HAL-CentraleSupelec
Not a member yet
77624 research outputs found
Sort by
Ten common misconceptions about Galaxy (and why they are wrong!)
No full textInternational audienceGalaxy is a widely used open-source platform for accessible, reproducible, transparent and scalable data analysis in the life sciences and beyond. Despite its growing adoption across domains, several misconceptions persist about its scope, usability, scalability and relevance to academia and industry. In this manuscript, we identify and address 10 common misconceptions about Galaxy, ranging from the belief that it is limited to genomics, lacks scalability, or is only useful for teaching, to doubts about its ability to support secure data analysis or maintain high software quality as a free and open-source project. We refute each misconception with present evidence based on Galaxy’s technical features, real-world use cases, user communities and governance structures. We show that Galaxy is a mature and versatile platform capable of supporting cutting-edge scientific research, education and even clinical workflows across a wide variety of disciplines. By clarifying existing misconceptions, we aim to help researchers, educators, developers and decision-makers better appreciate Galaxy’s capabilities and potential within their fields
Exploring Public Budgets with Symber Tree: An Interactive Hierarchical Visualization for Comparisons Across Orders of Magnitude
No full textInternational audienceWe present Symber Tree, a novel hierarchical visualization designed to support exploration and comparisons across values spanning multiple orders of magnitude. Common hierarchical visualizations, such as treemaps and icicle plots, struggle to represent such disparities across hierarchical levels and rely on interaction techniques that increase cognitive load by requiring users to recall previously seen values. To address these limitations, we adopt a hybrid node-link approach that encodes values along both edges and nodes. A Sankey-like representation conveys linear differences between categories, while glyphs inspired by symbolic number representation support cross-scale comparison and provide reference points on demand to facilitate evaluative meaning. We demonstrate the approach using the French national budget and report findings from a preliminary heuristic evaluation with eight visualization experts
Impact of N2∗ quenching process modeling on large-eddy simulation of plasma-assisted combustion
No full textInternational audienceNanosecond Repetitively Pulsed (NRP) discharges are an energy-efficient and promising technique to stabilize lean premixed flames. High-performance computing is an attractive and complementary approach to experiments to understand the effects of NRP discharges on the flame stability limit. The high computational cost of the complex plasma kinetics has led to the development of the phenomenological model of Castela et al., which aims to capture the main thermal and chemical effects induced by the discharges. Although it has been successfully applied and validated in many cases of flame ignition and stabilization by NRP discharges, limitations exist, such as the possible significance of the dissociation of species other than O2. Recently, Blanchard et al. proposed a new phenomenological model based on physics-based analytical closures of the plasma chemical rates. This article aims to assess the impact of Blanchard's model on numerical predictions in a 3-D configuration, the Mini-PAC bluff-body burner, by validating and comparing it against experimental data and Castela's model. Simulations are performed by combining the low-CPU cost plasma discharge model with an analytical reduced combustion mechanism. Both discharge models are comparable and agree with the measurements in the steady-state pulsing and flame-growing regimes. Some differences can be observed in transient regimes. Nevertheless, Blanchard's model is more detailed than Castela's, while having a similar CPU cost
Stabilization of Integral Difference Equations by solving a Corona problem
No full textThis paper proposes a stabilizing state-feedback control law for vector-valued state systems with a scalar control input, governed by a general class of integral difference equations that incorporate both pointwise and distributed input delays.The proposed controller is expressed through integral operators acting on the state and input histories over a finite time horizon. Closed-loop stability is established by characterizing the controller kernels as solutions to a convolution equation arising from a Corona problem. The existence of such solutions is ensured under a suitable spectral stabilizability condition, and a least-square procedure is implemented to find them numerically.The approach extends existing IDE stabilization results to more general settings, allowing for arbitrary numbers of pointwise delays affecting both the state and input, without requiring commensurability assumptions
Varied mutual growth inhibition between commensal yeasts and E. coli strains
No full textInternational audienceInterkingdom interactions between bacteria and fungi are an emerging research field that provides insights into pathological, environmental, and microbiota-related relationships. However, the mechanisms governing these interactions, particularly in the context of microbial resistance, remain largely unknown. This study aims to enhance our understanding of the complex interactions between different Candida, Nakaseomyces and Sacharomyces species from the human microbiota and two not isogenic strains of Escherichia coli (antibiotic-susceptible E. coli -ATCC and multidrug-resistant E. coli -OXA48). Forty-nine Candida strains were co-cultured with the two E. coli strains. Both bacterial and yeast growth was monitored using flow cytometry and compared to monocultures. The effect of yeast culture supernatants on E. coli proliferation was also investigated. Metabolomic fingerprints and metabolite identification were performed using mass spectrometry-based approaches followed by multiblock statistical analyses. The inhibitory powers (IP) of yeasts against E. coli and vice versa varied significantly among fungal species. N. glabrata exhibited the strongest inhibition against E. coli -ATCC, while Candida lusitaniae, C. kefyr, C. krusei, C. tropicalis , and C. dubliniensis showed lower IPs. C. parapsilosis and Saccharomyces cerevisiae had no inhibitory effects. Against E. coli -OXA48, most yeasts displayed no inhibition, except for N. glabrata . Conversely, E. coli inhibited yeast growth more effectively, particularly Candida albicans . Fungal supernatants from S. cerevisiae, C. lusitaniae , and N. glabrata showed the highest inhibitory effects on E. coli -ATCC, while S. cerevisiae, C. krusei , and C. lusitaniae were most effective against E. coli -OXA48. Unsupervised metabolite profiling data analysis with multiblock approach highlighted a clustering of samples according to yeast species. Regarding inhibitory power on E. coli (ATCC or OXA48), active supernatants tend to cluster together suggesting the presence of similar metabolites; some were further characterized. This study highlights the diverse interactions between E. coli and commensal yeasts. From an applied perspective, these findings pave the way for identifying probiotics or postbiotics with potential applications in combating multidrug-resistant bacteria through novel antimicrobial compounds
Resource-Aware Quantum Programming with General Recursion and Quantum Control
No full textThis paper introduces the hybrid quantum language with general recursion Hyrql, driven towards resource-analysis. By design, Hyrql does not require the specification of an initial set of quantum gates. Hence, it is well amenable towards a generic cost analysis, unlike languages that use different sets of quantum gates, which yield quantum circuits of distinct complexity.Regarding resource-analysis, we show how to relate the runtime of an expressive fragment of Hyrql programs with the size of the corresponding quantum circuits. We also manage to capture the class of functions computable in quantum polynomial time, which, by Yao's Theorem, corresponds to families of circuits of polynomial size. Consequently, this result paves the way for the use of termination and runtime-analysis techniques designed for classical programs to guarantee bounds on the size of quantum circuits
Améliorer la performance humaine grâce à une vision augmentée
No full textAugmented Vision (AV) is an emerging field that aims to enhance human sight by directly modulating a user's view of the real world. Unlike traditional visual aids such as spectacles or contact lenses that provide static correction, AV systems offer dynamic, real-time modulation of the real world. This capability to actively change the user's view opens up new possibilities for both correcting and augmenting human vision.One of the three most common applications for AV is vision correction. This is particularly relevant for presbyopia, an age-related condition affecting 1.8 billion people worldwide. Conventional solutions, such as progressive glasses or multifocal contact lenses, offer a static compromise with fixed focal zones or averaged focus, respectively. In contrast, AV systems use their inherently dynamic behaviour to address these limitations. For instance, some systems use automatic single focus approaches using focus-tunable lenses to focus at specific depths. However, this approach is limited to applying one optical power across the entire field of view at a time. The alternatives are multifocal systems, which provide granular, spatial control over focus. However, while the concept has been demonstrated, there is a lack of objective evaluation of the imaging quality of such systems.This thesis addresses two primary objectives. The first is to establish a conceptual framework for Augmented Vision by stating a formal definition and a new taxonomy that categorises existing systems by their light modulation method. The second is to address a gap identified by this framework: the lack of multifocal augmented vision systems and the evaluation of their imaging performance.To tackle the challenges of providing multifocal vision, this thesis proposes an optical architecture that combines a phase-only Spatial Light Modulator (SLM) with a Lohmann lens, allowing spatial control of the optical power over the Field Of View (FOV), providing see-through, real-world modulation. We validate the prototype by measuring the Modulation Transfer Function (MTF) for global and multifocal correction, placing objects at various depths and setting the system's focus at them. We use a camera with its focus set to infinity to simulate a presbyopic eye during the measurements. The results confirm that the system can bring objects from multiple depths into focus, demonstrating a higher MTF for each focused plane when the system is active compared to the inactive condition.This thesis provides two primary contributions: (1) a formal conceptual framework for AV systems that identifies the research gap, and (2) a prototype of an optical platform for multifocal vision and its evaluation. This platform serves as foundational technology for exploring a new class of interfaces based on spatially-variant optical power to enhance visual perception and address widespread visual impairments.La vision augmentée (VA) est un domaine émergent qui vise à améliorer la vision humaine en modulant directement la perception qu'a l'utilisateur du monde réel. Contrairement aux aides visuelles traditionnelles telles que les lunettes ou les lentilles de contact qui offrent une correction statique, les systèmes de VA permettent une modulation dynamique et en temps réel du monde réel. Cette capacité à modifier activement la perception de l'utilisateur ouvre de nouvelles possibilités pour corriger et augmenter la vision humaine. L'une des trois applications les plus courantes de la VA est la correction de la vision. Cela est particulièrement pertinent pour la presbytie, une affection liée à l'âge qui touche 1,8 milliard de personnes dans le monde. Les solutions conventionnelles, telles que les lunettes progressives ou les lentilles de contact multifocales, offrent un compromis statique avec des zones focales fixes ou un contraste réduit, respectivement. En revanche, les systèmes VA utilisent leur comportement intrinsèquement dynamique pour pallier ces limites. Par exemple, certains systèmes utilisent des approches à mise au point unique automatique avec des lentilles à mise au point réglable pour faire la mise au point à des profondeurs spécifiques. Cependant, cette approche se limite à l'application d'une seule puissance optique sur l'ensemble du champ de vision à la fois. Les alternatives sont les systèmes multifocaux, qui offrent un contrôle spatial granulaire de la mise au point. Cependant, bien que le concept ait été démontré, il manque une évaluation objective de la qualité d'image de ces systèmes. Cette thèse aborde deux objectifs principaux. Le premier est d'établir un cadre conceptuel pour la vision augmentée en énonçant une définition formelle et une nouvelle taxonomie qui classe les systèmes existants en fonction de leur méthode de modulation de la lumière. Le second est de combler une lacune identifiée par ce cadre : l'absence de systèmes de vision augmentée multifocaux et l'évaluation de leurs performances d'imagerie. Pour relever les défis liés à la vision multifocale, cette thèse propose une architecture optique qui combine un modulateur spatial de lumière (SLM) à phase uniquement avec une lentille de Lohmann, permettant un contrôle spatial de la puissance optique sur le champ de vision et offrant une modulation transparente et réaliste. Nous validons le prototype en mesurant la fonction de transfert de modulation (FTM) pour la correction globale et multifocale, en plaçant des objets à différentes profondeurs et en réglant la mise au point du système sur ceux-ci. Nous utilisons une caméra dont la mise au point est réglée sur l'infini pour simuler un œil presbyte pendant les mesures. Les résultats confirment que le système peut mettre au point des objets à plusieurs profondeurs, démontrant une FTM plus élevée pour chaque plan mis au point lorsque le système est actif par rapport à l'état inactif. Cette thèse apporte deux contributions principales : (1) un cadre conceptuel formel pour les systèmes VA qui identifie les lacunes de la recherche, et (2) un prototype de plateforme optique pour la vision multifocale et son évaluation. Cette plateforme sert de technologie de base pour explorer une nouvelle classe d'interfaces basées sur la puissance optique variable dans l'espace afin d'améliorer la perception visuelle et de remédier aux déficiences visuelles courantes
Décomposition parcimonieuse conjointe de signaux basée sur le transport optimal entre mélanges
No full textOptimal transport is now ubiquitous in data science. Its popularity stems from the fact that distances derived from optimal transport allow for the comparison of probability distributions of potentially different natures (discrete and/or continuous), and that there are efficient numerical methods for calculating these distances.A recent trend in signal processing is to integrate mathematical tools based on optimal transport to solve ill-posed inverse problems, in various application contexts ranging from neuroimaging and video analysis to identifying the impulse response of a room in terms of acoustics.In inverse problems, optimal transport is used to model prior knowledge about the distribution of the signal to be reconstructed. Signal reconstruction typically involves solving a convex, penalized least squares problem, which is significantly more challenging to solve than a simple optimal transport problem. One of the difficulties arises from the fact that the penalty term cannot be expressed analytically and requires solving a transport problem itself.This manuscript focuses on solving problems involving the decomposition or deconvolution of a sequence of signals, using regularizations that combine sparse representation in a dictionary and optimal transport on positive signals. Optimal transport is used to constrain the dynamic evolution of the signals in the sequence to be regular.A penalized least squares formulation is proposed, where the convex penalty promotes proximity between successive sparse decompositions. The originality of this formulation lies in its use of optimal transport distance computation methods between mixtures of distributions. Two optimal transport penalties are proposed, one of which is based on multi-marginal transport (to match signals from several consecutive channels). To promote greater sparsity, we propose a variant involving a slightly non-convex regularization term. Dedicated proximal optimization algorithms are proposed, along with their accelerated versions.This method is specified in the context of dynamic sparse deconvolution problems. The empirical study highlights the intrinsic limitations of the method for reconstructing linear dynamics, due to the discretization grid used to reconstruct the signal. It is empirically and theoretically demonstrated that the accuracy of the method improves significantly when applied in the context of super-resolution, where the sampling grid of the signal to be reconstructed is finer than that of the observed signals.Le transport optimal est aujourd'hui omniprésent en science des données. Sa popularité vient du fait que les distances issues du transport optimal permettent de comparer des distributions de probabilités de nature possiblement différentes (discrètes et/ou continues), et qu'il existe des méthodes numériques efficaces pour calculer ces distances.Une tendance récente en traitement du signal est d'intégrer des outils mathématiques issus du transport optimal pour résoudre des problèmes inverses mal posés, dans des contextes applicatifs variés allant de la neuro-imagerie, l'analyse de vidéos jusqu'à l'identification de réponse impulsionnelle d'une salle en acoustique.En problèmes inverses, le transport optimal sert à modéliser les connaissances a priori sur la distribution du signal à reconstruire. La reconstruction du signal conduit généralement à résoudre un problème de moindres carrés pénalisés convexe, mais largement plus difficile à résoudre qu'un simple problème de transport optimal. L'une des difficultés vient du fait que le calcul du terme de pénalisation n'a pas d'expression analytique, et nécessite lui-même de résoudre un problème de transport optimal.Dans ce manuscrit, on s'intéresse à la résolution de problèmes de type décomposition ou déconvolution d'une séquence de signaux, avec des régularisations couplant représentation parcimonieuse dans un dictionnaire et transport optimal sur des signaux positifs. Le transport optimal est exploité pour contraindre l'évolution dynamique des signaux de la séquence à être régulière.Une formulation de type moindres carrés pénalisés est proposée, où la pénalité, convexe, promeut la proximité entre les décompositions parcimonieuses successives. L'originalité de cette formulation est d'exploiter des méthodes de calcul de distances de type transport optimal entre mélanges de distributions. Deux pénalisations de type transport optimal sont proposées dont une basée sur le transport multi-marginales (pour mettre en correspondance des signaux émanant de plusieurs canaux consécutifs). Afin de promouvoir plus fortement la parcimonie, nous proposons une variante qui met en jeu un terme de régularisation légèrement non-convexe. Des algorithmes d'optimisation proximale dédiés sont proposés ainsi que des versions accélérées.Cette méthode est spécifiée dans le cadre de problèmes de déconvolution parcimonieuse dynamique. L'étude empirique met en évidence les limites intrinsèques de la méthode pour reconstruire des dynamiques linéaires, à cause de la grille de discrétisation du signal à reconstruire. On montre empiriquement et théoriquement que la précision de la méthode s'améliore nettement en travaillant dans le cadre de la super-résolution, où la grille d'échantillonnage du signal à reconstruire est plus fine que celle des signaux observés
Ferrite Plate Multi-Objective Topology Optimization in a Wireless Power Transfer System for UAV Charging
No full textInternational audienceWireless power transfer (WPT) systems offer a promising solution for automating the charging process and extending the mission duration of unmanned aerial vehicles (UAVs). However, the weight of the WPT system's receiving part mounted on the UAV can significantly impact its energy consumption and overall performance. In this regard, the ferrite plate is a critical component. While it is necessary for improving magnetic coupling and reducing the stray magnetic field, it also increases weight. Topology optimization (TO) is an effective tool for addressing this challenge and identifying the most suitable shape for the ferrite plate. In this study, a multi-objective topology optimization (TO) approach based on the solid isotropic material with penalization (SIMP) method is employed to design an optimized ferrite plate for the receiver side of a WPT system for UAV charging. The multi-objective optimization, implemented through a weighted-sum formulation, identifies ferrite configurations that maximize mutual coupling while minimizing material usage. Three representative Pareto-optimal designs are fabricated and experimentally validated against a full-ferrite reference plate. The results show that the reference configuration is a dominated solution in the Pareto sense, as equivalent mutual coupling can be achieved with approximately 40% less ferrite material. The trade-off between the resulting reduction in charging efficiency and the decrease in the required flight power due to ferrite reduction is analyzed, further demonstrating the potential of TO to substantially reduce system weight without compromising overall system performance
Image-Guided Autonomous Robotic Surgery in the Context of Therapies Managed by Intelligent Digital Technologies: A Narrative Review
No full textInternational audienceThis narrative review aims to highlight and analyze the supervision of precision robotic surgical interventions. These are autonomous, closed-loop procedures, assisted by images and managed by intelligent digital tools. These administered procedures are designed to be safe and reliable, adhering to the principles of minimal invasiveness, precise positioning, and non-toxicity. Thus, a precision intervention uses non-ionizing imaging-assisted robotics, controlled by a precise positioning device, forming an autonomous procedure augmented by artificial intelligence tools and supervised by digital twins. This intelligent digital management procedure allows staff to plan, train, predict, and execute interventions under human supervision. Patient safety and staff efficiency are linked to non-ionizing imaging, minimal invasiveness through image guidance, and strict delimitation of the intervention zone through precise positioning. This study includes, successively, sections covering an introduction, therapeutic and surgical interventions, imaging strategies integrating diagnostic and assistance functions, intelligent digital tools including digital twins and artificial intelligence, image-guided procedures including autonomous and precision robotic surgical interventions increased by machine learning, as well as augmented healthcare monitoring, and a discussion and conclusions of the review. All topics addressed in this analysis are supported by examples from the literature