Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
A One-Health Platform for Antimicrobial Resistance Data Analytics
No full textInternational audienceAntimicrobial resistance (AMR) poses potentially critical health issues for human and animal populations in the near future. To meet this challenge, we need to adopt a "One Health" strategy, which involves studying and linking information from human and animal populations, as well as from the environment.In this demonstration, we present an early prototype of Promise platform, which we are developing for One Health data management and analytics, to enable experts from different fields to gain insights into AMR. It is designed to handle data from 25 academic networks and 42 partners. Our demonstration illustrate the capabilities of our methodology for analyzing these data. The user is freed from considerations related to data heterogeneity, as interoperability issues are managed by the platform. Additionally, each data provider will be able to stay within his/her own vocabulary, whatever the taxonomy used by other data providers
A deep-learning framework for enhancing habitat identification based on species composition
No full textInternational audienceAims The accurate classification of habitats is essential for effective biodiversity conservation. The goal of this study was to harness the potential of deep learning to advance habitat identification in Europe. We aimed to develop and evaluate models capable of assigning vegetation-plot records to the habitats of the European Nature Information System (EUNIS), a widely used reference framework for European habitat types.Location The framework was designed for use in Europe and adjacent areas (e.g., Anatolia, Caucasus).Methods We leveraged deep-learning techniques, such as transformers (i.e., models with attention components able to learn contextual relations between categorical and numerical features) that we trained using spatial k-fold cross-validation (CV) on vegetation plots sourced from the European Vegetation Archive (EVA), to show that they have great potential for classifying vegetation-plot records. We tested different network architectures, feature encodings, hyperparameter tuning and noise addition strategies to identify the optimal model. We used an independent test set from the National Plant Monitoring Scheme (NPMS) to evaluate its performance and compare its results against the traditional expert systems.ResultsExploration of the use of deep learning applied to species composition and plot-location criteria for habitat classification led to the development of a framework containing a wide range of models. Our selected algorithm, applied to European habitat types, significantly improved habitat classification accuracy, achieving a more than twofold improvement compared to the previous state-of-the-art (SOTA) method on an external data set, clearly outperforming expert systems. The framework is shared and maintained through a GitHub repository.Conclusions Our results demonstrate the potential benefits of the adoption of deep learning for improving the accuracy of vegetation classification. They highlight the importance of incorporating advanced technologies into habitat monitoring. These algorithms have shown to be better suited for habitat type prediction than expert systems. They push the accuracy score on a database containing hundreds of thousands of standardized presence/absence European surveys to 88.74%, as assessed by expert judgment. Finally, our results showcase that species dominance is a strong marker of ecosystems and that the exact cover abundance of the flora is not required to train neural networks with predictive performances. The framework we developed can be used by researchers and practitioners to accurately classify habitats
Ingénierie agile de lignes de produits logiciels pour des applications d’aide à la décision pour l’agriculture
No full textThe industrialization of software development in the business sector promotes the creation of custom software solutions, of better quality, and at a lower cost. In this context, migrating to a Software Product Line (SPL) represents an interesting option for companies that already offer a range of similar software products.The industrial partner for this research project is the ITK company. This company develops a range of decision support software products for various agricultural cultures and is planning to migrate to an SPL.To facilitate a successful migration to a software product line, it is essential to prepare the various project stakeholders to adopt new methods and practices. This is why a thorough assessment of the expectations and readiness of the stakeholders is necessary to ease this transition.The first contribution of this thesis is to carry out this evaluation. It was conducted through interviews with project stakeholders and a detailed analysis of their responses.Migrating the existing software range to an SPL involves localizing and identifying the existing functionalities. Creating representative feature models of the system to be migrated is a crucial initial step. Many resources are linked to the source code and can be leveraged to enrich the migration, feature model generation, and SPL maintenance.Source code version control platforms, for example, provide means to link specifications to source code, primarily through the use of user stories and feature-based code merges. The second contribution of this thesis is an automated process that utilizes user stories to generate feature models. This method combines natural language processing, supervised artificial intelligence models, and formal and relational concept analysis to generate feature models for each user of the studied software. These models are enriched by considering constraints from code merges and a domain ontology.The third contribution of this thesis is a method for identifying variability in the data schemas of different simulators for existing software. In fact, the development of a new software at ITK begins with the integration of a simulator, developed by agronomists, for a new crop. This method is based on formal concept analysis and allows for enriching configurations to create new products.L'industrialisation du développement des logiciels en entreprise favorise la création de solutions logicielles sur mesure, de meilleure qualité et à moindre coût. Dans ce contexte, la migration vers une ligne de produits logiciels (LPL) représente une option intéressante pour les entreprises qui proposent déjà une gamme de produits logiciels similaires.Le partenaire industriel de ces travaux de recherche est l'entreprise ITK. Cette entreprise développe une gamme de produits logiciels d'aide à la décision pour différentes cultures agricoles et ont pour projet de migrer vers une LPL.Afin de favoriser une migration réussie vers une ligne de produits logiciels, il est essentiel de préparer les différents acteurs du projet à adopter de nouvelles méthodes et pratiques. C'est pourquoi une évaluation approfondie des attentes et de la préparation des parties prenantes est nécessaire, permettant ainsi de faciliter cette transition.La première contribution de cette thèse consiste à réaliser cette évaluation. Elle a été réalisée au moyen d'entretiens avec les acteurs du projet et d'une analyse approfondie de leurs réponses. La migration de la gamme de logiciels existants vers une LPL implique de localiser et d'identifier les fonctionnalités existantes. La création de modèles de fonctionnalités représentatifs du système à migrer constitue une première étape cruciale. De nombreuses ressources sont liées au code source et peuvent être exploitées pour enrichir la migration, la génération de modèles et la maintenance de la LPL.Les plateformes de gestion de versions du code source offrent par exemple des moyens de lier les spécifications au code source, notamment grâce à l'utilisation d'récits utilisateurs (user stories) et de fusions (merge requests) de code par fonctionnalités. La deuxième contribution de cette thèse est un processus automatisé qui utilise les récits utilisateurs pour générer des modèles de fonctionnalités. Cette méthode combine le traitement naturel du langage, des modèles supervisés d'intelligence artificielle et l'analyse formelle et relationnelle de concepts afin de générer des modèles de fonctionnalités pour chaque utilisateur des logiciels étudiés. Ces modèles sont enrichis en tenant compte des contraintes issues des fusions de code et d'une ontologie du domaine.La troisième contribution de cette thèse est une méthode identifiant la variabilité dans les schémas de données des différents simulateurs des logiciels existants. En effet, le développement d'un nouveau logiciel chez l'entreprise ITK débute d'abord par l'intégration d'un simulateur, développé par les agronomes, pour une nouvelle culture. Cette méthode est basée sur l'analyse formelle de concepts et permet d'enrichir les configurations pour de créer de nouveaux produits
Sea Bass (Dicentrarchus labrax) Tail-Beat Frequency Measurement Using Implanted Bioimpedance Sensing
No full textInternational audienceEstimating tailbeat frequency (TBF) is a crucial component of fish swimming kinematics and performance, particularly because it provides information about energetics and behavioral responses to environmental cues. The most commonly used technique for TBF estimation is based on accelerometers. This paper proposes a novel approach using bioimpedance technology. This is the first time bioimpedance has been measured in a freely moving animal. This was made possible by implanting a flexible electrode in the back muscle of seabasses and having them in a swimming tunnel. The experiment first demonstrates that it is possible to measure bioimpedance in an immersed fish despite the high conductivity of seawater. An agreement analysis was then performed to compare a video-based reference measurement of TBF with the newly proposed approach. Several bioimpedance settings, such as the configuration and the extracted electrical parameters, were considered. Data analysis highlights that a 4-point setup for modulus impedance measurement at frequencies over 10 kHz provides the best agreement (r > 0.98 and CCC > 0.97) with the video-based approach. These results attest to the significant benefits of integrating bioimpedance sensors in biologgers, especially considering the complementary parameters that can be extracted from bioimpedance measurements, such as length, weight, condition index, and fat content
Upper Limbs and Low-Back Loads Analysis in Workers Performing an Actual Industrial Use-Case with and without a Dual-Arm Cobot
No full textInternational audienceIn the industry 4.0 scenario, Human-Robot Collaboration (HRC) plays a key role in factories to reduce cost, increase production and help aged and/or sick workers maintain their job. The approaches of the ISO 11228 series commonly used for biomechanical risk assessment cannot be applied in Industry 4.0 as they do not involve interaction between workers and HRC technologies. The use of wearable sensor networks and software for biomechanical risk assessment could help us have a more reliable idea on the e^ectiveness of collaborative robot (coBots) in reducing biomechanical load for workers. The aim of the present study was to investigate some biomechanical parameters with 3D Static Strenght Prediction Program (3DSSPP) software, on workers executing a practical manual material handling task, by comparing a dual arm cobot assisted scenario with a no cobot scenario. The parameters investigated were percent of Maximum Voluntary Contraction, Maximum static (continuous) allowed exertion time, Low back spine orthogonal compression forces at L4/L5 level and Strength Percent Capable. In this study, we calculated mean and the standard deviation (SD) values from eleven participants for some 3DSSPP parameters. We considered the following parameters: Percent of Maximum Voluntary Contraction (%MVC), Maximum static allowed exertion time (MaxST), Low back spine compression forces (L4Ort) and Strength Percent Capable (SPC). The advantages in introducing the cobot, ac-cording to our statistics, concern trunk flexion (SPC from 85.8% without cobot to 95.2%; %MVC from 63.5% without cobot Vs. 43.4%; MaxST from 33.9s without cobot to 86.2s), left shoulder abdo-adduction (%MVC from 46.1% without cobot Vs. 32.6%; MaxST from 32.7s without cobot to 65s) and right shoulder abdo-adduction (%MVC from 43.9% without cobot Vs. 30.0%; MaxST from 37.2s without cobot to 70.7s) in Phase 1; right shoulder humeral rotation (%MVC from 68.4% without cobot Vs. 7.4%; MaxST from 873.0s without cobot to 125.2s), right shoulder abdo-adduction (%MVC from 31.0% without cobot Vs. 18.3%; MaxST from 60.3s without cobot to 183.6s)and right wrist flex/extension rotation (%MVC from 50.2% without cobot Vs. 3.0%; MaxST from 58.8s without cobot to 1200.0s) in Phase 2. Moreover, Phase 3, consisting of another manual handling, would be removed by using the cobot. In summary using the cobot in this industrial scenario, would reduce the biomechanical risk for workers particularly for trunk, both shoulders and right wrist. Finally, 3DSSPP software could be an easy, fast and costless tools for biomechanical risk assessment in industry 4.0 scenario where ISO 11228 series can't be applied for occupational medicine physicians and health and safety technicians and helping employers to justify a long-term investment
Uma Heurística para a Execuc ão de Workflows com Restricões de Confidencialidade em Ambientes Conteinerizados
No full textInternational audienceContainerized environments are ideal for running scientific workflows because they offer a flexible and easily instantiated setting. Although there are solutions for executing workflows in containerized environments, these were not designed to handle scientific workflows, especially those with confidentiality requirements. Non-compliance with these requirements allows malicious users to infer unpublished results or the workflow structure itself. Data dispersion and encryption can be adopted in this context, but not independently of workflow scaling, as this can increase the total execution time or the associated financial cost. In this paper, we present Okinawa, a heuristic for executing workflows in containerized environments with confidentiality constraints.Ambientes conteinerizados são ideais para a execução de workflows científicos, pois oferecem um ambiente flexível e de fácil instanciação. Embora existam soluções para execução de workflows em ambientes conteinerizados, estas não foram projetadas para lidar com workflows científicos, especialmente aqueles com requisitos de confidencialidade. A não conformidade com esses requisitos permite que usuários mal-intencionados infiram resultados não publicados ou a prõpria estrutura dos workflows. A dispersão de dados e a criptografia podem ser adotadas nesse contexto, mas não de forma independente do escalonamento, pois isso pode aumentar o tempo total de execução ou o custo financeiro associado. Neste artigo, apresentamos a Okinawa, uma heurística para execução de workflows em ambientes conteinerizados com restrições de confidencialidade
Collaborative Benchmarking Rule-Reasoners with B-Runner
No full textNational audienceConducting experimental analysis on rule reasoners is a mainstream task for validating novel algorithms and systems. Nevertheless, providing robust, veriable, and reproducible experiments can still raise a sensible challenge. We propose to demonstrate B-Runner, an open library for collaborative benchmarking focusing on the deployment of articulate tests for knowledge and rule-based systems with low cost and high robustness. B-Runner reduces the benchmarking setup time while guaranteeing experiment repeatability. At the same time, it improves the scrutability of experimental protocols thereby enhancing their robustness as well as fairness of system comparisons. This demonstration proposes to show- case the use of the tool for systematically testing a number of systems as well as to introduce its architecture and its extensibility to novel tools and experimental protocols
Critical Exponent of Binary Words with Few Distinct Palindromes
No full textInternational audienceWe study infinite binary words that contain few distinct palindromes. In particular, we classify such words according to their critical exponents. This extends results by Fici and Zamboni [TCS 2013]. Interestingly, the words with 18 and 20 palindromes happen to be morphic images of the fixed point of the morphism , ,
Explaining Metaphors in the French Language by Solving Analogies using a Knowledge Graph
No full textInternational audienceAn analogy is a relation which operates between two pairs of terms representing two distant domains. It operates by transferring meaning from a concept that is known to another that one would like to clarify or define. In this report, we address analogy both from the aspect of modeling and by automatically explaining it. We will then propose a system of resolution of analogical equations in their notation in symbol chains. The model, based on the common sense knowledge base JeuxDeMots (a semantic network), operates by generating a list of potential candidates from which it chooses the most suitable solution. We conclude by evaluating our model on a collection of equations, and reflecting upon future work
Adaptation de Yolov8 pour la détection d'objets avec peu d'exemples
No full textNational audienceRecent networks for object detection obtain excellent performance when trained on large databases but still have difficulties learning a new object with few examples. The methods of this domain using rather heavy architectures, we have decided to adapt the modules presented in DeFRCN into the newer and faster architecture of Yolov8. We show here the impact of this modification on the MSCOCO bechmark, and finally discuss about the biases existing in this method.Les réseaux récents pour la détection d’objets obtiennent d’excellentes performances quand ils sont entraînés sur de grandes bases de données, mais ont toujours des difficultés pour apprendre un nouvel objet avec peu d’exemples. Les méthodes de ce domaine utilisant plutôt des architectures lourdes, nous avons décidé d’adapter les modules présentés dans DeFRCN dans l’architecture plus récente et rapide de Yolov8. Nous montrons ici l’impact de cette modification sur le benchmark MSCOCO, et finalement parlons des biais de cette méthode