Mohamed Hebiri

Curriculum Vitae (detailed version)

Research interests

• Machine learning: Low consommation, algorithmic fairness, set-valued classification, semi-supervised methods, multi-class and multi-label classification, classification with reject option

• High dimensional statistic: Lasso-type methods, optimality of tuning parameters, Transductive approaches

• Online learning: Conformal Prediction, on-line Prediction, active learning

Publications

Journal papers

• Minimax semi-supervised set-valued approach for multi-class classification, with Evgenii Chzhen and Christophe Denis (2021). To appear in Bernoulli.

• Consistency of plug-in confidence sets for classification in semi-supervised learning, with Christophe Denis (2020), Journal of Nonparametric Statistics, vol. 32(1), 42-72.
  This paper is a long version of the Manuscript: Confidence Sets for Classification, with Christophe Denis (2015), Statistical Learning and Data Sciences (SLDS 2015), 301-312.

• On Lasso refitting strategies, with Evgenii Chzhen and Joseph Salmon (2019), Bernoulli, vol. 25(4A), 3175-3200.

• Confidence sets with expected sizes for Multiclass Classification, with Christophe Denis (2017), Journal of Machine Learning Research, vol. 18(102), 1-28.

• On the Prediction Performance of the Lasso, with Arnak Dalalyan and Johannes Lederer (2017), Bernoulli, vol. 23(1), 552-581.

• Rank penalized estimation of a quantum system, with Pierre Alquier, Cristina Butucea, and Katia Meziani. (2013) Physical Review A, vol. 88(3), 032113.

• How Correlations Influence Lasso Prediction, with Johannes Lederer. (2013), IEEE, Transactions on Information Theory, Volume 59(3), 1846-1854.

• Transductive versions of the LASSO and the Dantzig Selector, with Pierre Alquier (2012), Journal of Statistical Planning and Inference, Vo. 142(9), 2485-2500.

• Generalization of l₁ constraints for high dimensional regression problems, with Pierre Alquier (2011), Statistics and Probability Letters, Vol. 81, 1760-1765.

• The Smooth-Lasso and other l₁+l₂-penalized methods, with Sara van de Geer (2011), Electronic Journal of Statistics, Vol. 5, 1184-1226.
  This paper is a reviewed version of the Manuscript: Regularization with the Smooth-Lasso procedure.

• Sparse Conformal Predictors (2010), Statistics and Computing: Vol. 20(2), 253-266.

• Some theoretical results on the grouped variables Lasso, with Christophe Chesneau (2008), Mathematical Methods of Statistics, Vol. 17(4), 317-326.

Conference papers

• Regression with reject option and application to kNN, with Christophe Denis and Ahmed Zaoui (2020), Proceedings of Thirty-forth Conference on Neural Information Processing Systems (NeurIPS).

• Fair Regression with Wasserstein Barycenters, with Evgenii Chzhen, Christophe Denis, Luca Oneto, and Massimiliano Pontil (2020), Proceedings of Thirty-forth Conference on Neural Information Processing Systems (NeurIPS).

• Fair Regression via Plug-in Estimator and Recalibration With Statistical Guarantees, with Evgenii Chzhen, Christophe Denis, Luca Oneto, and Massimiliano Pontil (2020), Proceedings of Thirty-forth Conference on Neural Information Processing Systems (NeurIPS).

• Leveraging labeled and unlabeled data for consistent fair binary classification, with Evgenii Chzhen, Christophe Denis, Luca Oneto, and Massimiliano Pontil (2019), Proceedings of Thirty-third Conference on Neural Information Processing Systems (NeurIPS).

• Learning Heteroscedastic Models by Convex Programming under Group Sparsity, with Arnak Dalalyan, Katia Meziani and Joseph Salmon (2013), Proceedings of the International Conference on Machine Learning (ICML).

Preprints and documents

• Active learning algorithm through the lens of rejection arguments, with Christophe Denis, Boris Ndjia Njike, and Xavier Siebert (2022). Submitted.

• Prediction intervals with controlled length in the heteroscedastic Gaussian regression, with Christophe Denis and Ahmed Zaoui (2022). Submitted.

• Set-valued classification - overview via a unified framework, with Evgenii Chzhen, Christophe Denis, and Titouan Lorieul (2022). Submitted.

• Fairness guarantee in multi-class classification, with Christophe Denis, Romuald Elie, and François Hu (2022). Submitted.

• Layer Sparsity in Neural Networks, with Johannes Lederer (2020). Submitted.

• On the benefits of output sparsity for multi-label classification, with Evgenii Chzhen, Christophe Denis, and Joseph Salmon (2017). Submitted.

• Aggregation for Linear Inverse Problems, with Jean-Michel Loubes and Paul Rochet (2014). Preprint.

Thesis

Ph.D. in Mathematical Statistics: Quelques questions de sélection de variables autour de l'estimateur LASSO (2009), Université Paris Diderot-Paris 7.
Advisor: Nicolas Vayatis.

Habilitation à Diriger des Recherches (HdR)

Some contributions to high-dimensional statistics and learning with desired properties(2021) -- Université Gustave.

Teaching experience

Seminars:

• Since 2021. Organization of a seminar on statistic in LAMA laboratory with Chi Tran.

• 2019. Organization of a lecture group in the statistical team of LAMA laboratory (Université Gustave Eiffel) on the book Introductory lectures on convex optimization by Yurii Nesterov.

• 2013--2018. Organization of a weakly seminar probability, statistics, and their applications (ASPRo) (ASPRo) in LAMA laboratory with Matthieu Fradelizi and Miguel Martinez.

Conferences organizations:

• Organization of a session entitled Artificial Intelligence, Journées MAS - Rouen, 29 until 31 August 2022.

• Organization of a session entitled Algorithmic fairness with statistical guarantees, CMStatistics Conference (ERCIM 2020) - Virtual Conference, 19 until 21 December 2020.

• Organization of a session entitled Set-valued classification in CMStatistics Conference (ERCIM 2019) - University of London, UK, 14 until 16 December 2019.

Committee member:

• Sixth Symposium on Conformal and Probabilistic Prediction with Applications (COPA 2017). Karolinska Institutet, Stockholm, Sweden (14th until 16th June 2017).

• Special Session on Conformal Prediction and its Applications (COPA 2015 Workshop). Royal Holloway, University of London, Egham, Surrey, UK (20th until 21th April 2015).

• 3rd Conformal Prediction and its Applications Workshop (COPA 2014). Island of Rhodes, Greece (19th until 21st September 2014).

• 2nd Conformal Prediction and its Applications Workshop (COPA 2013). Paphos, Cyprus (30th September until 2nd October 2013).

• 1st Conformal Prediction and its Applications Workshop (COPA 2012). Halkidiki, Greece (27th until 30th September 2012).

• Statistics and Modeling for Complex Data. École des Ponts Paris-Tech (22 until 24 June 2011).

PhD Students

• Florian Valade - since 2022 : Generalization of energy-saving algorithms for recognition on board camera. Co-advised with Sébastien Loustau -- CIFRE with Fujitsu.

• Gayane Taturyan - depuis 2022 : Fairness and Robustness in Machine Learning. Co-dirigée avec Evgenii Chzhen et Jean-Michel Loubes.

• Ahmed Zaoui - 2018-2022 : Estimation of the variance function for regression and its applications. Co-advised with Christophe Denis.

• Evgenii Chzhen - 2016-2019 : Plug-in Approaches for classification. Co-advised with Joseph Salmon.

Institution

• Since 2018: member of Conseil de l’UFR de Mathématiques in Université Paris-Est -- Marne-la-Vallée.

• Since 2018: member of Commission Responsable du Budget de l’UFR de Mathématiques in Université Paris-Est -- Marne-la-Vallée.

• Since 2017: member of Comité Scientifique du LAMA.

• Since 2016: member of la Commission Permanente de Recrutement de la section 61 in Université Paris-Est -- Marne-la-Vallée.

• From 2012 until 2016: member of Commission permanente de recrutement de l'UFR de Mathématiques in Université Paris-Est -- Marne-la-Vallée.

• 2013 and 2015: member of the selection committee for a Probability-Statistic ``Maîtres de Conférences'' position in Université Paris-Nanterre.