Archive ouverte HAL de CY Cergy Paris Université
Not a member yet
38365 research outputs found
Sort by
Unified Anomaly Detection via Multi-Scale Contrasted Memory
No full textInternational audienceDeep anomaly detection aims to provide robust and efficient classifiers for zero-shot (unsupervised, UNS) and few-shot (imbalanced supervised, IMS) settings. However, current models still struggle on edge-case normal samples and are often unable to keep high performance over different scales of anomalies. Additionally, there is a lack of a unified framework that efficiently addresses both UNS and IMS settings. To address these limitations, we present a novel two-stage method which leverages multi-scale normal prototypes during training to compute an anomaly deviation score. First, we employ a novel memory-augmented contrastive learning to jointly learn representations and memory modules across multiple scales. This allows us to effectively capture subtle features of normal data while adapting to varying levels of anomaly complexity. Then, we train an efficient anomaly distance-based detector that computes spatial deviation maps between the learned prototypes and incoming observations. Our model outperforms the SoTA on a wide range of anomalies, including object, style, and local anomalies, as well as industrial inspection and face anti-spoofing, while being on par with SoTa out-of-distribution detectors. Notably, it stands as the first model capable of maintaining exceptional performance across both settings
Political attitudes differ but share a common low-dimensional structure across social media and survey data
No full textDoes polarization online reflect the state of polarization in society? We study ideological positions and attitudes on several issues in France, a country with documented issue nonalignment. We compare distributions on X/Twitter with a nationally representative sample, focusing on two key properties: ideological polarization and issue alignment. Despite significant issue-wise divergences, positions of both the X population and the nationally representative sample present a similar bi-dimensional structure along two dominant bundles of aligned issues: a Left-Right divide, and a Global-Local divide. We then study how our results vary when accounting for key structural parameters of the online public sphere: activity, popularity, and visibility. We find that the dimensionality of attitude distributions shrinks as ideological polarization increases when selecting more active users. The divergence between political attitudes on social media and in survey data is greatly mediated by the combination of activity and popularity of social media users: users benefiting from the most exposure are also the most representative of the general public. Together, our results shed light on the structural similarities and differences between political attitudes from social media users and the general public
L'abus de dépendance dix ans après la réforme : un bilan jurisprudentiel contradictoire
No full textInternational audienc
High-resolution evolution of sediment yields in the Valencia and Menorca basins during the Neogene: Relation with climate, tectonics and Ebro Basin opening
No full textInternational audienceBasin-wide stratigraphic correlation along the Valencia and Menorca basins enabled to provide a high-resolution quantitative sedimentary history for the last 23 Ma. The identified seismic markers were dated, thanks to well calibration, thickness maps of successive units were built, and associated sediment fluxes were calculated. This allowed to highlight the evolution of sediment routing through space and time over the Neogene and to discuss the local vs global triggers impacting this sediment yield, such as tectonics and climate, in particular in relation with the Ebro Basin opening to the Mediterranean Sea. Results of this study point to the successive trends: (i) a three-fold increase in sedimentary yields at the Burdigalian-Langhian transition in the Valencia Basin; (ii) an almost steady state during the Serravallian-Tortonian intervals; (iii) a nine-fold increase in sediment -flux during the Messinian Crisis, especially after the main intra-Messinian sea-level fall; (iv) a 2.5-fold sediment flux increase between the Miocene and the Pliocene and, finally, (v) a sedimentary flux increase by 1.5 at 0.45 Ma. The Mid-Miocene trend is likely linked to enhanced erosional processes during the ultimate phase of the Balearic orogeny and initiation of the subsidence in the adjacent area. The Climatic optimum at that time led also to the development of in-situ carbonate systems along the Ebro platform. The nine-fold increase in sediment flux during the Messinian Salinity Crisis reflect the huge erosional processes and sediment transfer from upstream into the basins; the fact that no increase is observed before the Late Messinian favours the hypothesis that the Ebro basin remains as an endorheic system until its catastrophic breaching that seems to occur during the crisis (with the uncertainty about the delay of the answer within the system). The 2.5-fold sediment increase between the Miocene and the Pliocene is in accordance with the observed increase on the Gulf of Lion margin, and on a worldwide scale. This increase is likely linked to the intense erosion of the Ebro Basin after the complete establishment of the modern Ebro River at the end of the Messinian Salinity Crisis and/or to the establishment of a more humid climate. The final increase, associated with an increase in sedimentary transfers from the Valencia Basin to the Menorca Basin through the Valencia channel system, mainly reflects the effects of the longer and larger eustatic cycles over the last 500 ka. These findings demonstrate that sediment yields are exceptional recorders of multi-timescale geomorphologic processes, even if they cannot always serve as precise chronometer. They also support a three-phases interpretation for the Ebro system: (1) a Mid-Miocene interval characterized by enhanced erosion in the Balearic orogen and subsidence initiation in the adjacent basins, without evidence of connection between the Ebro basin and the Mediterranean (even if Ebro tributaries are testified); (2) a catastrophic breaching during the Messinian Salinity Crisis, and (3) Pliocene and Quaternary intervals testifying of a fully Ebro River system connected to the Mediterranean
Euclid preparation. Using mock Low Surface Brightness dwarf galaxies to probe Wide Survey detection capabilities
No full textInternational audienceLocal Universe dwarf galaxies are both cosmological and mass assembly probes. Deep surveys have enabled the study of these objects down to the low surface brightness (LSB) regime. In this paper, we estimate Euclid's dwarf detection capabilities as well as limits of its MERge processing function (MER pipeline), responsible for producing the stacked mosaics and final catalogues. To do this, we inject mock dwarf galaxies in a real Euclid Wide Survey (EWS) field in the VIS band and compare the input catalogue to the final MER catalogue. The mock dwarf galaxies are generated with simple Sérsic models and structural parameters extracted from observed dwarf galaxy property catalogues. To characterize the detected dwarfs, we use the mean surface brightness inside the effective radius SBe (in mag arcsec-2). The final MER catalogues achieve completenesses of 91 % for SBe in [21, 24], and 54 % for SBe in [24, 28]. These numbers do not take into account possible contaminants, including confusion with background galaxies at the location of the dwarfs. After taking into account those effects, they become respectively 86 % and 38 %. The MER pipeline performs a final local background subtraction with small mesh size, leading to a flux loss for galaxies with Re > 10". By using the final MER mosaics and reinjecting this local background, we obtain an image in which we recover reliable photometric properties for objects under the arcminute scale. This background-reinjected product is thus suitable for the study of Local Universe dwarf galaxies. Euclid's data reduction pipeline serves as a test bed for other deep surveys, particularly regarding background subtraction methods, a key issue in LSB science
Targeted discovery of sesquiterpene indole alkaloids from Greenwayodendron suaveolens
No full textInternational audienceThroughout the past decades, annonaceous plants have been of particular interest to the natural product community because of their therapeutic value and their richness in isoquinoline alkaloids. Taking advantage from our laboratory historical collection of these compounds, a MS/MS database of 322 isoquinolines and other metabolites from Annonaceae was implemented and named IQAMDB. The present report describes the dereplication of known alkaloids from stem barks of Greenwayodendron suaveolens (Engl. & Diels) Verdc. leveraging IQAMDB-informed feature-based molecular networking further refined by in silico annotation and taxonomic weighting. This strategy annotated over 30 compounds and streamlined the isolation of three sesquiterpene indole alkaloids (SIA) (1–3). Structure elucidation and absolute configuration assignment by a combination of NMR, TDDFT-ECD or X-ray diffraction determined these compounds to be greenwaylactam D (1) and greenwaylactam E (2), previously undescribed Witkop-Winterfeldt oxidized diastereoisomers of the previously reported greenwaylactam A, and polyveodrine (3), that discloses an unprecedented relative configuration for a SIA. The isolated compounds were evaluated for their antibacterial and antifungal activities against Staphylococcus aureus, Pseudomonas aeruginosa, Candida albicans, Mycobacterium marinum as well as their antiviral activity against Zika virus. Polyveodrine exhibited moderate antimycobacterial activity against M. marinum, whereas greenwaylactam D (1) demonstrated moderate antiviral activity against Zika virus under non-cytotoxic concentrations.Graphical abstractThree new sesquiterpene indole alkaloids were targeted following a MS-based dereplicative logic and subsequently isolated from stem barks of Greenwayodendron suaveolens. The structure elucidation relied on a combination of spectroscopic techniques and computational methods. The isolated compounds were evaluated for their anti-infective activity
Deviation inequalities for contractive infinite memory processes
No full textInternational audienceIn this paper, we introduce a class of stochastic processes that encompasses many natural and widely used examples. A key feature of these processes is their infinite memory, which enables them to retain information from arbitrarily distant past states. Using the martingale decomposition method, we derive deviation and moment inequalities for separately Lipschitz functionals of such processes, under various moment conditions on certain dominating random variables. Our results extend those obtained for Markov chains by Dedecker and Fan [Stochastic Process. Appl., 2015], as well as recent results by Chazottes et al. [Ann. Appl. Probab., 2023] concerning specific infinite-memory models with sub-Gaussian concentration bounds. We also discuss an application to the stochastic gradient Langevin dynamics algorithm.</div
Introduction to arithmetic
No full textThis is an introduction to numbers, fractions, percentages and arithmetic. We first discuss what can be done with integers and their quotients, namely basic arithmetic, all sorts of counting results, and with a look into abstract algebra and quadratic residues too. We then upgrade our knowledge by introducing the real numbers, and exploring what can be done with them, in relation with number theory questions. Then we further upgrade our methods, by introducing and using the complex numbers. Finally, we provide an introduction to the zeta function, and the Riemann hypothesis
Lithospheric models supported by the Caribbean and Levant examples help rethink transpression at plate boundaries
No full textInternational audienceStrike-slip restraining bends, such as the Levant Fault, belonging to push-up systems and the Jamaican fault network, belonging to duplex systems, display a diversity of fault geometries and deformation patterns that reflect distinct modes of lithospheric-scale strain localization. To investigate the origin of this variability, we develop 3D numerical models of transpressional strike-slip systems using heterogeneous simple shear boundary conditions and thermally-dependent, non-linear rheology. Unlike classical analog or numerical models that impose velocity discontinuities, our approach allows spontaneous fault localization that naturally generates transpression. We systematically explore how the position and geometry of inherited weak zones influence fault development. We show that three distinct strike-slip systems emerge: (1) push-up systems with a single strike-slip fault and outward-propagating thrusts; (2) duplex systems with interacting parallel faults connected by P-shears; and (3) systems of non-interacting parallel faults. These results highlight spontaneous strike-slip localization and how initial heterogeneities control formation and evolution of long-term lithospheric deformation
