HAL Portal UO (Université d'Orléans)
Not a member yet
48841 research outputs found
Sort by
Pratiques éco-responsables, une question de génération ? Une étude exploratoire des pratiques post-usage dans le domaine du textile
No full textInternational audienceThe generation clash is a recurrent topic in the media, and the concern for ecology is often mentioned as one of the reasons. This research aims to capture the post-use practices, in the fashion sector, of two generations furthest apart in time, the “baby-boomers” and Generation Z. These post-consumer behaviors, still little studied in marketing, can shed light on more or less eco-responsible contributions to the circular economy. The study is based on an ethnographic approach, combining several tools and data collection sites from the two generations concerned. Our results, articulated around the post-use keep/separate dyad, reveal that different post-use practices predominate in each of the generations studied, and that some deserve to be encouraged because of their contribution to the circular economy. They also help to qualify the idea of opposition between generations.Le clash des générations est un sujet récurrent dans les médias et l’une des raisons souvent mentionnées de cette discorde est la préoccupation pour l’écologie. Cette recherche vise à mieux comprendre les pratiques post-usage, dans le secteur de la mode, de deux générations les plus éloignées dans le temps, les « baby-boomers » et la génération Z. Ces comportements post-consommatoires, encore peu étudiés en marketing, peuvent nous éclairer sur leurs contributions plus ou moins éco-responsables à l’économie circulaire. L’étude repose sur une approche ethnographique, qui combine plusieurs outils et sites de collectes de données auprès des deux générations concernées. Nos résultats, articulés autour de la dyade post-usage garder/se séparer, mettent au jour que des pratiques post-usage différentes prédominent dans chacune des générations étudiées, et que certaines méritent d’être encouragées en raison de leur contribution à l’économie circulaire. Ils viennent également nuancer cette idée d’opposition entre générations
How important is information about grandparents when selecting a dialect speaker?
No full textInternational audienceThis paper aims at setting a new criterion for selecting a local dialect speaker byinquiring about the origin of their parents, which accommodates to the reality inlarge cities. In Japanese dialectology, it is widely accepted that a dialect speakershould ideally be haenuki (‘native-born’), i.e. a person from a family who has livedin the survey area for three generations. However, this criterion has been usedhitherto without checking the assumptions that it is both justified and useful. After reexamining the data collected from three families by the National LanguageResearch Institute (1965) and conducting one survey, I conclude that the haenukicriterion is impractical, and not fully justified and that the most important criterionwhen selecting a representative speaker of a dialect should be having both parentsnative born
Higher-Spin Poisson Sigma Models and Holographic Duality for SYK Models
No full textInternational audienceSYK models provide an interesting playground for exploring the correspondence. We focus on a class of SYK models that exhibit higher-spin symmetry, whose gravity sector is described by a BF theory generalizing Jackiw--Teitelboim gravity to higher spins. We further develop this framework by constructing consistent interactions between higher-spin gauge fields and scalar matter fields. Two concrete realizations are proposed: Model A, arising from a deformation of the universal enveloping algebra of , and Model B, a perturbatively local Poisson sigma-model with an infinite-dimensional target space. While both capture higher-spin dynamics on , they differ in their algebraic structures and locality properties, thus offering complementary perspectives on higher-spin holography
Deep Learning-Based Composite Channel Estimation in Reconfigurable Intelligent Surfaces
No full textInternational audienc
Constraints on the mantle source of the recent Fani Maoré paroxysm (Mayotte) from noble gas (He-Ne-Ar) isotopes
No full textThe origin of intraplate magmatism remains a subject of active debate. Recent volcanic activity offshore of Mayotte, the easternmost volcanic island of the Comoros archipelago, has reignited extensive scientific investigations aimed at better understanding the geodynamic context of the region. Two main hypotheses are generally invoked to explain the observed volcanic activity: one involves a mantle plume, while the other attributes the magmatism primarily to tectonic processes, possibly linked to the East African Rift System. However, both models present limitations and are not entirely satisfactory. Noble gas isotopes are powerful tracers for distinguishing between these scenarios, as mantle plumes typically exhibit distinct isotopic signatures. We conducted stepcrushing analyses on pristine glassy, gas-rich basanitic samples from the most recent (2018-2021) submarine volcanic edifice of Mayotte, which erupted at a depth of over 3,000 meters. The resulting noble gas isotopic systematics are strikingly similar to those of a MORB-like mantle, with the notable exception of significant excesses in the radiogenic isotope 4 He. Our results suggest that a chemically distinct mantle plume is not required to explain the magmatism in the region, which may instead result from tectonic reactivation and thinning of an old, heterogeneous lithosphere
Dynamics of lean laminar premixed spherical hydrogen/air flames at varying initial temperatures
No full textInternational audienc
All-optical measurements of MORB-glass density at high pressure hint at a stiffness-composition relation in silicate glasses
No full textInternational audienc
Universal Prime Editing Therapeutic Strategy for RyR1-Related Myopathies: A Protective Mutation Rescues Leaky RyR1 Channel
No full textInternational audienceRyR1-related myopathies (RyR1-RMs) include a wide range of genetic disorders that result from mutations in the RYR1 gene. Pathogenic variants lead to defective intracellular calcium homeostasis and muscle dysfunction. Fixing intracellular calcium leaks by stabilizing the RyR1 calcium channel has been identified as a promising therapeutic target. Gene therapy via prime editing also holds great promise as it can cure diseases by correcting genetic mutations. However, as more than 700 variants have been identified in the RYR1 gene, a universal treatment would be a more suitable solution for patients. Our investigation into the RyR1-S2843A mutation has yielded promising results. Using a calcium leak assay, we determined that the S2843A mutation was protective when combined with pathogenic mutations and significantly reduced the Ca2+ leak of the RyR1 channel. Our study demonstrated that prime editing can efficiently introduce the protective S2843A mutation. In vitro experiments using the RNA electroporation of the prime editing components in human myoblasts achieved a 31% introduction of this mutation. This article lays the foundation for a new therapeutic approach for RyR1-RM, where a unique once-in-a-lifetime prime editing treatment could potentially be universally applied to all patients with a leaky RyR1 channel
Decentralized federated learning using validation loss for model sharing in crop disease classification
No full textInternational audienc
Extraction d'Informations dans les réseaux sociaux par apprentissage de représentations textes/images : application à la gestion des catastrophes naturelles
No full textWith the increasing use of social media, platforms like Twitter have become pivotal for real-time crisis information. Posts on these platforms serve as critical data sources for situational awareness and emergency response. To efficiently exploit this information, previous works have focused on developing automated methods to filter and categorize social media posts, primarily processing either text or images (unimodal data). The present work aims to integrate both images and text from tweets by fusing information from both modalities. Firstly, this thesis explores multiple unimodal encoders with various fusion techniques, along with a deep analysis of multimodal crisis tweets, highlighting the uniqueness of multimodal tweets in terms of modality relations. Secondly, it introduces a caption-based method that translates images into a text-compatible representation space through captioning, allowing for more efficient fusion and thus achieving better and more robust performance. This method also supports hybrid usage, enabling the model to function in both unimodal and multimodal settings without compromising performance. Our experiments revealed the presence of a well-known problem in multimodal learning, namely modality imbalance, where one modality dominates the learning process. We tackle this issue with a data-driven method that adapts to the learning dynamics and allows the dominated modality to catch up. In addition to methodological advancements, this thesis introduces M-CATNAT, a French multimodal dataset that fills a linguistic and multimodal gap in crisis-related resources. M-CATNAT provides detailed annotations for text, images, and combined content, offering a nuanced foundation for French multimodal crisis-related tweet processing.L'essor des réseaux sociaux, comme Twitter, a rendu ces plateformes essentielles pour diffuser des informations en temps réel lors de crises. Des méthodes automatisées de filtrage et de catégorisation de ces publications, basées sur le texte ou les images, ont été développées pour exploiter ces données. Ce travail propose d'intégrer simultanément texte et images des tweets, en fusionnant ces deux modalités. Dans un premier temps, cette thèse explore l'utilisation de plusieurs encodeurs unimodaux et techniques de fusion, tout en procédant à une analyse des tweets afin de mieux comprendre les relations entre les modalités visuelle et textuelle. Cette thèse introduit ensuite une méthode qui convertit les images dans un espace de représentation compatible avec le texte, rendant ainsi la fusion des deux modalités plus efficace et améliorant la robustesse et la performance du modèle. Cette approche est également hybride, permettant au modèle de traiter aussi bien des tweets unimodaux que multimodaux. Cette étude met en évidence un problème courant en apprentissage multimodal : le déséquilibre des modalités, où l'une d'elles domine le processus d'apprentissage. Une méthode d'adaptation a été proposée pour réguler cette dynamique et permettre une progression d'apprentissage plus équilibrée entre les modalités. En plus des avancées méthodologiques, cette thèse présente M-CATNAT, un jeu de données multimodal français qui comble un manque linguistique et multimodal dans les ressources liées aux crises. M-CATNAT offre des annotations détaillées pour le texte, les images et le contenu combiné, fournissant une base pour le traitement des tweets de crise multimodaux en français