2.1 Mathematical statistics and learning

Data science—a vast field that includes statistics, machine learning, signal processing, data visualization, and databases—has become front-page news due to its ever-increasing impact on society, over and above the important role it already played in science over the last few decades. Within data science, the statistical community has long-term experience in how to infer knowledge from data, based on solid mathematical foundations. The recent field of machine learning has also made important progress by combining statistics and optimization, with a fresh point of view that originates in applications where prediction is more important than building models.

The Celeste project-team is positioned at the interface between statistics and machine learning. We are statisticians in a mathematics department, with strong mathematical backgrounds, interested in interactions between theory, algorithms, and applications. Indeed, applications are the source of many of our interesting theoretical problems, while the theory we develop plays a key role in (i) understanding how and why successful statistical learning algorithms work—hence improving them—and (ii) building new algorithms upon mathematical statistics-based foundations. Therefore, we tackle several major challenges of machine learning with our mathematical statistics point of view (in particular the algorithmic fairness issue), always having in mind that modern datasets are often high-dimensional and/or large-scale, which must be taken into account at the building stage of statistical learning algorithms. For instance, there often are trade-offs between statistical accuracy and complexity which we want to clarify as much as possible.

In addition, most theoretical guarantees that we prove are non-asymptotic, which is important because the number of features $p$ is often larger than the sample size $n$ in modern datasets, hence asymptotic results with $p$ fixed and $n\to +\infty$ are not relevant. The non-asymptotic approach is also closer to the real-world than specific asymptotic settings, since it is difficult to say whether $p=1000$ and $n=100$ corresponds to the setting $p=10n$ or $p=n^{3/2}$.

Finally, a key ingredient in our research program is connecting our theoretical and methodological results with (a great number of) real-world applications. This is the reason why a large part of our work is devoted to industrial and medical data modeling on a set of real-world problems coming from our long-term collaborations with several partners, as well as various opportunistic one-shot collaborations.

3.1 General presentation

We split our research program into four research axes, distinguishing problems and methods that are traditionally considered part of mathematical statistics (e.g., model selection and hypothesis testing, see section 3.2) from those usually tackled by the machine learning community (e.g., multi-armed bandits and reinforcement learning, deep learning, clustering and pairwise-data inference, see section 3.3). Section 3.4 is devoted to industrial and medical data modeling questions which arise from several long-term collaborations and more recent research contracts. Finally, section 3.5 is devoted to algorithmic fairness, a theme of Celeste which we want to specifically emphasize. Despite presenting mathematical statistics, machine learning, and data modeling as separate axes, we would like to make clear that these axes are strongly interdependent in our research and that this dependence is a key factor in our success.

3.2 Mathematical statistics

One of our main goals is to address major challenges in machine learning in which mathematical statistics naturally play a key role, in particular in the following two areas of research.

3.3 Theoretical foundations of machine learning

Our distinguishing approach (compared to peer groups around the world) is to offer a statistical and mathematical point of view on machine-learning (ML) problems and algorithms. Our main focus is to provide theoretical guarantees for certain ML problems, with special attention paid to the statistical point of view, in particular minimax optimality and statistical adaptivity. In the areas of deep learning and big data, computationally-efficient optimization algorithms are essential. The choice of the optimization algorithm has been shown to have a dramatic impact on generalization properties of predictors. Such empirical observations have led us to investigate the interplay between computational efficiency and statistical properties. The set of problems we tackle includes online learning (expert aggregation, stochastic bandits, reinforcement learning), clustering and co-clustering, pairwise-data inference, semi-supervised learning, and the interplay between optimization and statistical properties.

3.4 Industrial and medical data modeling

Celeste collaborates with industry and with medicine/public health institutes to develop methods and apply results of a broadly statistical nature—whether they be prediction, aggregation, anomaly detection, forecasting, and so on—in relationship with pressing industrial and/or societal needs (see sections 4 and 5.2). Most of these methods and applied results are directly related to the more theoretical subjects examined in the first two research axes, including for instance estimator selection, aggregation, and supervised and unsupervised classification. Furthermore, Celeste is positioned well for problems with data requiring unconventional methods—for instance, non asymptotic analysis and data with selection bias—, and in particular problems that can give rise to technology transfers in the context of Cifre Ph.D.s.

3.5 Algorithmic fairness

Machine-learning algorithms make pivotal decisions which influence our lives on a daily basis, using data about individuals. Recent studies show that imprudent use of these algorithms may lead to unfair and discriminatory decisions, often inheriting or even amplifying disparities present in data. The goal of Celeste on this topic is to design and analyze novel tractable algorithms that, while still optimizing prediction performance, mitigate or remove unfair decisions of the learned predictor. A major challenge in the machine-learning fairness literature is to obtain algorithms which satisfy fairness and risk guarantees simultaneously. Several empirical studies suggest that there is a trade-off between the fairness and accuracy of a learned model: more accurate models are less fair. We are focused on providing user-friendly statistical quantification of such trade-offs and building statistically-optimal algorithms in this context, with special attention paid to the online learning setting. Relying on the strong mathematical and statistical competency of the team, we approach the problem from an angle that differs from the mainstream computer science literature.

4.1 Electricity load consumption: forecasting and control

Celeste has a long-term collaboration with EDF R&D on electricity consumption. An important problem is to forecast consumption, e.g., for electric vehicles (EVs). We currently work on hierarchical consumption data of EVs, for which we aim to output probabilistic forecasts, e.g., through conformal inference methods.

4.2 Reliability

Data collected on the lifetime of complex systems is often non-homogeneous, affected by variability in component production and differences in real-world system use. In general, this variability is neither controlled nor observed in any way, but must be taken into account in reliability analysis. We use latent structure models to identify the main causes of failure, and to predict system reliability as accurately as possible 10.

4.3 Neglected tropical diseases

Celeste collaborates with researchers at Institut Pasteur on encephalitis in South-East Asia, especially with Jean-David Pommier.

4.4 Explainability in change-points detection in high dimensional multivariate time series

Detecting changes in time series is essential in many areas, such as identifying anomalies in industrial processes, monitoring medical conditions, detecting variations in climatic conditions, or analyzing fluctuations in financial markets.

Numerous change-point detection approaches have been developed, both offline and online, and applied to univariate and multivariate series. In the multivariate context, where the components of the series can represent the measurements of thousands of sensors, an important question remains after the change-point has been estimated: which sensors are specifically involved in the detected change?

In Dhia El Haq Ouerfelli's PhD thesis, we develop post-hoc methods to identify the coordinates involved in a detected change and to evaluate the quality of this detection.

4.5 Cytometry

Celeste collaborates with Metafora to explore the use of multiple instance learning in flow cytometry as a means of early detection of specific cancers. This is in collaboration with Pascal Massart and Christine Keribin, in the context of Pierre-André Mikem's Cifre PhD, which follows Louis Pujol's thesis defended in 2022.

4.6 Railway operations

Following the CIFRE PhD of Rémi Coulaud, we continue our ongoing collaboration with SNCF–Transilien to exploit large datasets of railway operation and passenger flows, obtained by automatic recording devices (for passenger flows, these correspond to sensors at the door level). We model and forecast passenger movement inside train coaches, so as to be able to provide incoming passengers with information on how crowded wagons are. We link this problem with a neural network framework to improve performance. The next step is to take into account the behavior of passengers on platform. This will be examined in a new CIFRE PhD contract to start in 2025.

4.7 Ancient materials

Celeste collaborates with CNRS-IPANEMA (Ancient Materials Research Platform). The goal is to propose a new image segmentation method based on a dissimilarity which is particularly well adapted to XRF images. This will allow less exposure to radiation, which is important when dealing with antiques.

4.8 Education sciences

Ensuring student success is a central goal of universities, and a high success rate is seen as an indicator of the academic excellence of the institution. The transition from high school to university is seen as a critical time for first-time students, who must adapt not only to a different academic environment but also to greater autonomy. Universities face the challenge of facilitating this transition for a student population that is heterogeneous in terms of academic preparation, cultural and socio-economic background.

We currently collaborate with the EST laboratory (Univ. Paris-Saclay) on this topic. Our works are aimed at identifying the different factors that hinder success, identifying success profiles, and proposing welcoming and support solutions to effectively encourage the success of each student 14, 25.

5.1 Footprint of research activities

The carbon emissions of Celeste team members related to their jobs were very low and came essentially from:

• limited levels of transport to and from work, and a small amount for essentially land travel to conferences in France and Europe.
• electronic communication (email, Google searches, Zoom meetings, online seminars, etc.).
• the carbon emissions embedded in their personal computing devices (construction), either laptops or desktops.
• electricity for personal computing devices and for the workplace, plus also water, heating, and maintenance for the latter. Note that only 7.1% (2018) of France's electricity is not sourced from nuclear energy or renewables so team member carbon emissions related to electricity are minimal.

In terms of magnitude, the largest per capita ongoing emissions (excluding flying) are likely simply to be those from buying computers that have a carbon footprint from their construction, in the range of 100 kg Co2-e each. In contrast, typical email use per year is around 10 kg Co2-e per person, and a Zoom call comes to around 10g Co2-e per hour per person, while web browsing uses around 100g Co2-e per hour. Consequently, 2024 was a low carbon year for the Celeste team.

The approximate (rounded for simplicity) kg Co2-e values cited above come from the book, “How Bad are Bananas” by Mike Berners-Lee (2020) which estimates carbon emissions in everyday life.

5.2 Impact of research results

In addition to the long-term impact of our theoretical work—which is of course impossible to assess immediately—we are involved in several applied research projects which aim at having a short/mid-term positive impact on society.

First, we collaborate with the EST laboratory (Univ. Paris-Saclay) on questions related to student success in universities and how to maximize it (see Section 4.8).

Second, we collaborate with the Pasteur Institute on neglected tropical diseases; encephalitis in particular, with implications in global health strategies.

Third, the broad use of artificial intelligence/machine learning/statistics nowadays comes with several major ethical issues, one being to avoid making unfair or discriminatory decisions. Our theoretical work on algorithmic fairness has already led to several “fair” algorithms that could be widely used in the short term (one of them is already used for enforcing fair decision-making in student admissions at the University of Genoa).

Fourth, we expect short-term positive impact on society from several direct collaborations with companies such as EDF (forecasting and control of electricity load consumption, in particular, for electric vehicles) and Metafora (early detection of cancers).

6.1 Awards

• C. Keribin was promoted to full professor

7.1 New software

8.1 A tribute to Le Cam: On the van Trees inequality

Participants: Gilles Stoltz.

In 11, with Elisabeth Gassiat (Orsay) and in honor of the 100th birth anniversary of Lucien Le Cam (November 18, 1924—April 24, 2000), we developed a version of the van Trees (1968) inequality in the spirit of Hajek–Le Cam, i.e., under minimal assumptions that, in particular, involve no direct point-wise regularity assumptions on densities but rather almost-everywhere differentiability in the model's quadratic mean. Surprisingly, it suffices that the latter differentiability holds along canonical directions—not along all directions. Also, we identified a (slightly stronger) version of the van Trees inequality as an actual instance of a Cramér-Rao bound, i.e., the van Trees inequality is not just a Bayesian analog of the Cramér–Rao bound, as believed so far. We provided, as an illustration, an elementary proof of the local asymptotic minimax theorem for quadratic loss functions, again assuming differentiability in quadratic mean only along canonical directions.

8.2 Marginal and training-conditional guarantees in one-shot federated conformal prediction

Participants: Pierre Humbert, Sylvain Arlot.

In collaboration with Batiste Le Bars and Aurélien Bellet (Inria Lille, Magnet project-team), we study in 32 conformal prediction in the one-shot federated learning setting. The main goal is to compute marginally and training-conditionally valid prediction sets, at the server-level, in only one round of communication between the agents and the server. Using the quantile-of-quantiles family of estimators and split conformal prediction, we introduce a collection of computationally-efficient and distribution-free algorithms that satisfy the aforementioned requirements. Our approaches come from theoretical results related to order statistics and the analysis of the Beta-Beta distribution. We also prove upper bounds on the coverage of all proposed algorithms when the nonconformity scores are almost surely distinct. For algorithms with training-conditional guarantees, these bounds are of the same order of magnitude as those of the centralized case. Remarkably, this implies that the one-shot federated learning setting entails no significant loss compared to the centralized case. Our experiments confirm that our algorithms return prediction sets with coverage and length similar to those obtained in a centralized setting.

8.3 First-order ANIL provably learns representations despite over-parametrization

Participants: Etienne Boursier.

In collaboration with Oguz Yüksel and Nicolas Flammarion (EPFL), we give in 24 convergence guarantees for the well known meta-learning algorithm ANIL in the overparametrized regime. More precisely, we show in the limit of an infinite number of tasks, that first-order ANIL with an overparametrized linear two-layer network architecture successfully learns linear shared representations. The learned solution then yields a good adaptive performance on any new task after a single gradient step. Overall, this illustrates how well model-agnostic methods such as first-order ANIL can learn shared representations.

8.4 Estimating the history of a random recursive tree

Participants: Simon Briend, Christophe Giraud.

In collaboration with Gabor Lugosi and Déborah Sulen, we study in 9 the problem of estimating the order of arrival of the vertices in a random recursive tree. Specifically, we study two fundamental models: the uniform attachment model and the linear preferential attachment model. We propose an order estimator based on the Jordan centrality measure and define a family of risk measures to quantify the quality of the ordering procedure. Moreover, we establish a minimax lower bound for this problem, and prove that the proposed estimator is nearly optimal. Finally, we numerically demonstrate that the proposed estimator outperforms degree-based and spectral ordering procedures.

8.5 Minimax optimal seriation in polynomial time

Participants: Christophe Giraud.

In collaboration with Yann Issartel and Nicolas Verzelen, we investigate in 33 the statistical seriation problem, where the statistician seeks to recover a hidden ordering from a noisy observation of a permuted Robinson matrix. In this paper, we tightly characterize the minimax rate for this problem of matrix reordering when the Robinson matrix is bi-Lipschitz, and we also provide a polynomial time algorithm achieving this rate; thereby answering two open questions of [Giraud et al., 2021]. Our analysis further extends to broader classes of similarity matrices.

8.6 Computation-information gap in high-dimensional clustering

Participants: Bertrand Even, Christophe Giraud.

In collaboration with Nicolas Verzelen, we investigate in 21 the existence of a fundamental computation-information gap for the problem of clustering a mixture of isotropic Gaussian in the high-dimensional regime, where the ambient dimension $d$ is larger than the number $n$ of points. We provide evidence of the existence of such a gap generically in the high-dimensional regime $d\ge n$, by (i) proving a non-asymptotic low-degree polynomials computational barrier for clustering in high-dimension, matching the performance of the best known polynomial time algorithms, and by (ii) establishing that the information barrier for clustering is smaller than the computational barrier, when the number $K$ of clusters is large enough. These results are in contrast with the (moderately) low-dimensional regime $n\ge poly(d,K)$, where there is no computation-information gap for clustering a mixture of isotropic Gaussian. In order to prove our low-degree computational barrier, we develop sophisticated combinatorial arguments to upper-bound the mixed moments of the signal under a Bernoulli Bayesian model.

8.7 Active clustering with bandit feedback

Participants: Christophe Giraud.

In collaboration with Victor Thuot, Alexandra Carpentier, and Nicolas Verzelen, we investigate in 38 the Active Clustering Problem (ACP). A learner interacts with an $N$-armed stochastic bandit with $d$-dimensional subGaussian feedback. There exists a hidden partition of the arms into $K$ groups, such that arms within the same group, share the same mean vector. The learner's task is to uncover this hidden partition with the smallest budget - i.e., the least number of observation - and with a probability of error smaller than a prescribed constant $\delta$. In this paper, (i) we derive a non-asymptotic lower bound for the budget, and (ii) we introduce the computationally efficient ACB algorithm, whose budget matches the lower bound in most regimes. We improve on the performance of a uniform sampling strategy. Importantly, contrary to the batch setting, we establish that there is no computation-information gap in the active setting.

8.8 Policy optimization for adversarial Markov decision processes

Participants: Daniil Tiapkin, Evgenii Chzhen, Gilles Stoltz.

In 39, we consider policy optimization in adversarial $H$-episodic Markov decision processes (MDPs) and in the full information setting, Briefly, the learning agent interacts with an environment during $T$ episodes, each of which consists of $H$ stages, and each episode is evaluated with respect to a reward function that will be revealed only at the end of the episode; transition kernels are constant over episodes but unknown. The best regret known so far was of order $S\sqrt{AT}$, as far as the orders of magnitude of $T$ and the sizes $S$ and $A$ of the state and action spaces are concerned. We propose an algorithm, called APO-MVP, that achieves a regret bound of order $\sqrt{SAT}$, matching the minimax lower bound. The proposed algorithm leverages and combines two recent techniques: policy optimization based on online linear optimization strategies (Jonckheere et al., 2023) and a refined martingale analysis of the impact on values of estimating transitions kernels (Zhang et al., 2023).

8.9 Incentivized learning in principal-agent bandit games

Participants: Etienne Boursier.

In collaboration with A. Scheid (Polytechnique), D. Tiapkiin (LMO), A. Capitaine (Polytechnique), E-M El Mhamdi (Polytechnique), E. Moulines (Polytechnique), M. Jordan (Inria Paris, UC Berkeley), and A. Durmus (Polytechnique), we study in 23 a repeated principal-agent bandit game, where the principal can only interact with her environment through the agent. The principal and the agent have misaligned objectives and the choice of action is only left to the agent. However, the principal can influence the agent’s decisions by offering incentives which add up to her rewards. The principal aims to iteratively learn an incentive policy to maximize her own total utility. This framework extends usual bandit problems and is motivated by several practical applications, such as healthcare or ecological taxation, where traditionally used mechanism design theories often overlook the learning aspect of the problem. We present nearly optimal learning algorithms for the principal’s regret in both the multi-armed and linear contextual settings.

8.10 Learning to mitigate externalities: the Coase Theorem with hindsight rationality

Participants: Etienne Boursier.

In collaboration with A. Scheid (Polytechnique), A. Capitaine (Polytechnique), E. Moulines (Polytechnique), M. Jordan (Inria Paris, UC Berkeley), and A. Durmus (Polytechnique), we extend in 22 the result from Section 8.9 to the case where both the agent and principal are learning. In particular, we design a principal strategy that has low regret guarantees against any agent following a learning strategy inside a class of suitable no-regret strategies.

8.11 A survey on multi-player bandits

Participants: Etienne Boursier.

In collaboration with V. Perchet (ENSAE), we give in 8 a broad overview of recent developments that have appeared over the past decade to do with multiplayer bandits. Multiplayer bandits have recently been extensively studied because of their application to cognitive radio networks. Although considerable progress has been made on theoretical aspects, current algorithms are far from applicable and many obstacles remain between these theoretical results and a possible implementation of multiplayer bandits algorithms in real cognitive radio networks. This survey contextualizes and organizes the rich multiplayer bandits literature.

8.12 Unravelling in collaborative learning

Participants: Etienne Boursier.

In collaboration with A. Capitaine (Polytechnique), A. Scheid (Polytechnique), E. Moulines (Polytechnique), M. Jordan (Inria Paris, UC Berkeley), E-M El Mhamdi (Polytechnique), and A. Durmus (Polytechnique), we study in 19 the problem of collaborative learning with strategic agents. When such agents wish to train a model together but have sampling distributions of different quality, the coalition may undergo a phenomenon known as unravelling, wherein it shrinks up to the point that it becomes empty or solely comprised of the worst agent. We address this issue by proposing a novel method inspired by probabilistic verification. This approach makes the grand coalition a Nash equilibrium with high probability, despite information asymmetry, thereby hindering unravelling.

8.13 Statistical Learning Tools for Electricity Load Forecasting

Participants: Jean-Michel Poggi.

The monograph 26, written with Anestis Antoniadis (Institute of Applied Sciences and Intelligent Systems `Eduardo Caianiello', Naples), Jairo Cugliari (Lab. ERIC EA 3083, Lumière University Lyon 2), Matteo Fasiolo (School of Mathematics, University of Bristol) and Yannig Goude (EDF R& D & LMO), explores a set of statistical and machine learning tools that can be effectively utilized for applied data analysis in the context of electricity load forecasting. Drawing on their substantial research and experience with forecasting electricity demand in industrial settings, the authors guide readers through several modern forecasting methods and tools from both industrial and applied perspectives – generalized additive models (GAMs), probabilistic GAMs, functional time series and wavelets, random forests, aggregation of experts, and mixed effects models. A collection of case studies based on sizable high-resolution datasets, together with relevant R packages, then illustrate the implementation of these techniques. Five real datasets at three different levels of aggregation (nation-wide, region-wide, or individual) from four different countries (UK, France, Ireland, and the USA) are utilized to study five problems: short-term point-wise forecasting, selection of relevant variables for prediction, construction of prediction bands, peak demand prediction, and use of individual consumer data.

This text is intended for practitioners, researchers, and post-graduate students working on electricity load forecasting; it may also be of interest to applied academics or scientists wanting to learn about cutting-edge forecasting tools for application in other areas. Readers are assumed to be familiar with standard statistical concepts such as random variables, probability density functions, and expected values, and to possess some minimal modeling experience.

8.14 Diversity-preserving stochastic bandits

Participants: Gilles Stoltz.

In 13, with Hédi Hadiji (CentraleSupélec), we revisit the bandit-based framework for diversity-preserving recommendations introduced by Celis, Kapoor, Salehi, and Vishnoi (“Controlling polarization in personalization: an algorithmic framework”, at FAccT’2019).

Typically, in stochastic bandits, learning strategies focus quickly on the best-performing arm(s), and play only it (them), except for at most logarithmically many rounds. This phenomenom is called polarization. Celis et al. (2019) proposed a model to avoid polarization: by imposing that the player should play in two stages, first by picking a probability distribution over the arms located in a strict subset of the simplex (the set of all probability distributions)—typically, a polytope in the inner of the simplex. This modeling is interesting and we took it as a starting point.

Our contribution is to get sharper theoretical results (lower bounds and refined upper bounds). Celis et al. (2019) approached the problem in the case of a polytope mainly by a reduction to the setting of linear bandits. We instead design a direct UCB strategy using the specific structure of the setting and show that it enjoys a bounded distribution-dependent regret in the natural cases when the optimal mixed actions put some probability mass on all actions (i.e., when diversity is desirable). The regret lower bounds provided show that otherwise (at least when the model is mean-unbounded), a $lnT$ regret is suffered. We also discuss an example beyond the special case of polytopes.

8.15 Regression under demographic parity constraints via unlabeled post-processing

Participants: Evgenii Chzhen, Gayane Taturan.

In collaboration with Gayane Taturyan and Mohamed Hebiri, we study in 20 the problem of performing regression while ensuring demographic parity, even without access to sensitive attributes during inference. We present a general-purpose post-processing algorithm that, using accurate estimates of the regression function and a sensitive attribute predictor, generates predictions that meet the demographic parity constraint. Our method involves discretization and stochastic minimization of a smooth convex function. It is suitable for online post-processing and multi-class classification tasks only involving unlabeled data for the post-processing. Unlike prior methods, our approach is fully theory-driven. We require precise control over the gradient norm of the convex function, and thus, we rely on more advanced techniques than standard stochastic gradient descent. Our algorithm is backed by finite-sample analysis and post-processing bounds, with experimental results validating our theoretical findings.

9.1 Bilateral contracts with industry

• A. Janon: Contract with INSERM Toulouse (3,3 kE), on variable selection for identification of link between microbiotal dysbiosis and type-2 diabetes.
• C. Keribin: Ongoing Cifre PhD contract with Metafora (30 kE) on machine learning in flow cytometry for early detection of cancers.
• J.M. Poggi: Analysis and modelling of NO2 numerical model biases for data fusion of heterogeneous measurement networks, ATMO NORMANDIE, 20 kE.
• J.M. Poggi, G. Stoltz: Participation in the EDF–Inria Grand défi, with in particular a CIFRE PhD started in December 2023 and a post-doc to start in February 2025.
• G. Stoltz: Ongoing contract with BNP Paribas (3 x 10 kE), on stochastic bandits under budget constraints, with applications to loan management; annually since 2021.

10.1 National initiatives

Participants: Sylvain Arlot, Kevin Bleakley, Evgenii Chzhen, Christophe Giraud, Gilles Stoltz.

11.1 Promoting scientific activities

11.2 Teaching - Supervision - Juries

11.3 Popularization

12.1 Major publications

• 1 bookA.Anestis Antoniadis, J.Jairo Cugliari, M.Matteo Fasiolo, Y.Yannig Goude and J.-M.Jean-Michel Poggi. Statistical Learning Tools for Electricity Load Forecasting.Statistics for Industry, Technology, and EngineeringSpringer International Publishing2024HALDOI
• 2 articleE.Etienne Boursier and V.Vianney Perchet. A survey on multi-player bandits.Journal of Machine Learning ResearchJanuary 2024HAL
• 3 inproceedingsE.Evgenii Chzhen, M.Mohamed Hebiri and G.Gayane Taturyan. Regression under demographic parity constraints via unlabeled post-processing.PMLRNeurIPS 2024Vancouver, CanadaDecember 2024HAL
• 4 inproceedingsB.Bertrand Even, C.Christophe Giraud and N.Nicolas Verzelen. Computation-information gap in high-dimensional clustering.Proceedings of Thirty Seventh Conference on Learning Theory247Edmonton (Canada), CanadaJune 2024, 1646--1712HAL
• 5 articleÉ.Élisabeth Gassiat and G.Gilles Stoltz. The van Trees inequality in the spirit of Hajek and Le Cam.Statistical Science2024HAL
• 6 articleC.Christophe Giraud, Y.Yann Issartel, L.Luc Lehericy and M.Matthieu Lerasle. Pair-Matching: Link Prediction with Adaptive Queries.Mathematical Statistics and LearningMarch 2024HAL
• 7 inproceedingsO.Oguz Yuksel, E.Etienne Boursier and N.Nicolas Flammarion. First-order ANIL provably learns representations despite overparametrization.ICLR 2024 - The Twelfth International Conference on Learning RepresentationsVienne, AustriaMay 2024HAL

12.2 Publications of the year

12.3 Cited publications

• 41 inproceedingsP.Pierre Humbert, B.Batiste Le Bars, A.Aurélien Bellet and S.Sylvain Arlot. One-shot federated conformal prediction.International Conference on Machine LearningPMLR2023, 14153--14177back to text