HAL Portal UO (Université d'Orléans)
Not a member yet
48841 research outputs found
Sort by
GM-CSF drives IL-6 production by macrophages in polymyalgia rheumatica
No full textInternational audienceInsight into the immunopathology of polymyalgia rheumatica (PMR) is scarce and mainly derived from peripheral blood studies. The limited data available point towards macrophages as potential key players in PMR. This study aimed to identify the factors driving proinflammatory macrophage development and their functions in the immunopathology of PMR. Methods: Monocyte phenotypes were investigated by flow cytometry in peripheral blood (PMR, n = 22; healthy controls, n = 20) and paired subacromial-subdeltoid (SASD) bursal fluid (PMR, n = 9). Macrophages in SASD bursa were characterised by immunohistochemistry and immunofluorescence (PMR, n = 12; controls undergoing shoulder replacement surgery, n = 10). The</div
The non-Abelian Aharonov-Bohm-effect
No full textThe scattering of a nucleon beam around a cylinder containing a non-Abelian flux is studied.We confirm all the previsions of Wu and Yang. We consider the generalization to the gauge group SU (N ), and derive a classification scheme. Isospin precession is recovered also at the classical limit.</div
Features Leverage in Graph Models for Mineral Prospectivity Mapping
No full textInternational audienceMineral Prospectivity Mapping (MPM), the process of identifying areas with high potential for mineral deposits, can be divided into two main categories: knowledge-driven and data-driven. Knowledgedriven techniques rely on expert opinion on geological data, while data-driven techniques employ ML models to predict the probabilities of mineral occurrences based on known geological datasets. Recently, with the advancement of machine learning (ML) methods, data-driven MPM has gained significant improvements. Notably, graph-based approaches overcome the disadvantages of previously used approaches (pixel-based, image-based) and have demonstrated better performances. However, the graph construction in current methods is based solely on spatial distances between pixels, regardless of their geological attributes. In this paper, we introduce a novel graph construction approach that combines spatial distances with other distances obtained from feature mining. Our experiments show that this combination outperforms existing graphs, and can be considered as a promising approach to integrate feature mining into data-driven models in MPM
Analyses destructive et non-destructive des dommages directs produits par un choc foudre dans des stratifiés composites aéronautiques protégés et peints
No full textInternational audienceThe use of CFRP composite increased significantly since the last 40 years for aircraft structure. Unfortunately, such structures are subjected to significant damages if struck by lightning compared to metallic structure. This is mainly due to the low conductivity of this material, which cannot evacuate the current without high Joule heating. Lightning strike-induced damage in a composite laminate is composed of in-depth delamination, fibre breakage, and resin deterioration due to the surface explosion and the core current flow linked to interaction of the arc with the surface. But very rare previous studies dedicated to the analysis of damage as a direct effect of lightning have considered the spurious effect of the paint that always covers real aeronautic structures neither on the thermal nor the mechanical loads that are the root cause of these damages. We present in this paper a coupled non-destructive and destructive damage analysis to support the proposition of damage scenarios depending on the presence and thickness of the paint. The mechanical and thermal sources contribution in the global loading on the core damage is discussed, which confirms previous studies’ analysis and modelling and is in accordance with existing works in the literature
Apprentissage continu économe en mémoire avec contraste d'effondrement neuronal
No full textInternational audienceContrastive learning has significantly improved representation quality, enhancing knowledge transfer across tasks in continual learning (CL). However, catastrophic forgetting remains a key challenge, as contrastive based methods primarily focus on "soft relationships" or "softness" between samples, which shift with changing data distributions and lead to representation overlap across tasks. Recently, the newly identified Neural Collapse phenomenon has shown promise in CL by focusing on "hard relationships" or "hardness" between samples and fixed prototypes. However, this approach overlooks "softness", crucial for capturing intra-class variability, and this rigid focus can also pull old class representations toward current ones, increasing forgetting. Building on these insights, we propose Focal Neural Collapse Contrastive (FNC2 ), a novel representation learning loss that effectively balances both soft and hard relationships. Additionally, we introduce the Hardness-Softness Distillation (HSD) loss to progressively preserve the knowledge gained from these relationships across tasks. Our method outperforms state-of-the-art approaches, particularly in minimizing memory reliance. Remarkably, even without the use of memory, our approach rivals rehearsal-based methods, offering a compelling solution for data privacy concerns.L’apprentissage contrastif a considérablement amélioré la qualité des représentations, facilitant le transfert de connaissances entre les tâches dans l’apprentissage continu (CL). Cependant, l’oubli catastrophique reste un défi majeur, car les méthodes basées sur le contraste se concentrent principalement sur les « relations douces » ou la « douceur » entre les échantillons, qui évoluent avec les changements dans la distribution des données et entraînent un chevauchement des représentations entre les tâches. Récemment, le phénomène nouvellement identifié de l’Effondrement Neuronal (Neural Collapse) a montré un potentiel en CL en se focalisant sur les « relations dures » ou la « dureté » entre les échantillons et les prototypes fixes. Cependant, cette approche néglige la « douceur », essentielle pour capter la variabilité intra-classe, et cette focalisation rigide peut également attirer les représentations des anciennes classes vers celles actuelles, augmentant ainsi l’oubli. En nous appuyant sur ces observations, nous proposons Focal Neural Collapse Contrastive (FNC2), une nouvelle fonction de perte d’apprentissage de représentation qui équilibre efficacement à la fois les relations douces et dures. De plus, nous introduisons la perte Hardness-Softness Distillation (HSD) pour préserver progressivement les connaissances acquises de ces relations à travers les tâches. Notre méthode surpasse les approches de pointe, notamment en minimisant la dépendance à la mémoire. Remarquablement, même sans l’utilisation de mémoire, notre approche rivalise avec les méthodes basées sur la répétition, offrant une solution convaincante aux préoccupations liées à la confidentialité des données
The Convex Set Forming Game
No full textInternational audienceIn 1984, Frank Harary introduced the first graph convexity game, focused on the geodesic convexity. A set of vertices of a graph is convex if every shortest path between two vertices of is also included in . We introduce the Convex Set Forming Game \CFG: two players alternately select vertices in such a way that the set of selected vertices is always a convex set. In the normal (resp., mis\`ere) variant, the last player to be able to select a vertex wins (resp., loses). We also define a new graph invariant \gcon(G) as the largest integer such that the first player has a strategy ensuring that, at the end of the game, at least vertices of the graph have been selected. We first show that the problems of deciding the outcome (does the first player win?) of the game in both variants (normal and mis\`ere), as well as the problem of deciding whether \gcon(G)\geq k, are PSPACE-complete. As a by-product, we prove that the optimization variant of the classical \textsc{Kayles} game is PSPACE-complete. Then, we focus on convexable graphs, i.e., -node graphs for which \gcon(G)=n. For this purpose, we say that a set in a graph admits a Convex Elimination Ordering (CEO) if is convex for every . We show that the class of graphs whose vertex-set admits a CEO coincides with the chordal graphs and that this class strictly contains the convexable graphs. Moreover, every graph which is Ptolemaic (distance-hereditary chordal) or unit interval is convexable.Finally, we give a polynomial-time algorithm for computing a largest set admitting a CEO in outerplanar graphs, which gives upper bounds on \gcon(G) in outerplanar graphs
Tax Revenue Mobilization in Developing Countries: Do We Need More Political Veto Players?
No full textInternational audienc
“Il faut voir le mal où il est !” Le rôle des valeurs dans la résistance à la digitalisation d’un business model : les éditeurs français face à l’édition numérique (2000-2020)
No full textInternational audienc
Le bilan carbone : outil de conformité, de connaissance ou de passage à l’action ? Une étude exploratoire dans les établissements de santé
No full textInternational audienceMalgré le consensus autour de la nécessité de réduire nos émissions de gaz à effet de serre (GES), l’obligation, légale en France depuis 2010 pour les grandes entreprises et organisations publiques, de réaliser un bilan de ces émissions, est loin d’être remplie par toutes, sans que les causes de cette faible conformité soient entièrement claires. Le système de santé est concerné de façon particulière par les enjeux climatiques : d’un côté, ses émissions représentent une part non négligeable des émissions totales du pays, de l’autre, son activité (la demande de soins) est très sensible au réchauffement climatique et une augmentation de cette demande doit donc être anticipée. Face à ce double constat, la décarbonation des organisations de santé apparaît revêtir un enjeu particulier. Pourtant, ces dernières respectent aussi peu l’obligation de réalisation d’un bilan des émissions de GES que les autres organisations.Notre étude exploratoire auprès de cadres gestionnaires et dirigeants d’organisations de santé vise à appréhender la façon dont le bilan carbone est perçu et expérimenté. Elle montre que celui-ci, en cohérence avec la littérature existante, apparaît relativement peu maîtrisé et complexe à mettre en œuvre. Mais elle met aussi en avant les ambiguïtés d’une obligation réglementaire paradoxalement peu incitative, voire « désincitative ». Elle souligne enfin la dimension équivoque de l’outil et la façon dont il peut pâtir du fait d’être perçu comme un outil de connaissance bien plus que comme un outil opérationnel de passage à l’action
Synthesis and Reactivity of 1,4‐Ethano‐1,5‐Naphthyridine Derivatives Using Microwave Activation or Flow Chemistry
No full textInternational audienceThe design of some novel disubstituted 3,4‐dihydro‐2 H‐1,4‐ethano‐1,5‐naphthyridine derivatives is reported under classical and flow methodologies. The series are developed from quinuclidinone, which afford versatile platforms bearing one lactam function in position C‐2 that is then used to create C─N bond using the Chan–Lam coupling reaction or in situ C─O bond activation via palladium‐catalyzed cross‐coupling reactions. The conditions are optimized and a wide range of boronic acids are used to determine the scope and limitations of each method. To complete this study, a flow Suzuki–Miyaura process is established to afford polyfunctionalized 1,4‐ethano‐1,5‐naphthyridine derivatives in high yields with a very efficient process (10 min)