HAL Evry
Not a member yet
19237 research outputs found
Sort by
Contrôles et usages des sous-sols urbains au XIXe siècle, un exemple des enjeux de la définition d’espaces délaissés
No full textInternational audienceCommunication lors de la session "Mobilités & appropriations des espaces délaissés", discutée par Charlotte Vorm
Pourra-t-on financer les infrastructures nécessaires aux besoins de mobilité de demain?
No full textInternational audienceTable-ronde (16h45 - 17h44
Evaluation des politiques de tarification du carbone : équité sociale, coûts économiques et mécanismes mondiaux
No full textUsing data from the 2017 French Household Budget Survey and a multiregional input-output model, I examine in the first chapter the distribution of households' carbon footprints. Nested multivariate models and quantile regression techniques are used to explore how household socioeconomic characteristics (size, age, education) and household decisions (home energy, dwelling type, car ownership) influence carbon emissions beyond income. Quantile regressions are used to address the gap in the literature, which mostly rely on OLS estimations without assessing whether these influences vary across the emissions distribution. Results reveal that household characteristics and several consumption decisions have an important influence on carbon footprints, though their importance differs by emission levels and sources. In the second chapter, I focus on the implementation of a carbon tax in France, using the same sort of data as in the previous chapter but incorporating a demand system. The study examines the difficult trade-off between minimizing emissions while ensuring social fairness under different revenue-recycling scenarios. While lump-sum transfers distributed equally across households may seem a second-best policy, this analysis shows that such transfers can spur a resurgence in emissions among overcompensated households, thereby wiping off part of the initial environmental gains. I therefore propose three alternative scenarios: a vertical scenario targeting low-income households, a horizontal scenario focusing on vulnerable households, and a mixed scenario combining both aspects. The vertical scenario effectively narrows the dispersion of tax burdens among the most affected households without significantly offsetting emissions reductions. The horizontal and mixed scenarios, though costlier, successfully reduce overall tax burden dispersion without causing an uptick in emissions. A key conclusion is that tailoring policies to particular households can, in principle, overcome the trade-off but aligning recycling optimization with both vertical and horizontal inequalities simultaneously may prove impractical beyond these microsimulation models. In the third chapter, we undertake a comprehensive analysis of the economic repercussions of carbon taxes on the global supply chain. Building on Leontief's input-output framework, we incorporate heterogeneous pass-through mechanisms into the price model, allowing firms to transfer carbon costs along the supply chain. Our simulations show that a carbon tax of 100 dollar could yield large revenues (2.82% of global GDP) but partially offset by a net cost (2.18%) and inflationary pressures. The results suggest that, in practice, the carbon tax is less efficient than predicted by the polluter pays principle. There is a strong imbalance in how costs and liabilities are distributed among major emitters: emerging economies face higher burdens than developed economies. Then, we enrich the analysis by merging sector-level data with financial data, enabling an assessment of transition costs. The discrepancy in cost-sharing is confirmed at the microeconomic level, where carbon-intensive sectors benefit from earnings at the expense of less carbon-intensive downstream industries. In the last chapter, I extend the modeling framework to study the EU CBAM. Specifically, I implement a markup-based pricing model calibrated on empirical data to estimate how CBAM costs cascade through the supply chain. This modeling approach allows for estimating policy-induced impacts, cost-sharing distribution, and the emergence of windfall profits. The findings indicate that CBAM's effectiveness depends not only on its product scope or tax rate but critically on the methodological choices used to compute compliance costs. These choices can shift cost incidence, deepen distributional asymmetries, and potentially compromise the policy's environmental credibility.En utilisant les données de l'Enquête Budget de Famille de 2017 et un modèle d'entrées-sorties multirégional, j'examine dans le premier chapitre la répartition de l'empreinte carbone des ménages français. Des modèles multivariés imbriqués et des techniques de régression quantile sont employés pour analyser comment les caractéristiques socio-économiques des ménages et leurs choix influencent leurs émissions carbones au-delà du seul effet revenu. Les régressions révèlent que les caractéristiques des ménages et plusieurs décisions de consommation ont une influence importante sur l'empreinte carbone, bien que leur importance diffère en fonction des niveaux et des sources d'émissions. Dans le deuxième chapitre, je me concentre sur la mise en œuvre d'une taxe carbone en France. L'étude examine le compromis entre réduction des émissions et maintien de l'équité sociale dans différents scénarios de recyclage. Bien que des transferts forfaitaires distribués de manière égale entre les ménages puissent sembler être une solution viable, mon analyse montre qu'un tel système peut provoquer une résurgence des émissions chez les ménages surcompensés, annulant ainsi une partie des gains environnementaux initiaux. Je propose donc trois scénarios alternatifs : un scénario vertical ciblant les ménages à faibles revenus, un scénario horizontal axé sur les ménages vulnérables et un scénario mixte combinant les deux aspects. Le scénario vertical améliore efficacement la situation des ménages les plus affectés sans compromettre la réduction des émissions. Les scénarios horizontal et mixte, bien que plus coûteux, parviennent à réduire la dispersion globale du fardeau de la taxe sans entraîner de hausse des émissions. L'adaptation des politiques à des ménages particuliers peut, en principe, surmonter le compromis, mais l'optimisation du recyclage en tenant simultanément compte des inégalités verticales et horizontales pourrait s'avérer impraticable au-delà de ces modèles. Dans le troisième chapitre, nous entreprenons une analyse des répercussions économiques des taxes carbone sur la chaîne de valeur mondiale. En nous appuyant sur le cadre des modèles d'entrées-sorties, nous intégrons des mécanismes de transmission hétérogènes dans le modèle de prix, permettant aux entreprises de répercuter le coût du carbone. Nos simulations montrent qu'une taxe carbone de 100 dollar pourrait générer d'importantes recettes, bien qu'elle soit partiellement compensée par un coût net et des pressions inflationnistes. Les résultats suggèrent qu'en pratique, la taxe carbone est moins efficace que ne le prédit le principe du pollueur-payeur. On observe un fort déséquilibre dans la répartition des coûts et des responsabilités parmi les principaux émetteurs : les économies émergentes supportent des fardeaux plus élevés que les économies développées. Nous enrichissons ensuite l'analyse en fusionnant des données sectorielles avec des données financières, permettant ainsi l'évaluation les coûts de la transition. La disparité dans le partage des coûts est confirmée à cet échelle, où les secteurs à forte intensité carbone bénéficient d'une hausse des revenus au détriment des industries aval moins carbonées. Dans le dernier chapitre, j'étends le cadre de modélisation pour étudier le MACF de l'UE. Plus précisément, j'implémente un modèle de tarification fondé sur les marges, calibré à partir de données empiriques, afin d'estimer la manière dont les coûts se répercutent le long de la chaîne de valeur. Cette approche permet d'analyser l'impact de la politique sur le partage des coûts et la formation de rentes économiques des producteurs. Les résultats indiquent que l'efficacité du MACF ne dépend pas seulement de la couverture ou du prix du carbone, mais surtout des choix méthodologiques retenus pour le calcul des coûts de conformité. Ces choix peuvent modifier l'incidence des coûts, accentuer les asymétries distributives et remettre en cause la crédibilité environnementale du MACF
Detection method for improving shape perception of small object defects on metal surfaces
No full textInternational audienceDefects on metal surfaces often exhibit complexity with diverse shapes, small sizes, and irregular patterns, leading to frequent missed and false detections during inspection and posing significant challenges to automated detection systems. Existing advanced object detectors, when applied directly to small defect detection on metal surfaces, fail to achieve satisfactory results. To mitigate these issues, we proposed a detection method to enhance the shape perception of small object defects on metal surfaces, namely MetalYOLO. Firstly, a novel location-aware attention mechanism is designed to integrate deformable convolutions to form a new feature selection module to enhance the focus on key defect features, optimizes the generation of offsets, and improve the model’s ability to adapt to complex shape objects. Secondly, the standard up-sampling module is replaced with a dynamic sampling module to dynamically adjust the sampling pattern of the input feature distribution to improve computational efficiency and retain complex or small-scale object features, thereby improving detection accuracy. Finally, a new detail-enhanced detection head is designed to further improve the network’s ability to capture fine-grained details by introducing a detail-enhanced attention-sharing module so as to utilize contextual information to selectively suppress irrelevant features, thereby reducing information redundancy. The proposed model is compared with baseline models on the ILS-MB and NEU-DET datasets. and the experimental results show significant improvements in false detection and missed detection rates with only a slight loss in inference speed. Meanwhile, the mAP reached 80.4% and 79.0%, respectively, which is 1.7% and 3.2% higher than the baseline algorithm
Phages with a broad host range are common across ecosystems
No full textInternational audiencePhages are diverse and abundant within microbial communities, where they play major roles in their evolution and adaptation. Phage replication - and multiplication - is generally thought to be restricted within a single or narrow host range. Here we use published and newly generated proximity ligation-based metagenomic Hi-C (metaHiC) data from various environments to explore virus-host interactions. We reconstructed 4,975 microbial and 6,572 phage genomes of medium quality or higher. MetaHiC yielded a contact network between genomes and enabled assignment of approximately half of phage genomes to their hosts, revealing that a substantial proportion of these phages interact with multiple species and this, in environments as diverse as oceanic water column or the human gut. This observation challenges the traditional view of a narrow host spectrum of phages by unveiling that multi-host associations are common across ecosystems, with implications how they might impact ecology and evolution and phage therapy approaches
A new generative framework for fast parametric modeling of X- and \gamma - ray energy spectra using VAE
No full textInternational audienceThis paper proposes a variational autoencoder-based technique for the fast modeling of various X- X - or \gamma - γ - ray spectra, including the parametrization of the physical measurement conditions, and introduces the SpectroGAN simulation tool. The generated spectra reproduce the training data with high precision. Still, they may vary according to the parameters of interest, such as detector energy resolution, efficiency, attenuation, activity, enrichment, or isotopic composition. Our results demonstrate that machine learning-based spectra simulations have several potential advantages over Monte Carlo-based simulation methods. First, they are much faster and more efficient to calculate. Second, they can provide accurate spectra even if the expected conditions are not included in the training dataset. The data-based model can learn complex relationships between the detector characteristics, measurement configuration, and the spectrometric features examined (such as signature peaks of elements). Finally, they can generate large amounts of synthetic data for training other machine learning models, such as those used in data analysis or pattern recognition. Overall, the data-based spectrometry can be generalized to various other applied radiation measurement tasks, including passive and active non-destructive measurement techniques featuring prompt and delayed neutrons and or gamma-rays detection, PGNAA and PGAINS applications, offering new opportunities for measurement system configuration optimization, as well as detector design. Official implementation is available at https://github.com/Tarysa/fast-parametric-modeling-of-X-and-ray-energy-spectra-using-VAE
The immersive Debriefing : Comparative evaluation of Full and Segmented Redo methods in Virtual Reality
No full textInternational audienceThis study investigates the impact of two immersive debriefing strategies on learning outcomes in a virtual reality (VR) simulation focused on mobile cybersecurity. The simulation highlights everyday mistakes to raise awareness of risky behaviors in public settings. We developed and evaluated an immersive debriefing system that enables participants to review their performance and re-engage with the scenario through a redo phase. This redo was implemented in two formats: a full scenario redo (F-REDO) and a segmented redo targeting specific moments (S-REDO). We found that both redo formats were equally effective in learning outcomes and user satisfaction, using a mixed-methods approach combining standardized questionnaires (motivation, cognitive load, usability, knowledge retention) and qualitative trainer feedback. However, S-REDO demonstrated greater time efficiency without increasing cognitive load and was perceived by the trainer as more engaging and pedagogically relevant. These results support the integration of personalized, interactive debriefing tools in VR learning environments, particularly in domains requiring targeted remediation and critical decision-making
To Bubble or Not to Bubble: Asset Price Dynamics and Optimality in OLG Economies
No full textWe study an overlapping generations (OLG) exchange economy with an asset that yields dividends. First, we derive general conditions, based on exogenous parameters, that give rise to three distinct scenarios: (1) only bubbleless equilibria exist, (2) a bubbleless equilibrium coexists with a continuum of bubbly equilibria, and (3) all equilibria are bubbly. Under stationary endowments and standard assumptions, we provide a complete characterization of the equilibrium set and the associated asset price dynamics. In this setting, a bubbly equilibrium exists if and only if the interest rate in the economy without the asset is strictly lower than the population growth rate and the sum of per capita dividends is finite. Second, we establish necessary and sufficient conditions for Pareto optimality. Finally, we investigate the relationship between asset price behaviors and the optimality of equilibria
Discrimination Based on Race: The Story of Moroccan SNCF Workers in France
No full textInternational audienc
Machine learning in predictive biocatalysis: A comparative review of methods and applications
No full textInternational audienceIn recent years, machine learning has significantly advanced predictive biocatalysis, enabling innovative approaches to enzyme function prediction, biocatalyst discovery, reaction modeling, and metabolic pathway optimization. This review provides a comparative analysis of current methodologies, highlighting the intersection between computational tools and biochemical data for predictive biocatalysis applications. Key aspects covered include enzyme classification, reaction annotation, enzyme-substrate specificity, reaction outcomes, and kinetic parameter prediction. We discuss various machine learning approaches, such as neural networks with increased depth, convolutional networks, graph-based architectures, and transformer models, highlighting their respective strengths and limitations. The integration of large-scale data, representation and featurization techniques, and robust validation methods has accelerated enzyme discovery and the development of eco-friendly, sustainable biocatalytic processes. In the future, machine learning is anticipated to play a central role in connecting computational insights with practical enzyme engineering efforts, advancing applications in synthetic biology, metabolic engineering, and green biocatalysis