit from adjuvant treatment with corticosteroids. The literature search was registered on PROSPERO, CRD42022380746. Findings We included eight RCTs with 3224 patients. Across all eight trials, 246 (7·6%) patients died within 30 days (106 [6·6%] of 1618 in the corticosteroid group vs 140 [8·7%] of 1606 in the placebo group; odds ratio [OR] 0·72 [95% CI 0·56–0·94], p=0·017). The corticosteroid-effect model, which selected C-reactive protein (CRP), showed significant HTE during external validation in the two most recent trials. In these trials, 154 (11·4%) of 1355 patients died within 30 days (88 [13·1%] of 671 in the placebo group vs 66 [9·6%] of 684 in the corticosteroid group; OR 0·71 [95% CI 0·50–0·99], p=0·044). Among patients predicted to have no benefit (CRP ≤204 mg/L, nr5), no significant effect was observed (OR 0·98 [95% CI 0·63–1·50]), whereas for those with predicted benefit (CRP >204 mg/L, nc0), 39 (13·0%) of 301 patients died in the placebo group compared with 20 (6·1%) of 329 in the corticosteroid group (0·43 [0·25–0·76], pinteraction=0·026). No significant HTE was found between less severe CAP (PSI class I–III, n"9) and severe CAP (PSI class IV–V, n26). Corticosteroid therapy significantly increased hyperglycaemia risk (44 [12·8%] of 344 in the placebo group vs 84 [24·8%] of 339 in the corticosteroid group; OR 2·50 [95% CI 1·63–3·83], p<0·0001) and hospital re-admission risk (30 [3·7%] of 814 in the placebo group vs 57 [7·0%] of 819 in the corticosteroid group; 1·95 [1·24–3·07], p=0·0038). Interpretation Overall, adjuvant therapy with corticosteroids significantly reduces 30-day mortality in patients hospitalised with CAP. The treatment effect varied significantly among subgroups based on CRP concentrations, with a substantial mortality reduction observed only in patients with high baseline CRP. Full text: https://isdm.umontpellier.fr/authors.elsevier.com/a/1kWkq7tFB1XhtL" Machine Learning in Montpellier, Theory & Practice

Faculté des Sciences Campus Triolet bâtiment 16 salle SC.16.01. Juan Luis GASTALDI (ETH Zürich) L'impact des modèles neuronaux de langage dans les aspects les plus divers des pratiques sociales est si massif que l'élaboration d'une perspective critique à leur égard est devenue urgente. Pourtant, la complexité réputée de ces dispositifs et des principes formels qui les animent ne laisse souvent d'autre choix que celle d'une critique externaliste, entraînant presque invariablement leur dépréciation en tant qu'instruments de savoir sans qu'une véritable analyse épistémologique puisse avoir lieu. Ici je voudrais défendre l'idée qu'il est possible d'échapper à cet écueil tout en défendant une position critique. Cette perspective permettra, dans un deuxième temps, de proposer une analyse alternative des principes formels qui sous-tendent l'efficacité des réseaux neuronaux pour le traitement du langage, révélant des structures algébriques sous-jacentes aux objets statistiques qui les caractérisent.

Cité de l’Économie et des Métiers de Demain Cet événement marquera le lancement officiel de la Maison du Quantique en Occitanie dont l'Université de Montpellier est partenaire au travers de l'ISDM. Cette journée sera un moment privilégié pour renforcer les collaborations, échanger sur les projets en cours et explorer de nouvelles opportunités dans le domaine.

Campus Triolet - Amphithéâtre Bat. 20 ISDM Rejoignez-nous le 30 avril à partir de 15h sur le campus Triolet de l’Université de Montpellier pour célébrer la mise en service du nouveau équipement de calcul/cloud ISDM-MESO / Drocc-Est ! Grâce au financement obtenu dans le cadre du CPER par l’Etat, la Région Occitanie, la Métropole de Montpellier et l’Université de Montpellier, l'ISDM renouvelle ses infrastructures de calcul et vous invite à découvrir cette nouvelle configuration de pointe et les services associés. Un cocktail suivra la présentation. Inscription gratuite mais obligatoire - présence possible à distance.