Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Un environnement de développement d'applications sur un processeur à beaucoup de cœurs parallélisant
No full textDigital objects of the future (domestic robots, autonomous vehicles, automatic spacecraft, ...) will need both computing power and safety. The Little Big Processor (LBP) is suitable for this challenge: it has an innovative approach to parallelism which offers the advantages of computing power while guaranteeing a certain determinism of execution. This execution determinism brings a level of operational safety essential in most devices interacting with the world and humans. In the present thesis we created a development environment for LBP, with a compiler, a loader and a debugger. These tools are classic but in this case, they will have to be adapted to the implementation of parallelized OpenMP applications for LBP. Following the creation of the development environment, we defined a deterministic parallel model for embedded bare-metal. This model has been evaluated on a embedded bare-metal platform, and this allowed us to confirm that it is possible to have a deterministic parallel execution which keeps the performance speedups from parallelism.Les objets numériques du futur (robots domestiques, véhicules autonomes, engins spatiaux automatiques,...) auront besoin à la fois de puissance de calcul et de sûreté. Le Little Big Processor (LBP) est adapté à ce défi : il a une approche novatrice du parallélisme qui offre l'avantage de la puissance en garantissant un certain déterminisme de l'exécution. Ce déterminisme d'exécution donne une sûreté de fonctionnement indispensable dans la plupart des dispositifs interagissant avec le monde et l'humain. Dans cette thèse, nous avons réalisé un environnement de développement pour LBP, avec un compilateur, un « bootloader » et un débogueur. Ces outils sont classiques, mais en l'occurrence, ils devront être adaptés à la mise en oeuvre d'applications parallélisées avec OpenMP pour LBP. Suite à la réalisation de l'environnement de développement, nous avons défini un modèle de parallélisme déterministe pour de l'embarqué « bareme-tal ». Ce modèle a été évalué sur une plateforme embarquée « baremetal » et nous a permis de confirmer qu'il était possible d'avoir une exécution parallèle déterministe qui conserve les gains en performance du parallélisme
Interplay of Human Factors and Secure Architecture Design using Model-Driven Engineering
No full textInternational audienc
Some considerations to design an autonomous buoy system to enumerate pelagic sharks at Fish Aggregating Devices
No full textFish Aggregating Devices (FADs) are floating objects used by fishers to facilitate their catches. The majority of the global industrial tropical tuna purse-seine catches currently occurs at FADs. One of the main adverse ecological impacts of FAD fisheries is the bycatch of vulnerable species such as pelagic sharks. Detecting the presence of sharks at FADs remotely, constitutes a key step to reduce their catches.In this paper, we explain how an image-based shark detection module can be embedded on an autonomous buoy associated to a FAD. We discuss the general design, the hardware and software selection, the implementation of the detection algorithm based on Deep-Learning techniques and we detail the implementation inside the buoy. Some experiments allowed us to discuss energy consumption. This system could be integrated with other new technologies to mitigate bycatch in industrial tropical tuna purse-seine fisheries.</p
Discovering a Representative Set of Link Keys in RDF Datasets
No full textabbas2024aInternational audienceA link key is based on a set of property pairs and can be used to identify pairs of individuals representing the same real-world entity in two different RDF datasets. Various algorithms are aimed at discovering link keys which usually output a large number of candidates, making link key selection and validation a challenging task. In this paper, we propose an approach combining Formal Concept Analysis (FCA) for discovering link key candidates and building a link key lattice, and then hierarchical clustering over a given set of candidates for building a representative set of link keys. Such a link key set should minimize the number of candidates to be validated while preserving a maximal number of links between individuals. The paper also provides a series of experiments which are performed over different RDF datasets, showing the effectiveness of the approach and the ability of hierarchical clustering to return a concise and meaningful set of candidates while preserving the ordinal structure of the link key lattice
Gestion du déséquilibre de classe au sein du classifieur de séries d'images astronomiques ConvEntion
No full textProceedings en ligne https://projet.liris.cnrs.fr/coresa/National audienceEn classification de séquences d’images astronomiques, l’approche état-de-l’art (ConvEntion) repose sur l’utilisation d’une architecture basé sur une convolution 3D et un transformer. Cette architecture ConvEntion ne gère pas parfaitement le déséquilibre de classes. Dans cet article, nous proposons d’améliorer cela en s’appuyant sur les méthodologies auto-supervisées. Nous réduisons la variance intra-classe en passant par une architecture à deux branches. Chacune des deux branches traite une version augmentée de la donnée d’entrée. Dans le même temps, nous conservons la contrainte de classification ce qui nous permet de faire l’apprentissage sur un petit ensemble de données labellisées. Les résultats de notre modèle ICT-ConvEntion nous ont permis d’obtenir une amélioration de l’exactitude (accuracy) de 2.3% et du score F1 de 4.7% sur la base SDSS Supernova Survey
Beyond Facts 2024 - 4th International Workshop on Computational Methods for Online Discourse Analysis: Companion Proceedings of the ACM Web Conference 2024, May 13–17, 2024, Singapore, Singapore
No full textInternational audienceExpressing opinions and interacting with others on the Web has led to the production of an abundance of online discourse data, such as claims and viewpoints on controversial topics, their sources and contexts (events, entities). This data constitutes a valuable source of insights for studies into misinformation spread, bias reinforcement, echo chambers or political agenda setting. Computational methods, mostly from the field of NLP, have emerged that tackle a wide range of tasks in this context, including argument and opinion mining, claim detection, checkworthiness detection, stance detection or fact verification. However, computational models require robust definitions of classes and concepts under investigation. Thus, these computational tasks require a strong interdisciplinary and epistemological foundation, specifically with respect to the underlying definitions of key concepts such as claims, arguments, stances, check-worthiness or veracity. This requires a highly interdisciplinary approach combining expertise from fields such as communication studies, computational linguistics and computer science. As opposed to facts, claims are inherently more complex. Their interpretation strongly depends on the context and a variety of intentional or unintended meanings, where terminology and conceptual understandings strongly diverge across communities. From a computational perspective, in order to address this complexity, the synergy of multiple approaches, coming both from symbolic (knowledge representation) and statistical AI seem to be promising to tackle such challenges. This workshop aims at strengthening the relations between these communities, providing a forum for shared works on the modeling, extraction and analysis of discourse on the Web. It will address the need for a shared understanding and structured knowledge about discourse data in order to enable machine-interpretation, discoverability and reuse, in support of scientific or journalistic studies into the analysis of societal debates on the Web. Beyond research into information and knowledge extraction, data consolidation and modeling for knowledge graphs building, the workshop targets communities focusing on the analysis of online discourse, relying on methods from machine learning, natural language processing, large language models and Web data mining
Phylogenetic Inference: Distance-Based Methods
No full textInternational audienceWhen evolutionary distances between pairs of taxa can be estimated, they can be used to rapidly infer a phylogenetic tree. This chapter explains how distances should be defined and estimated, and then focus on the task of constructing a phylogenetic tree that corresponds to those distances. It discusses the properties that distances must satisfy in principle, and the classic approaches to tree inference, least squares and minimum evolution. The chapter focuses on the most commonly known methods, in particular neighbor joining and a wide range of algorithms inspired therefrom. It presents the major ideas underlying distance-based methodology. The study of tree distances is an old subject and is one of the foundations of mathematical phylogenetics. The component of a distance-based approach to tree inference is distance estimation. FastTree is faster and requires less memory than the distance-based methods, and it can easily handle datasets containing hundreds of thousands of sequences
A Dual-Domain Diffusion Model for Sparse-View CT Reconstruction
No full textInternational audienceA new deep-learning approach, dual-domain diffusion model (DDDM), is proposed for sparse-view CT reconstruction, which is composed of a sinogram upgrading module (SUM) and an image refining module (IRM) connected in series. In the sinogram domain, a novel degrading and upgrading framework is defined, in which SUM is trained to upgrade sparse-view sinograms step by step to reverse the degradation process of CT images caused by successive down-sampling of scanning views. In the image domain, IRM adopts an improved denoising diffusion framework to further reduce remaining artifacts and restore image details, where a skip connection from the original sparse-view sinogram is introduced to constrain the generation of details. Our DDDM shows significant improvement over deep-learning baseline models in both classical similarity metrics and perceptual loss, and has good generalization to untrained organs. We release our code at: https://github.com/YC-Markus/code-for-DDDM
Comparing angles in Euclid's Elements
No full textThe author is grateful to Andrei Rodin, AlexanderShtern, Andrei Shchetnikov and the participants of the Kolmogorov complexityseminar for their comments and explanations.The exposition in Euclid's Elements contains an obvious gap (seemingly unnoticed by most commentators): he often compares not just angles, but groups of angles, and at the same time he avoids summing angles (and considering angles greater than π), and does not say what such a comparison of groups could mean. We discuss the problem and suggest a possible interpretation that could make Euclid's exposition consistent