Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
A coclustering and computational intelligence-based approach for internet-of-things services composition
No full textInternational audienceThe Internet of Things (IoT) paradigm aims at interconnecting heterogeneous devices, called smart objects and seamlessly offering a multitude of services tailored to the user requirements. With the extremely rapid growth of the number of connected objects, the IoT services composition process becomes an NP-hard challenge due to the very high increase of the number of services offering similar functionalities but that may differ in their Quality of Service (QoS) parameter values. Various approaches have been proposed in the literature to obtain compositions with suboptimal QoS in a reasonable computation time. However, when the number of services and QoS parameters increases, the performance of these approaches is limited in terms of the composition time and/or the QoS utility of the composition. To address these limitations, a coclustering-based approach for QoS-constrained services composition (CoQSC) is proposed to reduce the composition space and improve the composition time as well as the composition utility. Unlike existing services composition algorithms where the composition space is reduced only in terms of the number of candidate services, the CoQSC approach exploits a coclustering method to reduce both the number of candidate services and the number of QoS parameters to be considered in the composition process. This reduction allows the composition process to find suboptimal compositions in a reduced computation time using eight among the most representative and recent computational intelligence (CI) techniques in the literature separately. The formulation of the CoQSC approach is complemented by a complexity analysis. Simulation scenarios show that the CoQSC approach significantly improves the QoS utility of composition and substantially decreases the composition time compared to recent and representative state-of-the-art composition approaches, making it suitable for large-scale IoT service environments
Spectral vs. Spatial Pruning: Fault Resilience Analysis for Extremely Sparse DNNs
No full textInternational audienc
Géométrie plane des q-rationnels et les opérations de Springborn
No full textInternational audienceWe study the geometry of q-rational numbers, introduced by Morier-Genoud and Ovsienko, for positive real q. In particular, we construct and analyse the deformed Farey triangulation and the deformed modular surface. We interpret every q-rational geometrically as a circle, similar to the famous Ford circles. Further, we define and study new operations on q-rationals, the Springborn operations, which can be seen as a quadratic version of the Farey addition. Geometrically, the Springborn operations correspond to taking the homothety centers of a pair of two circles. Contents 1. Introduction 1 2. From q-integers to q-rationals to q-reals 5 3. Hyperbolic geometry and deformed Farey tesselation 9 4. Deformed Farey determinants and operations 18 5. Classical Springborn operations 24 6. Springborn operations for q-rationals 32 7.</div
Effect of the temperature on the impedance control of pressure-based, current-driven electroacoustic absorbers: Addressing the loss of passivity using a viscoelastic material model
No full textInternational audienceIn active noise control, pressure-based control strategies forelectroacoustic absorbers depend on the loudspeakers’electromechanical properties, known as Thiele–Small parameters, toimplement impedance control. Due to the viscoelastic nature ofloudspeaker materials, these parameters are sensitive toenvironmental conditions, particularly temperature. This studyinvestigates the impact of temperature on the impedance control ofelectroacoustic absorbers. The acoustic impedance of severalabsorbers is measured over a broad temperature range, and ananalytical model is used to identify the variation of theThiele–Small parameters with temperature. A viscoelastic materialcharacterization framework is then proposed, employing theFractional Zener, Generalized Maxwell, and Generalized FractionalMaxwell models. These models are identified for individualabsorbers and compared in terms of accuracy and computational cost.A generalized approach through a normalized curve derived frommultiple absorbers is introduced to estimate the parameters ofunknown absorbers. The pressure-based control law is subsequentlyupdated to include temperature-dependent parameters, enablingevaluation of their influence on absorber passivity. Resultsdemonstrate that adapting the control strategy using either directmeasurements or model-based estimations enhances the acousticpassivity of electroacoustic absorbers
The calf holobiont under challenge: longitudinal microbiome-pathogen dynamics and respiratory health
No full textThe Bovine Respiratory Disease Complex (BRDC) is a major health and welfare challenge driven by multifactorial interactions among pathogenic viruses and bacteria, host and environmental factors, and the respiratory and gut microbiomes. Many implicated bacterial pathogens are also commensals of the respiratory tract, complicating diagnosis and prevention. Growing evidence indicates the roles of the respiratory and gut microbiomes in BRDC, yet their combined effects remain poorly understood. In this study, we dynamically followed the microbiome, pathogen load, and host response in 30 calves over 147 days under commercial rearing conditions. Nearly half developed BRDC, with elevated fever, cough, and lung sound scores, and peak symptoms at day 58 after feedlot arrival. Pathogen detection in nasal cavities revealed distinct patterns: Mycoplasma bovis and BCoV peaked during the first two weeks, IDV and Histophilus somni after one month, Mannheimia haemolytica after two, while Pasteurella multocida peaked after one month and persisted. Importantly, M. haemolytica and Pasteurella multocida loads correlated with higher BRDC scores, whereas BCoV and Mycoplasma bovis were associated with diarrhea, suggesting systemic effects. Nasal beta-diversity diverged between groups at the symptomatic window, and healthy animals exhibited higher fecal diversity and evenness early in life. The respiratory pathobionts Pasteurella and Corynebacterium were enriched in diseased calves, whereas potentially protective families (Lachnospiraceae, Oscillospiraceae) were more abundant in healthy ones. Multivariate analyses confirmed that antibiotic treatments and short-chain fatty acids, especially isovalerate and isobutyrate, further modulated both fecal and nasal microbiomes, with consistently stronger impacts in diseased animals. Together, these findings demonstrate that BRDC outcomes are shaped not by pathogen burden alone, but by the interplay among respiratory and digestive microbiotas, pathogens, environment, and host factors. Our study highlights the importance of a holobiont perspective that integrates both gut and respiratory microbiotas to better elucidate the complexity of BRDC. Such an inclusive framework may provide new insights into disease mechanisms and inform the development of innovative therapeutic strategies
Des réseaux de neurones sur graphes auto-explicatifs basés sur la logique
No full textInternational audienceLes graphes sont des structures complexes et non euclidiennes qui nécessitent des modèles spécialisés comme les réseaux de neurones sur graphes (Graph Neural Networks, GNNs) pour capturer efficacement les motifs relationnels associés à la variable de classe. Cette complexité intrinsèque rend particulièrement difficile l’explication des décisions prises par les GNNs. La plupart des méthodes actuelles d’intelligence artificielle explicable (XAI) appliquées aux GNNs se concentrent sur l’identification de nœuds influents ou l’extraction de sous-graphes pertinents, sans toutefois clarifier comment ces éléments contribuent réellement à la prédiction finale. Pour dépasser cette limite, les approches à base logique visent à dériver des règles explicites reflétant le raisonnement du modèle. Cependant, les méthodes logiques existantes demeurent majoritairement post-hoc et se limitent à la classification de graphes, laissant un manque important en matière d’architectures intrinsèquement explicables. Dans cet article, nous intégrons le raisonnement logique directement au sein du modèle d’apprentissage sur graphes. Nous introduisons LogiX-GIN, une nouvelle architecture de GNN auto- explicable qui incorpore des couches logiques afin de produire des règles logiques interprétables au cœur même du processus d’apprentissage. Contrairement aux approches post-hoc, LogiX-GIN fournit des explications transparentes, fidèles et cohérentes avec les calculs internes du modèle. Évalué sur plusieurs tâches basées sur des graphes, LogiX-GIN atteint des performances prédictives compétitives tout en explicitant son processus décisionnel. Ces travaux ont été acceptés à NeurIPS 202
A complete answer to the strong density problem in Sobolev spaces with values into compact manifolds
No full textPermanent degradation of p-GaN HEMTs due to repetitive overvoltage stress during hard turn-off switching
No full textInternational audienceThis study investigates the long-term impact of dynamic overvoltage stress on GaN HEMTs using a newly designed test circuit, UIS3, a variant of classic UIS, which isolates key stress factors. Devices were subjected to short-duration repetitive overvoltage stress near and below their dynamic breakdown voltage. Characterization before and after stress reveals permanent degradation in CDS, CDS and IGSS, suggesting deep-trapping or structural damage within the device. A distinct alteration in the CDS curve is observed, may indicate less spreading of the electric-field within the device. RDS,on degradation is also noted, likely due to trapping effects, with partial recovery at room temperature. Higher stress levels accelerate failure. Waveform analysis and post-failure characterization indicate a short-circuit failure mode, likely due to partial dielectric breakdown during overvoltage events. These results provide new insights into GaN HEMT degradation mechanisms under high-voltage stress
Diffusion-Based Authentication of Copy Detection Patterns: A Multimodal Framework with Printer Signature Conditioning
No full textInternational audienceCounterfeiting affects diverse industries, including pharmaceuticals, electronics, and food, posing serious health and economic risks. Printable unclonable codes, such as Copy Detection Patterns (CDPs), are widely used as an anti-counterfeiting measure and are applied to products and packaging. However, the increasing availability of high-resolution printing and scanning devices, along with advances in generative deep learning, undermines traditional authentication systems, which often fail to distinguish high-quality counterfeits from genuine prints. In this work, we propose a diffusion-based authentication framework that jointly leverages the original binary template, the printed CDP, and a representation of printer identity that captures relevant semantic information. Formulating authentication as multi-class printer classification over printer signatures lets our model capture fine-grained, device-specific features via spatial and textual conditioning. We extend ControlNet by repurposing the denoising process for class-conditioned noise prediction, enabling effective printer classification. On the Indigo 1 × 1 Base dataset, our method outperforms traditional similarity metrics and prior deep learning approaches. Results show the framework generalizes to counterfeit types unseen during training
Automatic seal quality inspection using deep learning in mono material flexible packaging
No full textInternational audienceThe automated detection of seal flaws with thermal imaging represents one of the major problems in the manufacturing process. This paper suggests a thermal imaging based deep learning method of detecting sealing faults of mono-material flexible packaging. We compare multiple pre-trained Convolutional Neural Networks to detect more accurately and run faster. Three experimental tests are carried out to enhance the classification process. Another way of reducing model size and computational load with pruning and quantization is required to run it on edge devices. With high-performance NVIDIA GPUs, the adjusted models have an accuracy of 98.7\%, precision of 80.69\%, and recall of 88.89\%. This method performs effectively in real-time seal defect identification and may be applied in packaging quality checks in the industrial line. The model requires tuning to operate more effectively and adapt to a variety of production circumstances and packaging