HAL Portal UBS (Université Bretagne Sud)
Not a member yet
14201 research outputs found
Sort by
Training LayoutLM from Scratch for Efficient Named-Entity Recognition in the Insurance Domain
No full textInternational audienceGeneric pre-trained neural networks may struggle to produce good results in specialized domains like finance and insurance. This is due to a domain mismatch between training data and downstream tasks, as in-domain data are often scarce due to privacy constraints. In this work, we compare different pre-training strategies for LAYOUTLM. We show that using domain-relevant documents improves results on a named-entity recognition (NER) problem using a novel dataset of anonymized insurance-related financial documents called PAYSLIPS. Moreover, we show that we can achieve competitive results using a smaller and faster model
Visualiser les proximités à l’aide de l’intelligence artificielle générative : le cas des nomades numériques à Bali, Indonésie
No full textInternational audienceThe paper examines proximity, both physical and digital, within Bali’s digital nomad landscape. Despite the allure of business prospects, unregulated digital nomadism poses challenges. Using AI, an image illustrates physical closeness but underscores social isolation, lacking community bonds. The beach, typically a social hub, appears solitary, reflecting a future depicted by the AI—an interplay of prophecy and uncertainty. This highlights the significance of visual representations in grasping the nuances of proximity, both physical and digital.Cet article propose de s'emparer de l'outil de génération d'image par intelligence artificielle en vue de mettre en perspective les proximités géographiques et organisées, en l'espèce celles des nomades numériques. L'idée générale consiste à analyser comment à partir d'un simple texte, ou prompt, l'on peut offrir une meilleure représentation visuelle et concrète de ces concepts, les rendre utiles et leur donner du sens. L’exemple des nomades numériques nous semble dans ce cadre pertinent car l’une des caractéristiques de ces individus est qu’ils souffrent de solitude (Miguel et al., 2023), forme de distance sociale inhérente à leur forte mobilité. Paradoxalement, ils ont tendance à se rassembler en communauté dans les mêmes destinations comme l’île de Bali en Indonésie (Thompson, 2019)
Vibrio harveyi uses both type III secretion system and quorum sensing for the colonization of the European abalone
No full textInternational audienceThe vibriosis of the European abalone, Haliotis tuberculata, is characterized by the rapidity of the infection by the pathogen Vibrio harveyi ORM4, leading to death of animals only after two days. The lethality of the pathogen is linked to the production of the type III secretion system (T3SS) and to genes regulated by quorum sensing (QS). The aim of this study was to investigate the colonization of the European abalone by both virulent and avirulent V. harveyi strains, as well as the involvement of T3SS and QS during infection. Our results emphasize the importance of gills for the bacterial establishment as the bacterial concentration of the avirulent V. harveyi strain significantly decreased from 189.3 ± 98.6 CFU/mg to 0.8 ± 0.5 CFU/mg between 24 and 48 h post-infection (hpi). In opposition, the pathogen V. harveyi ORM4 was able to maintain itself on the gills, with a concentration of 461.9 ± CFU/mg at 48 hpi, which was allowed by the production of T3SS and a functional QS. Following the infection cycle of V. harveyi ORM4 inside H. tuberculata, we also demonstrated that QS is essential for the ability of V. harveyi ORM4 to colonize the abalone hemolymph and to maintain in it. In response to the presence of V. harveyi, an increase in reactive oxygen species production was recorded, while the phagocytosis activity remained unchanged. We also highlighted the involvement of both QS and T3SS to escape the immune system activity, and that an overproduction of T3SS induced hemocyte mortality. This study provides the evidence that both T3SS and QS are essential for the establishment of V. harveyi ORM4 inside the European abalone
Chaotic behaviour in billiards
No full textInternational audienceAfter recalling classical probabilistic limit theorems for sums of independent identically distributed random variables, we present analogous results in a dynamical context. Motivated by examples coming from statistical mechanics, we are mostly interested in the Sinai billiard and in the Z^2-periodic Lorentz gas. We will also consider the Bunimovich stadium billiard and dispersive billiards with cusps. All these billiards are chaotic, with different behaviours.Additional explanations are given in four independent appendices
Direct evidence of predation on endangered sea lamprey in estuarine tidal area
No full textInternational audienceThe sea lamprey Petromyzon marinus is a anadromous species that has undergone a dramatic decline in many European rivers over recent decades, observed for instance as substantial decrease in the number of returning spawners (i.e., the reproductive stock), observed in most major river systems. Consequently, in Western Europe, its conservation status varies from Near Threatened (NT) in Ireland, to Least concern (LC) in Great Britain, Vulnerable (VU) in Spain, Endangered (EN) in France, and Critically Endangered (CR) in Italy, according to national IUCN assessments. This decline is attributed to multiple pressures, including habitat degradation, obstacles to migration, climate change, and overfishing. However, the potential impact of predation on this endangered species remains poorly documented, especially in estuarine areas. This study aimed to evaluate predation on sea lampreys during their upstream migration in the tidal oligohaline zone of the Loire estuary, using acoustic telemetry over two years. Fish were captured, tagged with acoustic tags (with or without digestion sensors), and released into the river. The predation rates (certified by digestion tag and estimated with classical tags) ranged from 8.8 to 20 %, suggesting that predation could represent a significant, yet overlooked, threat in estuarine zones. Based on indirect evidence and catfish fishing data, the European catfish Silurus glanis — an invasive and opportunistic predator increasingly abundant in the Loire system — appears to be the most likely predator responsible for these events. This study provides the first evidence suggesting sea lamprey predation by European catfish in estuarine tidal zones. It highlights the need to consider the impact of invasive species on migratory fish populations, not only in upstream sections of rivers but also from the very first stages of their freshwater migration
Behavioral analytics for optimized self-scheduling in sustainable local multi-carrier energy systems: A prospect theory approach
No full textInternational audiencePower coefficients of the CHP unit (kW). SoC E min/maxLower and upper bounds for the SoC of electrical energy storage (kWh). SoC H min/max Lower and upper bounds for the SoC of thermal energy storage (kWh). Slop Pos/Neg l Slope of the value function in the positive (gain) and negative (loss) parts. Variables u Pos/Neg s,t Binary variables of the positive and negative parts of the value function.</div
Characterisation of Friction Between Glass Tows: Comparison of Experiments and Finite Element Simulations Using an ALE Formulation
No full textInternational audienceThe phenomenon of inter-yarn friction plays a crucial role in determining the behavior of woven materials. However, understanding and modeling this phenomenon remain complex due to the numerous parameters involved-such as normal force, velocity, angles, and sizing agents. To better analyze the influence of these various parameters on friction, a friction test bench without force sensors was developed, along with an associated finite element model. Motor control is used to vary carriage speed. An automatic image analysis method is used to measure the slippage between fiber tows. The model relies on an Arbitrary Lagrangian Eulerian (ALE) formulation, which provides simplified handling of inter-yarn contacts and slippage, while offering enhanced computational speed and robustness compared to a model based on a Lagrangian formulation. By comparing simulation results with experimental curves, a contact law was formulated, and its parameters were identified
Material Decomposition in Photon-Counting Computed Tomography with Diffusion Models: Comparative Study and Hybridization with Variational Regularizers
No full textInternational audiencePhoton-counting computed tomography (PCCT) enables spectral imaging and material decomposition (MD) but often suffers from low signal-to-noise ratios due to constraints like low photon counts and sparse-view settings. Traditional variational methods depend heavily on handcrafted regularizers, while AI-based approaches, particularly convolutional neural networks (CNNs), have become state-of-the-art. More recently, diffusion models (DMs) have gained prominence in generative modeling by learning distribution functions, which can serve as priors for inverse problems. This work explores DMs as regularizers for MD tasks in PCCT using diffusion posterior sampling (DPS). We evaluate three DPS-based approaches: image-domain two-step DPS (im-TDPS), projection-domain two-step DPS (proj-TDPS), and one-step DPS (ODPS). Im-TDPS first samples spectral images via DPS, then performs image-based MD; proj-TDPS applies projection-based MD before sampling material images via DPS; ODPS directly samples material images from measurement data. Results show ODPS outperforms im-TDPS and proj-TDPS, producing sharper, noise-free, and crosstalk-free images. Additionally, we propose a hybrid ODPS method integrating DM priors with variational regularizers to handle materials absent from the training dataset. This approach enhances material reconstruction quality over standard variational methods
FBI Transform in Gevrey Classes and Anosov Flows
No full textInternational audienceAn analytic FBI transform is built on compact manifolds without boundary, that satisfies all the expected properties. It enables the study of microlocal analytic regularity on such manifolds. This tool is then used to study the Ruelle spectrum of Anosov flows with Gevrey coefficients. In particular, a Weierstrass factorization of their dynamical determinant is derived
