HAL Portal UDL Université de Lyon
Not a member yet
327890 research outputs found
Sort by
Enseigner la prononciation de l'anglais en 2025 : tensions entre les prescriptions institutionnelles, les réalités sociolinguistiques et les perceptions des futurs enseignants
No full textInternational audienc
Generating Efficiently Realistic Counterfactual Explanations
No full textInternational audienceThis article introduces VCNet -Variational CounterNet -a method for generating realistic counterfactuals, and its extension ImmutableVCNet. VCNet aims to generate counterfactuals that are representative of their predicted classes in the context of tabular data. Moreover, it aims to overcome the limitations related to the posthoc character and optimization procedure of the state-of-the-art approaches, by reducing the computing time of the counterfactual generation and reaching high levels of validity. However, state-of-the-art methods that succeed in overcoming the aforementioned limitations suffer from a lack of realism with regard to the counterfactuals generated. VCNet addresses this concern by adding realism constraints to the counterfactual generation process. Our approach is based on a conditional variational autoencoder (cVAE) to model the distributions for every class at once. Thus, generated counterfactuals not only belong to the data distribution but belong to the data distribution of a given class. The ImmutableVCNet extends VCNet to overcome the limitation of handling immutable features. We conducted several ablation studies to investigate the impact of the different choices within the VCNet architecture. Furthermore, we conducted empirical studies that demonstrate the significant benefits of our approach in generating realistic counterfactuals. We evaluate VCNet against ImmutableVCNet and also ImmutableVCNet against a variety of state-of-the-art counterfactual methods
Factors Influencing STEM Self-Efficacy Among Primary School Pupils: A Gender Comparison
No full textInternational audienceThis study examines the influence of gender-related beliefs on STEM self-efficacy among primary school pupils in Malaysia, a context shaped by cultural diversity and hierarchical norms. While gender gaps in Science, Technology, Engineering, and Mathematics (STEM) are often associated with early-formed stereotypes and mindsets, their specific influence on children’s STEM self-efficacy remains under-explored, particularly in Southeast Asian settings. This study compares male and female pupils’ gender stereotypes in STEM, male-power stereotype, perceptions of STEM, growth mindset, and STEM self-efficacy. It also investigates whether the relationships between these factors and STEM self-efficacy differ by gender. A quantitative comparative design was employed, involving 415 primary school pupils aged 8–9 years (Year 2–3). Five instruments were used; the Gender Stereotypes Measure, Gender-Power Attribution Task, Self-Perception Scale in STEM, Growth Mindset Sub-scale from the Implicit Theories of Intelligence Scale for Children (ITIS), and the Self-Efficacy Scale in STEM. Data were analysed using multivariate analysis of variance (MANOVA), Chi-square tests of independence, and multiple regression analysis. Findings revealed that male pupils held stronger gender stereotypes in STEM, while no significant gender differences were found in other variables. Regression analysis showed that a growth mindset and positive perceptions of STEM were significant predictors of STEM self-efficacy. Gender stereotypes in STEM had a weaker but statistically significant positive effect, while male-power stereotype were not significant. Gender does not significantly moderate the relationships between the predictors and STEM self-efficacy. The findings highlight the importance of strengthening growth mindsets and fostering positive perceptions of STEM to support self-efficacy among pupils of both genders. These insights contribute to a deeper understanding of the factors influencing STEM self-efficacy and offer implications for addressing gender disparities in STEM education
Federated dynamic modeling and learning for spatiotemporal data forecasting
No full textInternational audienceThis paper presents an advanced Federated Learning (FL) framework for forecasting complex spatiotemporal data, improving upon recent state-of-the-art models. In the proposed approach, the original Gated Recurrent Unit (GRU) module within previous Dynamic Spatial-Temporal Graph Convolutional Recurrent Network (DSTGCRN) modeling is first replaced with a Long Short-Term Memory (LSTM) network, enabling the resulting model to more effectively capture long-term dependencies inherent to time series data. The resulting architecture significantly improves the model's capacity to handle complex temporal patterns in diverse forecasting applications. Furthermore, the proposed FL framework integrates a novel Client-Side Validation (CSV) mechanism, introducing a critical validation step at the client level before incorporating aggregated parameters from the central server into local models, ensuring only the most effective updates are retained and improving both the robustness and accuracy of the forecasting model across clients. The efficiency of our approach is demonstrated through extensive experiments on real-world applications, including public datasets for multimodal transport demand forecasting and private datasets for Origin-Destination (OD) matrix forecasting in urban areas. The results demonstrate substantial improvements over conventional methods, highlighting the framework's ability to capture complex spatiotemporal dependencies while preserving data privacy. This work not only provides a scalable and privacy-preserving solution for real-time, region-specific forecasting and management but also underscores the potential of leveraging distributed data sources in a FL context. We provide our algorithms as open-source on GitHub 1
Adult male patients with DKC1 mutations present early-onset pulmonary fibrosis and severe prognosis
No full textInternational audienceBackground Mutations in telomere-related genes (TRGs) are the main cause of monogenic familial pulmonary fibrosis. Dyskerin, encoded by the X-localised gene DKC1 , is involved in telomere maintenance. Methods This retrospective study aimed to further characterise the pulmonary phenotype of DKC1 -deficient patients with pulmonary fibrosis identified between 2010 and 2025 in our laboratory. Results We reported eight, as yet undescribed, to our knowledge, probands affected by pulmonary fibrosis associated with X-linked DKC1 deficiency. The median age at interstitial lung disease (ILD) diagnosis was 47 years. Four had idiopathic pulmonary fibrosis, two unclassifiable fibrosis, one idiopathic nonspecific interstitial pneumonia and one unknown. All patients displayed ectodermal abnormalities (premature hair greying, nail dystrophies, reticulated hyper- or hypopigmentation, tooth abnormalities and oral leukoplasia). Haematological abnormalities were found in four patients. Seven patients died during follow-up and one patient received a transplant. Median survival after ILD diagnosis was 22 months (range 3–81 months). Conclusions Compared with patients with pulmonary fibrosis associated with other TRG variants, those with pulmonary fibrosis associated with X-linked recessive DKC1 variants are younger and have a shorter survival time. Early identification and referral of these patients to an expert transplantation centre should be considered
Premiers pas vers la récupération de métaux en stations d'épuration domestiques : optimisation de l'étape de sorption
No full textInternational audienceIncreasing metal demand requires to move toward circular economy. This study explores domestic wastewater as a potential metal resource, assessing the effective recovery of 16 metals, 2 metalloids and phosphorus by 4 adsorbents (activated alumina, activated carbon, clinoptilolite, ferric hydroxide) and two cation-exchange resins (weak and strong). It is the first to evaluate the removal of such a wide range of elements-19 in total-comparing six different sorbents taking into account the complexity of a real wastewater, and thus providing tangible insights for the development of a recovery process. A design of experiments using batch-tests was set-up to study the effects of the material dosage, contact time and pH, and has determined the best operating conditions. We have shown that three materials have captured; 60% of the elements: P, Si, As, V, Zn, Mo, Cr and Sb for the ferric hydroxide, B, Co, Cr and Zn for the activated carbon, and Ca, Mg, Sr and Zn for the weak cation-exchange resin. A dosage of 20 g.L -1 of material was shown optimal for the capture by activated carbon and ferric hydroxide, while the weak cation-exchange resin performed best at 4 g.L -1 . A contact time of 1 h was identified as optimal for most elements and tested materials. The recovery would be substantial for Ca, Na, K, Mg, P, Si, Sr, but is still a challenge for others. Poor elements retention were shown for activated alumina and zeolite-clinoptilolite, and rapid saturation for the strong cation-exchange resin.La demande croissante en métaux nécessite de s'orienter vers une économie circulaire. Cette étude explore le potentiel des eaux usées domestiques comme source de métaux, en évaluant l'efficacité de la récupération de 16 métaux, 2 métalloïdes et du phosphore à l'aide de quatre adsorbants (alumine activée, charbon actif, zéolite clinoptilolite et hydroxyde ferrique) et de deux résines échangeuses d'ions (faibles et fortes). Il s'agit de la première étude à évaluer l'élimination d'un éventail aussi large d'éléments (19 au total) en comparant six sorbants différents, tout en prenant en compte la complexité des eaux usées réelles. Cette étude fournit ainsi des informations concrètes pour le développement d'un processus de récupération. Une série d'expériences utilisant des tests par lots a été mise en place pour étudier l'impact du dosage des matériaux, du temps de contact et du pH, et pour déterminer les meilleures conditions de fonctionnement.Nous avons montré que trois matériaux ont capturé 60 % des éléments métalliques : Le charbon actif a capturé le P, le Si, l'As, le V, le Zn, le Mo, le Cr et le Sb, la résine échangeuse de cations faibles le Ca, le Mg, le Sr et le Zn, et l'hydroxyde ferrique le B, le Co et le Cr. Un dosage de 20 g.L-1 de matériau s'est avéré optimal pour la capture par le charbon actif et l'hydroxyde ferrique, tandis que la résine échangeuse de cations faibles a donné les meilleurs résultats à 4 g.L-1. Un temps de contact d'une heure a été identifié comme optimal pour la plupart des éléments et des matériaux testés. La récupération est importante pour le calcium, le sodium, le potassium, le magnésium, le phosphore, le silicium et le strontium, mais reste difficile pour les autres éléments. Une faible rétention des éléments a été observée pour l'alumine activée et la zéolite clinoptilolite, et une saturation rapide pour la résine échangeuse de cations forts
GLIMPSE-D: An Exotic Balmer-Jump Object at z=6.20? Revisiting Photometric Selection and the Cosmic Abundance of Pop III Galaxies
No full textInternational audienceWe present deep JWST/NIRSpec G395M spectroscopy of GLIMPSE-16043, a promising Pop III candidate originally identified through NIRCam photometry as having weak [OIII] emission. Our follow-up reveals clear [OIII] emission, ruling out a genuine zero-metallicity nature. However, the combination of the measured line fluxes and photometry indicates that its spectral energy distribution requires an extraordinarily strong Balmer jump ( mag) and H equivalent width ( Å), features that cannot be reproduced by current stellar+nebular or pure nebular photoionization models. The only models approaching the observations to almost within involve a hot ( K) single blackbody embedded in a low- nebular environment, suggestive of scenarios such as a tidal-disruption event or a microquasar with strong disk winds. This cautions that photometric Pop~III selections are vulnerable to contamination when the rest-frame optical continuum is undetected. Motivated by this, we refine the photometric Pop III selection criteria to exclude the locus of extreme Balmer-jump objects. The revised criteria also recover the recently reported spectroscopic candidate AMORE6, demonstrating that the updated selection preserves sensitivity to genuine Pop III-like sources while removing key contaminants. Applying the refined criteria across legacy survey fields and five newly released CANUCS lensing cluster fields, we revisit the Pop III UV luminosity function and estimate the Pop III cosmic star-formation rate density to be --~~yr~cMpc at --7, falling in the range of current theoretical predictions
Tierce opposition effectuée par une société, cogérance et désistement
No full textInternational audienc
Exploring Bell nonlocality with extremal non-signaling boxes
No full textExtremal non-signaling (ENS) boxes are correlations that correspond to vertices of the non-signaling polytope of a Bell scenario. Neither quantum theory nor any theory for ideal measurements allows for ENS boxes. That is, according to quantum theory, ENS boxes are nonphysical. Still, ENS boxes are crucial for addressing a number of problems in Bell nonlocality. Here, we obtain ENS boxes in arbitrary bipartite Bell scenarios and present the complete list of ENS boxes for several unexplored scenarios. Equipped with the boxes, we revisit several foundational questions. We find that already two copies of any ENS box violate the exclusivity (or local orthogonality) and Specker's principles. We provide the minimal decomposition of the magic square correlation-the simplest known perfect correlation in nature-in terms of ENS boxes. We identify the minimal scenario in which a dit of communication (with d ⩽ 5) is insufficient to simulate ENS boxes. Our results show that the ENS boxes approach leads to new results and opens new avenues for research