Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
Variable selection in sparse multivariate GLARMA models: application to germination control by environment
No full textInternational audienceWe propose an iterative variable selection approach in multivariate sparse GLARMA models for modeling multivariate discrete-valued time series. The estimation in our approach is performed in two steps: firstly, our approach estimates the autoregressive moving average (ARMA) coefficients of multivariate GLARMA models, followed by variable selection in the coefficients of the Generalized Linear Model using regularized methods. We provide a detailed description of the implementation of our approach. Subsequently, we study its performance on simulated data and compare it with other methods. Finally, we illustrate its application on RNA-Seq data resulting from polyribosome profiling to determine translational status for all mRNAs in germinating seeds. The proposed approach benefits from a number of attractive features: it has a low computational load and outperforms other methods in accurately performing variable selection and, consequently, recovering the null and non-null coefficients. Furthermore, being implemented in the MultiGlarmaVarSel R package and openly accessible on the CRAN, our variable selection method holds significant appeal for broader applications across diverse scientific disciplines
Path-Based Explanations for Knowledge Graph-Driven Course Recommendation
No full textInternational audienceRecommendation systems (RS) play a key role in e-learning by guiding learners toward relevant educational resources. Yet, the integration of domain knowledge to enhance both accuracy and explainability remains underexplored. This paper presents an explainable e-learning RS grounded in an educational Knowledge Graph (KG). The KG is constructed by extracting and linking key course concepts, and leveraged through embedding techniques to improve recommendation quality. To ensure transparency, we propose a path-based explanation mechanism that identifies and ranks user-course connections using a scoring function combining similarity measures and random walk probabilities. A case study demonstrates that the approach not only improves recommendation accuracy but also generates diverse, interpretable explanations, contributing to more transparent and trustworthy systems.</div
Systèmes de traitement et de stockage distribué confidentiels basés sur les TEEs
No full textData is the new fuel of the digital landscape, driving innovation, decision-making, and business strategies across industries. In this data-driven world, information has become a powerful asset, but it also comes with great responsibility. Many datasets contain sensitive personal, financial, or corporate information that requires strict protection. Safeguarding this data is crucial, as breaches can lead to significant privacy violations, financial losses, and damage to trust. Therefore, handling data responsibly is not only a matter of security but also of ethics and compliance in today's interconnected world. To ensure robust data security, protection measures must be applied across all three states of data: in transit, at rest, and during processing. Data in transit, moving between systems or across networks, is vulnerable to interception and requires encryption and secure transmission protocols. Data at rest, stored on servers, databases, or other storage media, must be safeguarded with strong encryption, access controls, and physical security measures. Meanwhile, data in use, actively being processed by applications or systems, demands strict access controls, monitoring, and secure computing environments to prevent unauthorized access or leaks. Comprehensive security across these states is essential to maintaining confidentiality in an increasingly data-centric world. In particular, Trusted Execution Environments (TEEs) provide secure, isolated computing environments that protect sensitive data and code in main memory from unauthorized access, even from privileged system processes or the operating system. This research is divided into two main parts. The first part focuses on securing data in use, studying federated learning (FL) as a particular use case for data processing systems. FL is chosen due to its inherent predisposition for privacy-preserving machine learning, despite its susceptibility to security threats. Spyware can monitor the main memory of FL client devices to extract information about the private data used to train the FL model through various leakage attacks. In response to this threat, we developed gs, a TEE-based defense mechanism to secure FL against leakage attacks. gs leverages ARM TrustZone, a TEE for mobile devices, to secure sensitive FL model layers based on each surveyed leakage attack. The second part focuses on securing data at rest, leveraging in-memory key-value stores (KVSs) as a storage medium. The choice of KVSs is driven by their schema-less design, which allows for storing various types of data by serializing them when necessary. In this part, we conducted a comprehensive survey of existing TEE-based KVSs in literature, identifying common architectures and building blocks while emphasizing their vulnerability to side-channel leakage attacks despite TEE protection. To mitigate this vulnerability, we propose s, a confidential distributed KVS that harmoniously combines TEEs and secret sharing to split sensitive data across multiple TEE nodes, making data leakage significantly more challenging for an attacker. These research objectives collectively aim to address security challenges, particularly confidentiality, in untrusted environments where data may be processed and stored.Les données sont devenues un moteur essentiel du monde numérique, alimentant l'innovation et les stratégies commerciales. Toutefois, leur manipulation s'accompagne d'importants enjeux de sécurité et d'éthique, car elles peuvent contenir des informations sensibles nécessitant une protection stricte. Toute violation peut entraîner des atteintes à la vie privée, des pertes financières et une perte de confiance. Pour garantir leur sécurité, il est crucial de les protéger dans leurs trois états : en transit, au repos et en cours d'utilisation. Les données en transit doivent être chiffrées pour éviter toute interception, celles au repos sécurisées par des contrôles d'accès stricts et un chiffrement robuste, tandis que celles en cours de traitement nécessitent un environnement protégé pour empêcher tout accès non autorisé. Les environnements d'exécution fiables jouent un rôle clé en isolant les données sensibles, même vis-à-vis des processus système ou du système d'exploitation. Cette recherche se divise en deux parties principales. La première porte sur la sécurisation des données en cours d'utilisation, en étudiant l'apprentissage fédéré comme cas d'usage spécifique. Ce paradigme d'apprentissage automatique préservant la confidentialité reste vulnérable aux attaques visant la mémoire principale des dispositifs clients, pouvant révéler des informations sensibles sur les données utilisées pour l'entraînement du modèle. Face à ces menaces, GradSec a été développé pour sécuriser l'apprentissage fédéré contre ces attaques. Cette solution exploite ARM TrustZone, un environnement d'exécution fiable conçu pour les appareils mobiles, afin de protéger les couches critiques du modèle en fonction des attaques identifiées. La seconde partie traite de la sécurisation des données au repos en s'appuyant sur les bases de données clé-valeur en mémoire, choisies pour leur flexibilité et leur capacité à stocker divers types de données via sérialisation. Une analyse approfondie des bases de données clé-valeur basées sur des environnements d'exécution fiables a mis en évidence leurs architectures et composants fondamentaux, tout en révélant leur vulnérabilité aux attaques par canaux auxiliaires. Pour atténuer ce risque, TruShare a été conçu comme un système de stockage distribué confidentiel combinant environnements d'exécution fiables et partage de secret. En fragmentant les données sensibles entre plusieurs nœuds, cette approche rend les fuites de données beaucoup plus complexes pour un attaquant.Ces travaux de recherche visent à répondre aux défis de la sécurité des données dans des environnements non fiables, en garantissant une confidentialité renforcée aussi bien lors du traitement que du stockage des informations
A Graphical Tool for Predicting Class EF Inverter Behavior Including Non-Ideal Load Conditions
No full textInternational audienceThis paper presents a novel analytical framework for the design and understanding of class EF inverters under both optimal and non-optimal load conditions. Unlike conventional approaches that rely heavily on numerical simulations, the proposed method provides a fast, visual, and intuitive tool for analyzing inverter operation. Its effectiveness is demonstrated experimentally on a 15 MHz class EF inverter across three distinct load conditions, showing good agreement with theoretical predictions. To highlight the robustness and broad applicability of the approach, a class Φ2 inverter—a lumped-element analog of the class EF inverter—is also implemented and successfully analyzed. By combining theoretical insight, experimental validation, and generalization to alternative topologies, the proposed framework offers an efficient, accessible, and versatile tool for high-frequency resonant inverter design
How transfer film formation in bronze/silver-graphite sliding contact drives its electrical performance
No full textInternational audienceRecent interest in improving lifespan of electrical slip rings for application in wind turbines has renewed the need for deeper understanding of sliding electrical contact operation. In contact slip rings using graphite-baséd brush téchnology, thé formation of transfér film, or "patina", on thé surface of the metal ring is a well-known phenomenon [1], [2]. In this scope, the study of its formation in a bronze/silver-graphite contact has been conducted. The test bench is based on a real industrial slip ring specific for wind turbine application. The evolution of the transfer film was investigated by means of different surface characterization methods. EDX analyses permitted to establish a direct relation between transfer chemical composition, transfer thickness (h) and electrical resistance of the contact (Rc), highlighting the role of the transfer film in slip ring electrical performance. In particular, the presence of contaminating Cu2O oxides is emerging as a key factor in controlling both h and Rc. Optical microscopy and profilometry, allowed to propose a scenario of build-up following by destruction of the transfer film, explaining at the same time the origin of these Cu2O oxides. Focused Ion Beam (FIB) nanomachining, followed by Transmission Electron Microscopy (TEM) enrich this scenario and correlates the surface observations, revealing a strong evolution of the composition and morphology of the transfer pads from very carbonaceous, poorly graphitized transfers to a much more metallic and graphitized composition. In this last state, a lamellar structure is observed, with successive lamellae of silver separated by thin graphite layers. Raman spectroscopy also permits to study the crystalline structure of the graphite transferred on the ring surface and shows a progressive re-organization of the crystallites. While this reorganisation should be beneficial in a tribological perspective, it is important to note that this mechanism is entirely independent of both the growth-destruction of the transfer film and the evolution of Rc
Erbium doped yttrium oxide thin films grown by chemical vapour deposition for quantum technologies
No full textInternational audienceThe obtention of quantum-grade rare-earth doped oxide thin films that can be integrated with optical cavities and microwave resonators is of great interest for the development of scalable quantum devices. Among the different growth methods, Chemical Vapour Deposition (CVD) offers high flexibility and has demonstrated the ability to produce oxide films hosting rare-earth ions with narrow linewidths. However, growing epitaxial films directly on silicon is challenging by CVD due to a native amorphous oxide layer formation at the interface. In this manuscript, we investigate the CVD growth of erbium-doped yttrium oxide (Er:Y2O3) thin films on different substrates, including silicon, sapphire, quartz or yttria stabilized zirconia (YSZ). Alternatively, growth was also attempted on an epitaxial Y2O3 template layer on Si (111) prepared by molecular beam epitaxy (MBE) in order to circumvent the issue of the amorphous interlayer. We found that the substrate impacts the film morphology and the crystalline orientations, with different textures observed for the CVD film on the MBEoxide/Si template (111) and epitaxial growth on YSZ (001). In terms of optical properties, Er 3+ ions exhibit visible and IR emission features that are comparable for all samples, indicating a high-quality local crystalline environment regardless of the substrate. Our approach opens interesting prospects to integrate such films into scalable devices for optical quantum technologies
In-Mold Electronics with high electrical conductivity circuits on Poly(Lactic Acid) and Polycarbonate using photonic curing
No full textInternational audienceIn-Mold Electronics (IME) integrates printed electronics into 3D-molded polymer parts, to produce lightweight, functional and easy to assemble devices. IME makes use of conductive inks that require heat curing making the process not well adapted for low glass transition (Tg) temperature polymers. This study presents the adaptation of IME to Poly(Lactic Acid) (PLA), a biosourced, and biodegradable polymer, replacing traditional heat curing with photonic curing.Characteristics of the electronic circuit using both curing methods were compared for PLA and Polycarbonate. The use of pulsed light heats locally conductive inks and adhesives preventing deforming or degrading the polymer substrate due to PLA's low Tg temperature. Our approach ensured high electrical conductivity of ME603 silver ink and ME902 conductive adhesive while maintaining conductive track continuity after stretching by thermoforming of PLA. Fully functional IME demonstrators present potential applications. This work highlights photonic curing's capability for enabling sustainable electronics manufacturing with recyclable polymers, maintaining performance, and reducing production time
Iterative Collaboration Network Guided by Reconstruction Prior for Medical Image Super-Resolution
No full textInternational audienceHigh-resolution medical images can provide more detailed information for better diagnosis. Conventional medical image super-resolution relies on a single task which first performs the extraction of the features and then upscaling based on the features. The features extracted may not be complete for super-resolution. Recent multi-task learning, including reconstruction and super-resolution, is a good solution to obtain additional relevant information. The interaction between the two tasks is often insufficient, which still leads to incomplete and less relevant deep features. To address above limitations, we propose an iterative collaboration network (ICONet) to improve communications between tasks by progressively incorporating reconstruction prior to the super-resolution learning procedure in an iterative collaboration way. It consists of a reconstruction branch, a super-resolution branch, and a SR-Rec fusion module. The reconstruction branch generates the artifact-free image as prior, which is followed by a super-resolution branch for prior knowledge-guided super-resolution. Unlike the widely-used convolutional neural networks for extracting local features and Transformers with quadratic computational complexity for modeling long-range dependencies, we develop a new residual spatial-channel feature learning (RSCFL) module of two branches to efficiently establish feature relationships in spatial and channel dimensions. Moreover, the designed SR-Rec fusion module fuses the reconstruction prior and super-resolution features with each other in an adaptive manner. Our ICONet is built with multi-stage models to iteratively upscale the low-resolution images using steps of 2× and simultaneously interact between two branches in multi-stage supervisions. Quantitative and qualitative experimental results on the benchmarking dataset show that our ICONet outperforms most state-of-the-art approaches
An Ultra-low-loss Compact Phase-Change Material-based Hybrid-mode Interferometer for Photonic Memories
No full textInternational audienceWe propose a novel hybrid mode interferometer (HMI) leveraging the interference of hybridized TE-TM modes in a silicon-on-insulator (SOI) waveguide integrated with a GeSe phase change material (PCM) layer. The SOI waveguide's dimensions are optimized to support the hybridization of the fundamental transverse magnetic () and the first higher transverse electric () mode. This design allows for efficient and nearly equal power coupling between these two modes, resulting in high-contrast interference when starting from the amorphous PCM state. The PCM's phase transition induces a differential change in the modal effective index, enabling high-contrast transmittance modulation. Our numerical simulations demonstrate a multilevel transmission with a high contrast of nearly 14 dB, when the amorphous region's length is varied incrementally, enabling multi-bit storage. The transmittance is maximized in the fully crystalline state with an insertion loss below 0.1 dB. The HMI can also operate as a quasi-pure phase shifter when partially amorphized, making it suitable for Mach-Zehnder interferometers. These characteristics make the proposed device a promising candidate for applications in photonic memories and neuromorphic computing
Acoustofluidique de bulles de cavitation en milieu confiné
No full textPression de radiation, streaming, dynamique de bulles; GAPSUS - Acoustique Physique, Sous-Marine et Ultra-SonoreNational audienceDe nombreuses applications en ingénierie ou en santé utilisent des bulles de gaz soumises à des champs ultrasonores en tant que systèmes résonnants à haut facteur de qualité, qui permettent le nettoyage de surfaces ou la destruction de calculs rénaux par implosion de bulles, entre autres. L’étude des propriétés acoustiques (oscillation, émission acoustique) et fluidiques (mise en place d’un écoulement dans leur voisinage, appelé microstreaming) de ces bulles acoustiques permet de mieux appréhender leur interaction avec le milieu environnant. Une cuve de lévitation acoustique permettant de piéger une bulle entre deux parois et de déclencher ses oscillations à l’aide d’un transducteur ultrasonore a été conçue et sera présentée. Les interactions entre une bulle et des parois seront étudiées à deux niveaux : d'abord, l'effet des parois sur la dynamique de la bulle sera analysé, en se concentrant sur ses oscillations, sa fréquence de résonance et les seuils d'instabilité des ondes de surface. Ensuite, les écoulements générés par les oscillations de la bulle en présence de parois à proximité seront caractérisés. Par ailleurs, un modèle théorique basé sur le principe de la bulle image sera exposé, intégrant les interactions multiples entre la bulle, ses images, et les parois. Ce modèle permettra de prédire la réponse acoustique (dynamique) de la bulle ainsi que le champ de vitesse Lagrangien associé à ses oscillations