Interests

Extreme values, high dimensional statistics, statistical learning theory and applications.

Extreme value theory, Unsupervised learning for extremes (dimension reduction, clustering), Statistical learning theory for rare events, Supervised learning algorithms and statistical guarantees with extreme covariates or targets, Functional Data, Applications (environmental and industrial risks, anomaly detection)

Projects, grants

• 2024-2029: Project coordinator of the ANR project EXSTA (EXtremes, STatistical learning and Applications)
• 2021-2024: Scientific coordinator for ANR project T-REX coordinated by Clément Dombry
• 2019-2023: Scientific investigator for ANR project MELODY coordinated by Ronan Fablet
• 2015 Young researcher grant (`Projet Exploratoire de Premier Soutien', PEPS JCJC INS2I) for the projet "AGREED"

Working papers

• Clémençon, S., Huet, N., & Sabourin, A. Regular Variation in Hilbert Spaces and Principal Component Analysis for Functional Extremes. arXiv preprint arXiv:2308.01023.

• Huet, N., Clémençon, and Sabourin, A., On Regression in Extreme Regions, arXiv preprint arXiv:2303.03084 .

• Aghbalou, A., Bertail, P., Portier, F., & Sabourin, A., Cross Validation for Extreme Value Analysis. arXiv preprint arXiv:2202.00488 .

Publications

• Aghbalou, A., Portier, F, & Sabourin, A. (2024) Sharp error bounds for imbalanced classification: how many examples in the minority class? To appear, proceedings of AISTATS

• Aghbalou, A. , Portier, F., Sabourin, A., Zhou, C. (2024) Tail inverse regression for dimension reduction with extreme response. Bernoulli arXiv preprint arXiv:2108.01432.

• Aghbalou, A., Sabourin, A., & Portier, F. (2023). On the bias of K-fold cross validation with stable learners. Proceedings of AISTATS

• Clémençon, S., Jalalzai, S. Lhaut, H., Sabourin, A., & Segers, J. (2023). Concentration bounds for the empirical angular measure with statistical learning applications. Bernoulli arXiv preprint arXiv:2104.03966.

• Lhaut, S., Sabourin, A., & Segers, J. (2022). Uniform concentration bounds for frequencies of rare events. Statistics & Probability Letters, 189, 109610 .

• Drees, H. and Sabourin, A., Principal Component Analysis for Multivariate Extremes (2021), Electronic Journal of Statistics 15 (1), 908-943 preprint

• Jalalzai, H., Colombo, P., Clavel, C., Gaussier, E., Varni, G., Vignon, E., Sabourin, A. (2020). Heavy-tailed Representations, Text Polarity Classification \& Data Augmentation. NeurIPS 33 . link

• M. Chiapino, S. Clémençon, V. Feuillard, A. Sabourin. A Multivariate Extreme Value Theory Approach to Anomaly Clustering and Visualization (2020), Computational Statistics 35(2), 607-628 preprint

• Jalalzai, H., Clémençon, S., Sabourin, A., On binary Classification in Extreme regions, NIPS, 2018 preprint

• Chiapino, M., Sabourin, A., Segers, J. Identifying groups of variables with the potential of being large simultaneously. Extremes, 1-30 (2018) preprint

• Achab, M., Clémençon, S., Garivier, A., Sabourin, A., & Vernade, C. (2017), Max K-armed bandit: On the ExtremeHunter algorithm and beyond proceedings of ECML-PKDD, 2017 arXiv

• S. Clémençon, A. Gramfort, A. Sabourin and A. Thomas (2017) Anomaly Detection in Extreme Regions via Empirical MV-sets on the Sphere proceedings of AISTATS 2017

• Sabourin, A., Segers, J. (2017) Marginal standardization of upper semicontinuous processes. With application to max-stable processes Journal of Applied Probability 54.3 hal preprint

• Goix, N., Sabourin, A., Clémençon, S. (2017) Sparse representation of multivariate extremes with applications to anomaly detection Journal of Multivariate Analysis, 2017 hal preprint

• Chiapino, M., Sabourin, A.(2016). Feature clustering for extreme events analysis, with application to extreme stream-flow data ECML-PKDD 2016, workshop NFmcp2016 preprint

• Goix, N., Sabourin, A., Clémençon, S. (2016). Sparse Representation of Multivariate Extremes with Applications to Anomaly Ranking AISTATS 2016

• Sabourin, A., Renard, B. (2015). Combining regional estimation and historical floods: A multivariate semi-parametric approach with censored data. Water Resources reseach. preprint

• Goix, N., Sabourin, A., Clémençon, S (2015). Learning the dependence structure of rare events: a non-asymptotic study. Proceedings of the 28th Conference on Learning Theory (COLT) arXiv preprint

• Goix, N., Sabourin, A., Clémençon, S (2015). On Anomaly Ranking and Excess-Mass Curves. 18th International Conference on Artificial Intelligence and Statistics (AISTATS) arXiv preprint

• Sabourin, A. (2015). Semi-parametric modelling of excesses above high multivariate thresholds with censored data.
  Journal of Multivariate Analysis. arXiv preprint

• Sabourin, A. , Naveau, P. (2014). Bayesian Dirichlet mixture model for multivariate extremes: A re-parametrization.
  Computational Statistics and Data Analysis. HAL preprint

• Sabourin, A. , Naveau, P., Fougères, A.-L. (2013). Bayesian Model averaging for Multivariate extremes. Extremes. preprint

Students

• Florian Lamalle (2023- ...) , CIFRE - Renault, Telecom Paris-Master MVA), co-advised with Stéphan Clémençon and Vincent Feuillard.
• Nathan Huet (2021 - ...) , grant from ANR IA / chair DSAIDIS, master MDA Orsay), co-advised with Stéphan Clémençon
• Anass Aghbalou (2020 - ...), grant from the chair DSAIDIS, Ecole Centrale Lyon), co-advised with François Portier, Patrice Bertail
• Hamid Jalalzai, (2017-2020, grant from the chair MLBD, M2 Data science Ecole Polytechnique), co-advised with Chloé Clavel
• Robin Vogel (2017-2020) co-advised with Stéphan Clémençon
• Mastane Achab (2016-2020), co-advised with Stéphan Clémençon
• Maël Chiapino ( 2014- 2018, grant from the Labex LMH, master MVA), co-advised with François Roueff.
• Nicolas Goix (2013-2016, grant AMN, Ens Cachan/UPMC) co-advised with Stéphan Clémençon

Packages (R)

• BMAmevt : Bayesian model averaging of angular measures for multivariate extremes.
  Available on CRAN packages repository here .

• DiriXtremes : Dirichlet Mixture model for multivariate extremes, inference with unknown number of components : implementation of a reversible-jump algorithm in a re-parametrization version of the model.
  Development version available upon request

• DiriCens : Extension of DiriXtremes for inference with censored data.
  Development version available upon request