HAL portal of the University of Lorraine
Not a member yet
161261 research outputs found
Sort by
Data driven prediction of thermal hysteresis in NiTi-based high entropy shape memory alloys
No full textInternational audienceThis study provides a data-driven approach to predict thermal hysteresis in NiTi-based high-entropy shape memory alloys (HESMAs), using a dataset of NiTi-based shape memory alloys that includes the four transformation temperatures (M s , M f , A s , A f ). Machine learning models based on extremely randomized trees were proposed to predict M s and the thermal hysteresis (ΔT = A f -M s ), using 11 physical material descriptors as inputs. M s prediction is dominated by mixing enthalpy, the valence electron ratio, and the difference in electronegativity. In contrast, hysteresis depends equally on 8 additional balanced features such as the melting temperature, electronegativity, atomic radius, and the mixing entropy. The predictions models are highly reliable with mean absolute errors of 31 • C for M s and 11 • C for ΔT. Finally, a design strategy is proposed for (NiCuCo)(TiZrHf) HESMAs that optimises both M s and ΔT with targeted values.</div
Marchés prédictifs en ligne : miser sur un résultat sportif… ou sur la guerre en Ukraine
No full textInternational audienceCet article publié sur The Conversation France interroge les limites morales et éthiques des marchés prédictifs, où de nombreux parieurs spéculent désormais sur la guerre en Ukraine
The Lucie-7B LLM and the Lucie Training Dataset: Open resources for multilingual language generation
No full textWe present both the Lucie Training Dataset and the Lucie-7B foundation model. The Lucie Training Dataset is a multilingual collection of textual corpora centered around French and designed to offset anglo-centric biases found in many datasets for large language model pretraining. Its French data is pulled not only from traditional web sources, but also from French cultural heritage documents, filling an important gap in modern datasets. Beyond French, which makes up the largest share of the data, we added documents to support several other European languages, including English, Spanish, German, and Italian. Apart from its value as a resource for French language and culture, an important feature of this dataset is that it prioritizes data rights by minimizing copyrighted material. In addition, building on the philosophy of past open projects, it is redistributed in the form used for training and its processing is described on Hugging Face and GitHub. The Lucie-7B foundation model is trained on equal amounts of data in French and English -- roughly 33% each -- in an effort to better represent cultural aspects of French-speaking communities. We also describe two instruction fine-tuned models, Lucie-7B-Instruct-v1.1 and Lucie-7B-Instruct-human-data, which we release as demonstrations of Lucie-7B in use. These models achieve promising results compared to state-of-the-art models, demonstrating that an open approach prioritizing data rights can still deliver strong performance. We see these models as an initial step toward developing more performant, aligned models in the near future. Model weights for Lucie-7B and the Lucie instruct models, along with intermediate checkpoints for the former, are published on Hugging Face, while model training and data preparation code is available on GitHub. This makes Lucie-7B one of the first OSI compliant language models according to the new OSI definition
Le Jeune dessinateur de Jan Lievens : support d’une réflexion sur la contemporanéité éducative
No full textInternational audienceCet article propose une réflexion sur la contemporanéité de l’éducation à partir d’une analyse du tableau Le Jeune dessinateur de J. Lievens (1607-1674). En s’appuyant notamment sur les travaux de G. Agamben et H. Arendt, il envisage la peinture non comme un objet historique, mais comme un déclencheur intempestif de pensée éducative. À rebours d’une pédagogie centrée sur la transparence, la lumière et l’efficacité, l’article défend une pensée de l’ombre comme puissance éducative, valorisant la suspension, l’indétermination et l’inachèvement propres à l’atelier en tant que dispositif de formation. L’ombre, loin d’être un simple défaut de lumière, devient ici un espace de création et de profanation des usages scolaires dominants. À travers une lecture philosophique du tableau, l’article invite à penser l’éducation comme une expérience – entendue au sens de skholé – ouvrant un autre rapport au savoir, aux objets et au temps
Projet SARHAE - Le site archéologique des Rothenstauden à Voellerdingen (67) étude d’un habitat antique et de son environnement - Rapport d’opération 2025
No full textLe site archéologique des Rothenstauden est localisé sur la commune de Vœllerdingen, dans le département du Bas-Rhin. Il fait depuis 2021 l’objet du programme pluridisciplinaire « SARHAE : Le site des Rothenstauden – Vœllerdingen (67) - Étude pluridisciplinaire d’un habitat antique et de son environnement ». L’année 2025 a notamment été marquée par la poursuite de la fouille extensive dans le secteur 100 (avec la réalisation d’un conséquent protocole de prélèvement). En raison des conditions météorologiques, cette opération n’a pas pu être achevée comme prévue. Elle a tout de même permis de mettre au jour les premiers niveaux d’occupations au sein de ce secteur de la pars rustica de la villa et de constater qu’il a connu de nombreuses phases d’évolutions. Suite à une première période d’aménagement (laténienne ?), la zone est dévolue à des activités liées au feu au début de l’Antiquité. Entre la fin du Ier s. ap. J.-C. et le IIe s., le bâtiment maçonné est construit. Il va connaître plusieurs périodes d’aménagements et sans doute plusieurs fonctions jusqu’au IIIe s., où il sera abandonné.En parallèle de la fouille, d’autres travaux ont été menés au cours de l’année 2025. De nouvelles études géomorphologiques et pédo-sédimentaires ont été réalisées et l’analyse des données LiDAR et géophysiques a été approfondie. Ce volet permet désormais d’avoir une bonne image de l’étendue du site et de la composition de son environnement qui est structuré par un vaste parcellaire. À partir de 2026, il s’agira de réaliser de nouveaux sondages exploratoires pour valider certaines observations et hypothèses.Si des zones d’ombres persistent encore et si les problématiques scientifiques sont encore loin d’avoir trouvé toutes leurs réponses, le projet SARHAE a déjà permis d’obtenir, depuis son lancement en 2021, de nombreuses données sur le site archéologique des Rothenstauden et sur son environnement. En outre, la démarche employée a permis de tracer plusieurs pistes méthodologiques sur l’étude des vestiges archéologiques sous forêt. Les travaux prévus au cours des prochaines années viendront progressivement compléter les données sur les relations entre une villa et son environnement, sur son fonctionnement et plus globalement sur l’histoire des espaces qui sont actuellement sous forêt
The industrial cost of fixed exchange rate regimes
No full textInternational audiencePremature deindustrialization in most emerging and developing economies is one of the most striking stylized facts of the recent decades. In this paper, we provide solid empirical evidence supporting that the choice of a fixed exchange rate regime accelerates this phenomenon. Relying on a panel of 146 developed, emerging, and developing countries over the 1974–2019 period, we show that fixed exchange rate regimes have had a negative, significant, and robust effect on the size of the manufacturing sector—developing countries being the most affected by the industrial cost of such a regime. Additional gravity model regressions show that the impact of fixed regimes passes through the trade channel. In particular, this regime has kept countries with low relative productivity in a state of structural dependence on imports of manufactured products to the detriment of the emergence of a strong local manufacturing sector
Ribosomal protein L5 (RPL5/uL18) I60V mutation is associated to increased translation and modulates drug sensitivity in T-cell acute lymphoblastic leukemia cells
No full textInternational audienc
Clinical Phenotypes of Critically Ill Patients with COVID-19 Infected with Omicron: A Nationwide Prospective Cohort Study
No full textInternational audienceINTRODUCTION: The clinical presentation of critically ill patients with coronavirus disease 2019 (COVID-19) has evolved significantly with the emergence of the Omicron variant. Current intensive care unit (ICU) admissions involve patients with diverse comorbidities and immune statuses, highlighting the need to redefine homogeneous phenotypic subgroups within this population. This study aimed to characterize distinct clinical phenotypes among critically ill patients with COVID-19 and acute respiratory failure. METHODS: This multicenter prospective substudy of the SEVARVIR cohort included adult patients from 39 French ICUs between December 2021 and October 2024 with acute respiratory failure and infected with the Omicron variant. Clustering analysis was conducted using Kohonen’s self-organizing maps (SOMs) and validated with ClinTrajan, two unsupervised clustering methods, to identify homogeneous patient phenotypes. RESULTS: During the study period, 777 patients with Omicron infection were included, and 7 distinct clinical clusters were identified. Clusters 1 and 2 included patients with metabolic and cardiovascular comorbidities. Cluster 3 featured younger, mildly ill patients with isolated chronic respiratory failure, while cluster 4 comprised older male patients with isolated respiratory failure. Cluster 5 included patients with isolated hematologic malignancies, cluster 6 patients with multiorgan failure, and cluster 7 organ transplant recipients, with high severity scores and impaired renal function. ICU management varied substantially across clusters. Patients in clusters 5 and 7 had the highest requirements for organ support, with frequent use of invasive mechanical ventilation, vasopressors (cluster 6), and renal replacement therapy (cluster 7). Dexamethasone and tocilizumab were most commonly prescribed in cluster 4 (91.3% and 30.2%, respectively). Mortality at day 28 varied significantly across clusters, ranging from 13.1% in cluster 3 to 41.1% in cluster 6. CONCLUSIONS: This clustering analysis highlights, for the first time, the clinical heterogeneity of critically ill patients infected with Omicron, identifying seven distinct clusters with varying clinical presentations, management strategies and outcomes. These findings underscore the relevance of a phenotype-driven approach to support personalized treatment strategies and guide future clinical trials. TRIAL REGISTRATION: Clinicaltrials.gov, NCT05162508. A Graphical Abstract is available for this article
