HAL Portal de Univ. Gustave Eiffel
Not a member yet
72847 research outputs found
Sort by
An Illustrated Guide to Automatic Sparse Differentiation
No full textInternational audienceIn numerous applications of machine learning, Hessians and Jacobians exhibit sparsity, a property that can be leveraged to vastly accelerate their computation. While the usage of automatic differentiation in machine learning is ubiquitous, automatic sparse differentiation (ASD) remains largely unknown. This post introduces ASD, explaining its key components and their roles in the computation of both sparse Jacobians and Hessians. We conclude with a practical demonstration showcasing the performance benefits of ASD
Les jeux vidéo libres et open source dans l'enseignement
No full textNational audienceLes jeux vidéo de type Free/Libre Open Source Software (FLOSS) sont souvent méconnus car peu mis en avant par leurs communautés respectives. Pourtant, dans un contexte de formation, ils peuvent être de meilleures alternatives que certains produits propriétaires. C'est le cas des jeux présentés dans cette conférence : Minetest/Luanti et FlightGear
Représentations des risques d’inondation, adaptation et évacuation en Île-de-France
No full textInternational audienceThe study of risk perception determinants and their impacts on the levels of preparedness of exposed populations often leads to inconclusive results across case studies. Numerous competing theories and hypotheses have been formulated across different disciplines, but their empirical validation remains a challenge. The current fragmentation of research on risk perception hampers the cross-validation of previous results, which might make the field appear as a collection of independent case studies in some instances. We use a flood risk perception and emergency behaviour survey in Paris, France(n=2976), to discuss a theoretical and methodological approach to consolidating the position of geography in the field, better disentangling social and spatial factors. The research questions are structured around the role of hazard exposure, prior experience, place of residence, and social vulnerability on risk perception and emergency behaviour. The main hypothesis is that previous inconclusive results published in the literature on risk perception determinants stem from divergent disciplinary approaches and theories, as well as from the at times uncertain place of geography in the field.El estudio de las determinantes de la percepción del riesgo y sus efectos sobre los niveles de preparación de la población expuesta, entregan resultados contradictorios según el caso analizado. Si bien han sido propuestas una amplia gama de hipótesis y teorías, éstas restan contrapuestas entre disciplinas y ninguna ha podido prevalecer empíricamente. La validación cruzada de los resultados resulta compleja, y el campo de la percepción del riesgo se asocia generalmente a una colección de estudios de casos independientes y con una baja capacidad de hacerlos comparables. A partir de una encuesta realizada a 2976 personas referente a las representaciones del riesgo de inundación aplicada en la ciudad de París y en la región de Île-de-France, este artículo propone un enfoque teórico y metodológico en vías de consolidar la posición de la geografía en el campo de las representaciones del riesgo, con el fin de dilucidar de manera más robusta los factores sociales y territoriales. Las preguntas de investigación se estructuran en torno al rol de la exposición al peligro, la experiencia previa, las situaciones de vulnerabilidad, el lugar de residencia en las representaciones y los comportamientos. La hipótesis principal expresa que los resultados publicados en la literatura internacional para identificar las determinantes de estas representaciones son contradictorios, lo cuales derivan tanto de enfoques y teorías disciplinariamente divergentes, como del rol en ocasiones incierto que toma la geografía en el ámbito internacional de la percepción del riesgo.L’étude des déterminants des représentations liées aux risques et leurs effets sur les niveaux de préparation des populations exposées conduit à des conclusions souvent contradictoires d’une étude de cas à une autre. De nombreuses théories et hypothèses concurrentes ont été formulées dans des disciplines différentes, mais aucune n’est parvenue à s’imposer empiriquement. La validation croisée des résultats est complexe et le champ des risk perception s’apparente souvent à une collection d’études de cas indépendantes et peu comparables. À partir d’une enquête par questionnaire sur les représentations liées au risque inondation à Paris et en Île-de-France sur un échantillon de grande taille (n=2976), cet article propose une réflexion théorique et méthodologique pour consolider le positionnement de la géographie dans le champ des représentations des risques, pour mieux démêler les facteurs sociaux et territoriaux. Les questions de recherche se structurent autour du rôle de l’exposition aux aléas, de l’expérience préalable, des situations de vulnérabilité et des territoires de résidence sur les représentations puis sur les comportements. L’hypothèse principale est que les résultats contradictoires publiés dans la littérature internationale pour cerner les déterminants de ces représentations découlent d’approches disciplinaires et de théories divergentes, ainsi que de la place parfois peu assurée de la géographie dans le champ international des risk perception
Le bien-être avec le numérique, entre promesses de progrès, discours critiques et pratiques contrastées
No full textInternational audienceThis introduction recontextualizes the recurrence of questions about the positive effects of information and communication technologies throughout the history of Interfaces Numériques magazine. Then, delving into the roots of the notion of well-being, we look at how technological deployments relate to it, on a practical level, through the pleasures or facilities they bring, or through stories of improved lives. The growing concern for health, including mental health, intersects with the consequences of confinement, as artificial intelligence and new alienations become widespread. Based on the subjective position of the contemporary subject, outlined in broad strokes, a three-pronged approach is proposed, depending on whether well-being is achieved with, through or in the digital world. The collection of contributions, presented in pairs, starts with the user experience, examining the world of video games, then moves on to creativity and psychological support, and finally to learning and education.Cette introduction recontextualise au fil de l’histoire de la revue Interfaces Numériques la réccurrence d’un questionnement sur les effets positifs des technologies de l’information et de la communication. Plongeant ensuite aux racines de la notion de bien-être, il s’agit de saisir en quoi les déploiements technologiques s’y réfèrent, au niveau pratique, par les plaisirs ou facilités apportés ou des récits d’amélioration de vie. Le souci croissant de santé, y compris mentale, croise les conséquences des confinements, alors que se généralisent les intelligences artificielles et nouvelles aliénations. À partir de la position subjective du sujet contemporain, brossée à grands traits, une triple approche est proposée, selon que le bien-être se réalise avec, à travers ou dans le numérique. La collecte des contributions, présentées deux par deux, s’ordonne en partant de l’expérience utilisateur, en interrogeant les univers des jeux vidéo, pour aller vers la créativité, le soutien psychologique et enfin l’apprentissage et l’éducation
Innover dans les organisations grâce au wargaming analytique
No full textInternational audienceAvec les premières journées d’étude transdisciplinaires dédiées au wargaming , la création de l’Association Française d’Étude du Wargaming (AFEW) et la parution d’un numéro double sur ce sujet dans la revue Stratégique, l’année 2024 a acté formellement la renaissance française des pratiques du wargaming et, surtout, le souhait de concevoir des méthodologies et pratiques rigoureuses (Fouillet, 2022, Lépinard, 2024 ; 2025, Vallat, 2024). Le travail présenté dans cette proposition de communication s’inscrit dans cette volonté et détaille l’expérimentation d’une méthodologie complète de wargaming analytique afin de comprendre en quoi cette pratique peut soutenir les innovations dans les organisations ? Notre étude a été menée en février 2025 dans le cadre du projet pédagogique et de recherche en ludopédagogie EdUTeam avec un groupe de quinze alternants de 2nde année de master Management de la sécurité des systèmes d‘information (MSSI) de l’IAE Paris-Est, l’école de management de l’Université Paris-Est Créteil (UPEC)
Estimating Spatial Mean Speeds from Local Sensors: a Machine Learning Approach
No full textInternational audienceAccurate Network Macroscopic Fundamental Diagram (NMFD) estimation requires accurate spatial mean speed estimates that are hard to get from loop detector devices (LDD) as they capture local speeds only. This study introduces a correction method able to reconstruct the mean speed from loop data, leveraging floating car devices (FCD) during the training phase. This significantly improves LDD-based NMFD estimation. Unlike previous studies focusing on local speed corrections, our approach integrates LDD and FCD using machine learning techniques to enhance link-average speed estimation. We compare the performance of Ordinary Least Squares Regression (OLSR), Random Forest (RF), Multilayer perceptron (MLP), and XGBoost models, using a comprehensive dataset to evaluate their effectiveness in aligning LDD speeds with FCD as ground truth average speed. The RF model showed the highest accuracy, with a 37.44% improvement over original LDD speeds and a Root Mean Squared Error (RMSE) of 9.54 km/hr. Additionally, we investigated the impact of LDD positions and demand by separating loading and unloading periods. The RF model's adjustments were especially effective under these conditions, resulting in an improved accuracy of 44.9%. Our analysis of NMFDs further validated the RF model's robustness and accurate estimation of NMFDs using corrected LDD. We validate our methodology across two urban networks, Athens and Lyon, demonstrating its spatial and temporal transferability. Results show that bias correction significantly improves LDD speed accuracy, leading to more reliable NMFD estimation. This work advances data fusion methods for urban traffic monitoring and provides a robust framework for addressing LDD positional bias in large-scale traffic studies
Plastic debris dataset on the Seine riverbanks: up to 38 000 pre-production plastic pellets reported per square meter
No full textInternational audiencePlastic pollution in rivers is a major source for plastic pollution into the ocean. However, it is now recognized that plastics may accumulate in rivers for years, especially in estuaries, before reaching the ocean. This long residence time favours fragmentation of macroplastics into smaller and smaller pieces, but relative data are still carse. Here we present data from the downstream part of the Seine estuary in a historical deposit full of plastic debris, with the highest concentration of industrial plastic pellet ever reported in France. Plastic debris (down to 6 mm according to sieving limits) were classified using the updated European classification J-list. The sampled site is located close to the river mouth within a national natural reserve, surrounded by international harbour activities and two major industrial plastic producers: Total Energy and Exxon Mobil. A surface of only 1 m2 was sampled in a visual maximum of plastic pollution. Over 100 000 plastic debris were counted or estimated when it comes to plastic debris <6 mm. Items were classified and weighted by category for a total mass higher than 4 kg. By mass (count), 24% (38%) of total plastic debris were pre-production plastic pellets, 21% (1%) were unidentified, colourful, plastic fragments (≥2.5 cm), and 19% (33%) were unidentified, colourful, plastic fragments (<2.5 cm)
Addressing medical students' health challenges: codesign and pilot testing of the Preventive Remediation for Optimal MEdical StudentS (PROMESS) program
No full textInternational audienceBackground Medical students often encounter challenges in maintaining healthy habits. The Preventive Remediation for Optimal MEdical StudentS (PROMESS) project seeks to support students to adopt healthier lifestyles throughout their curriculum by implementing a program focused on three modules (stress, sleep, and physical activity). Prior to implementation, it was essential to gain insights into students' needs. For this purpose, a comprehensive approach was adopted to ensure that the proposed program aligns with students' needs, to identify barriers and facilitators for implementation, to propose adjustments, and to test the program.Methods A three-step study was conducted. First, two focus groups sessions, one involving students and the other involving university staff members were conducted to identify medical students' needs and obstacles to change regarding their health. After verbatim transcription, a framework thematic analysis was performed with the use of MAXQDA. Second, a co-construction workshop was conducted with participants from both groups to develop the PROMESS program. Third, some of the medical students who participated in the prior steps tested the co-constructed program. The study was conducted at the Lyon-Est Faculty of Medicine (France, IRB2023070404). Results (i)The medical students cited a heavy academic workload and demanding internships as the main factors that contribute to their limited selfcare, heightened stress levels and sleep disruptions, and reduced physical activity. The university staff members noted that students struggled to acknowledge their needs, accept limitations, and seek assistance.(ii) The participants provided advice for adapting PROMESS program to the students' specific needs (e.g., individualized advice, one-on-one meetings, peer coaching, and signing a commitment contract).(iii) The students who tested the program reported being more aware of their health behaviors and reported improvement in stress levels, sleep, and physical activity. They believed that the changes could be long-lasting
Developing Cancer Quality of Care Indicators to Quantify the Impact of a Global Destabilization of the Care System (COLLAT-COVID)
No full textInternational audienceBackground/Objectives: The COVID-19 pandemic led to significant disruptions in healthcare systems, particularly impacting cancer care through delays in diagnoses and treatments. Quality indicators (QIs) are essential tools for monitoring healthcare performance, yet existing QIs may not be suited for crises. This study aimed to develop a set of hospital-based QIs tailored to assess the impact of care reorganization during health crises across four cancer types: breast cancer, hepatocellular carcinoma, gynecological cancers (excluding ovarian cancer), and peritoneal carcinomatosis. Methods: A multidisciplinary steering committee (SC) conducted a five-stage process, including a literature review, indicator selection, content validation via the RAND/UCLA method, final validation by the SC, and a pilot feasibility study. QIs were assessed based on clinical relevance, reproducibility, sensitivity to change, and feasibility. Expert panels evaluated and validated the indicators in two rounds of voting. Results: Among 150 initially identified QIs, 49 were validated: 12 for breast cancer, 11 for hepatocellular carcinoma, 8 for gynecological cancers, and 18 for peritoneal carcinomatosis. Most (92%) were process indicators, covering diagnosis, treatment, and care delays. Two common indicators were identified across all four cancers: multidisciplinary team meeting discussions and psychological support consultations. Conclusions: This study demonstrates the feasibility of developing crisis-responsive QIs to monitor cancer care during health system disruptions. Future work will focus on their real-time implementation, validation in international settings, and integration into healthcare policies to enhance crisis preparedness