INRIA a CCSD electronic archive server
Not a member yet
122212 research outputs found
Sort by
Increasing the Lifetime of HPC Machines: Issues, Implications, and Open Challenges
No full textExtending the lifetime of High-Performance Computing (HPC) machines is becoming an important concern for a variety of reasons. These include the environmental and human costs associated with chip manufacturing, the rising demands by AI workloads, the soaring prices of accelerator chips, political blocks, and delays in the delivery of next-generation supercomputers. As a community, we must reconsider the traditional HPC paradigm and explore new strategies for making existing HPC infrastructure viable for longer periods. In this work, we highlight the current barriers in prolonging HPC machines lifespan and discuss key technical and operational challenges towards this goal
Fostering Responsibility in Email Marketing: A Contextual Restless Bandit Framework
No full textInternational audienceEmail marketing is increasingly criticized due to ethical concerns, as bulk email campaigns often result in spam, reduced engagement, and negative user experiences. In addition, there is increasing awareness of the environmental impact, as these large-scale campaigns contribute to carbon emissions. To address these issues, we introduce QWIC-Fair (Q-learning Whittle Index with Context and Fairness), an algorithm that operates within a Contextual Restless Multi-Armed Bandit framework. QWIC-Fair leverages implicit feedback to learn the dynamics of user interactions and thus target users with relevant content. In this model, each user represents an arm of the bandit, evolving as a Markov Decision Process that captures state transitions reflecting their interactions with email contents, while accounting for contextual information. The algorithm also incorporates a fairness constraint to ensure balanced selection and to avoid repetitive targeting of the same users. The experiments conducted, using synthetic and real-world data, show that QWIC-Fair outperforms existing email marketing approaches
Mathematical and numerical analysis of the mixed formulation of single phase flow in three-dimensional fractured porous media
No full textThree-dimensional fractured-porous media are typically modeled using a reduced approach known as the Discrete Fracture Matrix (DFM) model. In this model, the rock matrix is represented in three dimensions, while the fracture network is represented as a two-dimensional (co-dimension one) structure. The system is governed by a set of partial differential equations (PDEs) that couple the flow in the 3D rock matrix with the flow in the 2D fracture network. Mixed finite element methods are particularly well-suited for discretizing this system because they ensure local mass conservation. This paper presents an existence and uniqueness result for the PDEs in mixed form, assuming that the pressure is continuous across fractures (i.e., the fractures are conductive). Additionally, error estimates for the mixed finite element discretization are derived under a regularity assumption, and a mixed-hybrid approximation for the coupled problem is developed. The theoretical findings are validated through comparison with an analytical solution
Coypu: Music on-the-fly with Pharo
No full textInternational audienceCoypu is a live coding package and domain-specific language for Pharo, designed to be easy to install, simple to learn, and fun to use.It follows the principles of:• Iconicity – code that resembles its meaning • Economy – minimal syntax, maximal clarity • Synonymic Equivalence – flexibility in expressionThe goal: code that feels more like natural human language
FraPPE: Fast and Efficient Preference-based Pure Exploration
No full textInternational audiencePreference-based Pure Exploration (PrePEx) aims to identify with a given confidence level the set of Pareto optimal arms in a vector-valued (aka multi-objective) bandit, where the reward vectors are ordered via a (given) preference cone . Though PrePEx and its variants are well-studied, there does not exist a computationally efficient algorithm that can optimally track the existing lower bound for arbitrary preference cones. We successfully fill this gap by efficiently solving the minimisation and maximisation problems in the lower bound. First, we derive three structural properties of the lower bound that yield a computationally tractable reduction of the minimisation problem. Then, we deploy a Frank-Wolfe optimiser to accelerate the maximisation problem in the lower bound. Together, these techniques solve the maxmin optimisation problem in time for a bandit instance with arms and dimensional reward, which is a significant acceleration over the literature. We further prove that our proposed PrePEx algorithm, FraPPE, asymptotically achieves the optimal sample complexity. Finally, we perform numerical experiments across synthetic and real datasets demonstrating that FraPPE achieves the lowest sample complexities to identify the exact Pareto set among the existing algorithms
Modèles stochastiques individu-centrés avec des dynamiques allométriques : branchement, convergence, simulations numériques
No full textThe first part of this work focuses on the design and study of an individual-based model, structured in a trait called energy, describing a population of individuals consuming a resource assumed to be constant over time. In order to be able to compare the behaviour of various living species, we introduce the typical energy of an individual at birth as a parameter of the model. The mechanisms involved are of two types: random jumps corresponding to birth and death events; and a continuous and deterministic evolution of individual energies between jump times. The jump rates of the process depend on the energy of the individuals over time and, although our model is formulated in a general context, we especially focus on the case of allometric rates (i.e. they are assumed to be power functions). Individual trajectories are independent conditionally to the initial state of the population, because there is no competition for resources. We therefore study a branching process, and obtain necessary conditions on the allometric parameters of the model for this process to be supercritical, and for individual trajectories to be biologically relevant (individual energies do not explode, do not reach 0, and individuals die in finite time), for every living species we consider (i.e. for every typical energy of an individual at birth). In the second part, we modify the previous model by allowing the resource to vary over time, adding a renewal term independent of the state of the population, and a consumption term corresponding to indirect competition between individuals. After a preliminary study to construct well-defined objects, we consider a sequence of renormalizations of the underlying process and show a tightness result for the associated laws in a large population asymptotic. We characterize the accumulation points of this sequence as solutions of an integro-differential system of equations, which proves the existence of measure solutions to this system. Furthermore, if such a measure solution is unique, then our tightness result becomes a convergence result towards this unique process. The originality of our results lies in the fact that we study unbounded jump rates, and in particular the allometric case. In a final part, we show additional results for the integro-differential system of equations described in the second part. We present numerical simulations illustrating in particular the convergence (if the solution to the previous system is unique) result in a large population limit of the second part in the allometric case, and discuss the numerous difficulties encountered when we implement a numerical scheme within the framework of our model.La première partie de ce travail s'attache à la conception et à l'étude d'un modèle individu-centré, structuré en un trait appelé énergie, décrivant une population d'individus consommant une ressource supposée constante au cours du temps. Afin de pouvoir comparer les comportements de différentes espèces vivantes, nous introduisons notamment l'énergie typique d'un individu à la naissance comme paramètre du modèle. Les mécanismes en jeu sont de deux types : des sauts aléatoires correspondant à des événements de naissance et mort ; et une évolution continue et déterministe des énergies individuelles entre ces instants de saut. Les taux de saut du processus dépendent de l'énergie des individus au cours du temps et, bien que notre modèle soit formulé de manière générale, nous nous intéressons en particulier au cas de taux allométriques (i.e. au cas de fonctions de type puissance). Les trajectoires individuelles sont indépendantes conditionnellement à l'état initial de la population, du fait de l'absence de compétition pour la ressource. Nous étudions donc un processus de branchement, et obtenons des conditions nécessaires sur les paramètres allométriques du modèle pour que ce processus soit surcritique, et que les trajectoires individuelles soient biologiquement acceptables (i.e. les énergies individuelles n'explosent pas, ne s'annulent pas, et les individus meurent en temps fini), et ce quelle que soit l'espèce considérée (c'est-à-dire quelle que soit l'énergie typique d'un individu à la naissance). Dans une deuxième partie, nous modifions le modèle précédent en autorisant la ressource à varier au cours du temps, via un terme de renouvellement indépendant de l'état de la population, et un terme de consommation qui correspond à une compétition indirecte entre les individus. Après une étude préliminaire de la bonne définition des objets étudiés, nous considérons une suite de renormalisations du processus sous-jacent, et démontrons un résultat de tension pour les lois associées dans une asymptotique de grande population. Nous caractérisons les valeurs d'adhérence de cette suite comme des solutions d'un système d'équations intégro-différentiel, ce qui démontre au passage l'existence de solutions mesures à ce système. De plus, sous réserve que ce dernier admette une unique solution mesure, notre résultat de tension devient un résultat de convergence en loi vers cet unique processus. La nouveauté de nos résultats réside dans le caractère non-borné des taux de sauts, et nous étudions en particulier le cas allométrique. Dans une dernière partie, nous montrons des résultats complémentaires sur le système d'équations intégro-différentiel décrit dans la deuxième partie. Nous présentons des simulations numériques illustrant notamment le résultat de convergence (sous réserve d'unicité de la solution mesure au système étudié) en grande population de la deuxième partie dans le cas allométrique, et discutons des nombreuses difficultés de mise en œuvre d'un schéma numérique dans le cadre de notre modèle
From blades to tracks: a case study in structural reuse of curved surfaces for circular design
No full textInternational audienceWe explore the fabrication of curved surfaces by reusing panels extracted from decommissioned wind turbine blades, using cycling pumptracks as a case study. We first present real-world prototypes of pumptrack modules that we manufactured to evaluate the practicality of this reuse scenario and to define the boundary conditions for harvesting blade panels and assembling a track. We then propose an algorithm to optimize the segmentation of a wind turbine blade into quadrilateral panels whose sides fall within a small set of compatible boundaries. These panels form a library of modules that designers can connect side by side to create pumptracks of various lengths and curvatures. Together, these contributions provide a proof-of-concept of how computer-aided design and manufacturing can support circular design through the reuse of curved surfaces
Unsupervised Deep Generative Models for Anomaly Detection in Neuroimaging: A Systematic Scoping Review
No full textInternational audienceUnsupervised deep generative models are emerging as a promising alternative to supervised methods for detecting and segmenting anomalies in brain imaging. Unlike fully supervised approaches, which require large voxel-level annotated datasets and are limited to well-characterised pathologies, these models can be trained exclusively on healthy data and identify anomalies as deviations from learned normative brain structures. This PRISMA-ScR–guided scoping review synthesises recent work on unsupervised deep generative models for anomaly detection in neuroimaging, including autoencoders, variational autoencoders, generative adversarial networks, and denoising diffusion models. A total of 49 studies published between 2018 and 2025 were identified, covering applications to brain MRI and, less frequently, CT across diverse pathologies such as tumours, stroke, multiple sclerosis, and small vessel disease. Reported performance metrics (Dice, AUROC, AUPRC) are compared alongside architectural design choices such as dimensionality, masking, patching, and loss formulations. Across the included studies, generative models achieved encouraging performance for large focal lesions and demonstrated steady progress in addressing more subtle and heterogeneous abnormalities. While supervised methods remain the benchmark, unsupervised approaches are advancing rapidly, with increasing adoption of 3D architectures and anatomy-aware designs. A key strength of generative models is their ability to produce interpretable pseudo-healthy (also referred to as counterfactual) reconstructions, which is particularly valuable when annotated data are scarce, as in rare or heterogeneous diseases. Looking ahead, these models offer a compelling direction for anomaly detection, enabling semi-supervised learning, supporting the discovery of novel imaging biomarkers, and facilitating within- and cross-disease deviation mapping in unified end-to-end frameworks. To realise clinical impact, future work should prioritise anatomy-aware modelling, development of foundation models, task-appropriate evaluation metrics, and rigorous clinical validation
Towards Key Contributing Factors in Identifying Dark Pattern Autonomy Violations under the EU Digital Services Act
No full textInternational audienceDark patterns refer to design practices which undermine users' ability to make autonomous and informed choices in relation to digital systems. The recent EU Digital Services Act (DSA) aims to protect users from such dark patterns and their effects. DSA Article 25 prohibits three autonomy violation types: deception, manipulation and distortion/impairment. However, for regulation of dark patterns, it is important to reason about why an observed design practice constitutes a particular autonomy violation type, to show that it indeed violates the DSA. In this work-in-progress, two experts (with HCI, CS and legal background) mapped 59 known dark patterns onto these three autonomy violation types. We then analysed our rationale for this mapping to identify eight design factors which can help determine the dark pattern autonomy violation(s). Our analysis aims to situate existing dark patterns knowledge within the DSA legal framework, to support regulation and compliance of such design practices
Optimal sub-Gaussian variance proxy for 3-mass distributions
No full textWe investigate the problem of characterizing the optimal variance proxy for sub-Gaussian random variables, whose moment-generating function exhibits bounded growth at infinity. We apply a general characterization method to discrete random variables with equally spaced atoms. We thoroughly study 3-mass distributions, thereby generalizing the well-studied Bernoulli case. We also prove that the discrete uniform distribution over N points is strictly sub-Gaussian. Finally, we provide an open-source Python package that combines analytical and numerical approaches to compute optimal sub-Gaussian variance proxies across a wide range of distributions