HAL - La Rochelle Université, Archives Ouvertes
Not a member yet
19516 research outputs found
Sort by
Deep metric learning for end-to-end document classification
No full textInternational audienceDocument classification systems become more and more complex with the need to deal with new document formats or categories while obtaining a low error rate when classifying more and more documents. Such systems need to have important features including 1) the ability to eliminate ambiguity to improve precision or reduce the error rate, 2) the capability to detect and reject documents belonging to new categories or new variations. Previous studies often focused on closed datasets or solely on the problem of novelty detection, without evaluating the ability to reject ambiguous results after the novelty detection. In this paper, we propose an end-to-end document classification algorithm including both novelty and ambiguity rejection. The proposed algorithm utilizes deep metric learning to compact the knowledge space, and then uses the last hidden layer's features as input for an unsupervised KNN-based method for novelty and ambiguity rejection. Extensive experiments and analysis on private and public benchmark datasets demonstrate the effectiveness of our proposed algorithm. The algorithm provides the capability to handle new documents while effectively rejecting ambiguity to enhance the precision, recall of known categories, and coverage rate of the end-to-end document classification system
Evaluation in vitro de l’impact sur le microbiote du colon humain et la stéatose hépatique d’un complexe végétal conçu pour la prise en charge de la MASLD
No full textInternational audienc
Au Moyen Âge et par-delà: Passeurs, transferts et échanges en Méditerranée. Hommage à John Tolan
No full textInternational audiencePassé d’une rive à l’autre de l’Atlantique, d’Occident en Orient, John Tolan est depuis son arrivée à Nantes, à la fin du xxe siècle, l’un des historiens qui, au plan international, s’est le plus consacré aux transferts et aux échanges dans la Méditerranée médiévale et aux échos contemporains de cette histoire complexe. Pour lui rendre hommage, une vingtaine de chercheurs a souhaité mettre en relief et interroger des figures de passeurs. Au fil des pages, dans des contextes fort différents, se dessinent individus, groupes ou institutions et, à travers eux, les interrelations, toujours riches et éventuellement conflictuelles, qu’ils ont permises. Une temporalité ample, dépassant le Moyen Âge placé au cœur du livre, et un cadre géographique large s’imposaient, propres à souligner les changements de contexte que John Tolan, dans ses travaux, a si bien analysés en opérant à sa suite une approche pluridisciplinaire qui réunit l’histoire, l’étude des langues et l’ensemble des sciences sociales
Questionner la pérennité du désir de rivage dans le contexte des changements globaux
No full textInternational audienceThis paper looks at the durability of the desire for the shoreline in France in the context of global changes, whether environmental or societal. It explores how the attractiveness of coastal areas, initially a driving force for development, has become a source of tension and place struggles. After defining attractiveness as a socio-political construct and demonstrating the systemic dimension of the notion of global changes, the article examines recent residential and tourism dynamics, focusing on the case of Charente-Maritime (France). It also discusses the responses of public policies put in place to regulate this desire for the shore, such as legislation on short-term rentals and the surcharge on second homes, highlighting the local contrasts in their application. In conclusion, the article questions the longevity of the paradigm of coastal attractiveness in the face of the ‘polycrisis’ and highlights the enduring strength of the desire for the shoreline.Cet article interroge la pérennité du désir de rivage en France dans le contexte des changements globaux, qu'ils soient environnementaux ou sociétaux. Il explore comment l'attractivité des zones côtières, initialement un moteur de développement, est devenue une source de tensions et de "luttes des places". Après avoir défini l’attractivité comme une construction socio-politique et montré la dimension systémique de la notion de changements globaux, l'article examine les dynamiques résidentielles et touristiques récentes, en se concentrant sur le cas de la Charente-Maritime. Il traite aussi des réponses des politiques publiques mises en place pour réguler ce désir de rivage, telles que la législation sur les locations de courte durée et la surtaxe des résidences secondaires, soulignant les contrastes locaux dans leur application. En conclusion, l’article questionne la longévité du paradigme de l'attractivité littorale face à la « polycrise » et souligne la force persistante du désir de rivage
Long-term vertical deformations in the Ryukyu subduction zone, Japan
No full textInternational audienceIn subduction zones, the seismic cycle may induce plastic deformations that can accumulate through time leading to the formation of permanent reliefs. Studying long-term deformations can lead to a better understanding of the megathrust seismic cycle, its interface and its areas of past and potential future seismic ruptures. In the Ryukyu subduction zone, the tectonic behavior at the plate interface is not well understood. Its seismic coupling is debated since the recent discovery of both aseismic events and past earthquakes. We propose to investigate long-term behavior of the overriding Eurasian plate to better understand the geometry and behavior of the megathrust and better assess its seismic hazard. We studied marine terraces, that are exceptionnaly distributed along the archipelago. Marine terraces are formed by a combination of eustatic variations and tectonic vertical motions. Using very high resolution satellite imagery to compute Digital Surface Models, we obtained unique and new insights on the morphology of 12 islands between Kikai and Ishigaki islands. We identified and mapped a total of 13 Pleistocene terraces. We measured their elevation and correlated them with glacio-eustatic curves, as well as previous absolute datation from literature in order to establish their chronology. An average uplift has been estimated for each studied island, and constrained using reef growth modelling.Our study hilights the main role of the megathrust in shapping the long-term topography on the overriding plate. Variations in the uplift rates reveal the spatial and vertical segmentation of the megathrust. Finally, we explored) the link between long-term uplift and short to very short-term vertical deformations of the Ryukyu subduction zone to discuss its behavior over multiple time scales
Multidimensional integration using machine learning and Monte Carlo methods for acoustic predictions
No full textInternational audienceTo predict underwater noise radiated by a ship, various numerical methods are available. In underwater acoustics, the most effective prediction methods consist in solving an acoustic analogy using an integral formulation. In this study, we propose a machine learning surrogate-based method, combined with Monte Carlo integration, to efficiently estimate volume integrals that arise in acoustic analogies. We use three machine learning surrogate models: multi-layer perceptrons, Gaussian processes and gradient-boosted decision trees. For each model, a theoretical background is presented. We conduct numerical experiments to compare the state-of-the-art classical Monte Carlo quadrature method with our new machine learning based method. We first apply our method to simple canonical functions, which are analytically integrable, to evaluate the accuracy of our method. We then use a multi-layer perceptron-based surrogate model to approximate a fabricated function that mimics the characteristics of noise sources found in acoustic prediction models, such as those related to turbulent flows near geometrical singularities. Numerical experiments demonstrate that the proposed machine learning-based approach achieves performance levels comparable to state-of-the-art Monte Carlo quadrature methods, demonstrating the potential of ML techniques in this domain
