HAL-EMSE
Not a member yet
9132 research outputs found
Sort by
L’optimisation de la production de semi-conducteurs par l’IA : quelles transformations des interactions homme/système
No full textNational audienc
Exploring Climate Change Through the Lens of Records Theory
No full textInternational audienceThe global phenomenon of climate change has become one of the most pressing challenges of our time.Amidst this complex landscape, the scientific community seeks innovative methodologies to quantify and understand the occurrence and severity of climate change. The theory of records in extreme value theory provides a promisingframework for detecting and verifying climate change. This research spans multiple geographic locations, employing a univariate scoring approach to analyze temperature and precipitation trends, and providing valuable insights into the occurrence, frequency, and intensity of extreme climate events. Our findings reveal a global warming pattern,with temperature scores declining mid-century before rising sharply, often surpassing previous levels. Precipitation scores vary across regions, with some experiencing drier conditions while others face increased rainfall. These results highlight shifting climate extremes, reinforcing the urgency of climate change analysis.Le phénomène mondial du changement climatique est devenu l'un des défis les plus urgents de notre époque. Dans ce contexte complexe, la communauté scientifique cherche des méthodologies innovantes pour quantifier et comprendre l'occurrence et la gravité du changement climatique. La théorie de records en statistique des extrêmes offre un cadre prometteur pour détecter et vérifier le changement climatique. Notre recherche couvre plusieurs lieux géographiques et utilise une approche de notation univariée pour analyser les tendances de température et de précipitations, fournissant ainsi des informations précieuses sur l'occurrence, la fréquence et l'intensité des événements climatiques extrêmes. Nos résultats révèlent un schéma de réchauffement global, avec des scores de température en baisse au milieu du siècle avant d'augmenter brusquement, dépassant souvent les niveaux précédents. Les scores de précipitations varient selon les régions, certaines connaissant des conditions plus sèches tandis que d'autres font face à une augmentation des précipitations. Ces résultats mettent en évidence des extrêmes climatiques changeants, soulignant l'urgence d'analyser le changement climatique
Rank-based Linear-Quadratic Surrogate Assisted CMA-ES
No full textInternational audienceIn this poster, we introduce a rank-based surrogate-assisted variant of CMA-ES. Unlike previous methods that employ rank information as constraints to train an SVM classifier, our approach employs a linear-quadratic regression on the ranks. We investigate the method's invariance empirically. While this first algorithm outperforms CMA-ES with a few exceptions, it falls short to entirely meet the lq-CMA-ES performance levels. To address this, we propose an enhanced variant that handles together two alternative surrogates, one based on the ranks and one based on the original function values. Although this variant sacrifices strict invariance, it gains in robustness and achieves performance comparable to, or even exceeding, lq-CMA-ES on transformed problems. This last algorithm shows how simply incorporating new transformations of rank values could improve any surrogate-based CMA-ES variant.Dans ce poster, nous présentons une variante de CMA-ES assistée par un métamodèle basé sur les rangs. Contrairement aux approches précédentes qui exploitent l'information de rang comme contrainte pour entraîner un classifieur SVM, notre méthode repose sur une régression linéaire-quadratique appliquée aux rangs. Nous étudions empiriquement l'invariance de cette approche. Bien que ce premier algorithme surpasse CMA-ES dans la plupart des cas, il n’atteint pas pleinement les performances de lq-CMA-ES. Pour pallier cette limitation, nous proposons une variante améliorée combinant deux modèles alternatifs : l’un fondé sur les rangs, l’autre sur les valeurs originales de la fonction objectif. Cette nouvelle version renonce à l’invariance stricte, mais gagne en robustesse et atteint des performances comparables, voire supérieures, à celles de lq-CMA-ES sur des problèmes transformés. Ce dernier algorithme montre comment l’intégration simple de nouvelles transformations des rangs peut améliorer toute variante de CMA-ES reposant sur un métamodèle
Efficient Bayesian Linear Models for a Large Number of Observations
No full textInternational audienceBayesian linear models are widely used as efficient approaches for nonparametric function estimation. In this paper, we present a Bayesian method for generating finite-dimensional linear models that can handle large datasets. This method is based on an efficient Markov chain Monte Carlo algorithm.The advantage of this approach is that sampling is performed before conditioning, rather than after. This enables the use of efficient samplers when the prior covariance matrix exhibits special properties, such as being Toeplitz, block-Toeplitz, or sparse. The performance of the proposed approach is evaluated in the context of nonparametric function estimation with large datasets.A numerical comparison with direct approaches based on Cholesky factorization is provided to illustrate its efficiency in terms of computational runtime
Orchestrating an interoperable sovereign federated Multi-vector Energy data space built on open standards and ready for GAia-X: D6.2 OMEGA-X Use case family implementation report
No full textThis report summarises the project impact evaluation results conducted within task 6.6 for the different pilot sites of the different use case families according to the Key Performance Indicators (KPIs) defined in deliverable OMEGA-X_D3.5. In addition, it describes the lessons learnt, best practises and recommendations obtained both from a technical and SSH (Social Sciences and Humanities) perspective
Orchestrating an interoperable sovereign federated Multi-vector Energy data space built on open standards and ready for GAia-XD4.4 Data ingestion, Common Information Model and semantic interoperability. Final version
No full text"D4.4 - Data ingestion, Common Information Model and semantic interoperability" is the second edition of OMEGA-X_D4.1 - Data ingestion, Common Information Model and semantic interoperability" [1]. Deliverable OMEGA-X_D4.4 documents the implementation for OMEGA-X federated infrastructure of the needed data adapters, tailored to the demo case requirements, and the agreement and implementation of a Common Information Model ruling the data exchange in the Data Space
Crystalline high aspect ratio nano-pillars generated by ultrafast Bessel beam from sapphire
No full textInternational audienceUltrafast lasers deliver high energy density through laser-matter interaction generating extreme pressures and temperatures. The energy deposited in the bulk of transparent materials can lead to the amorphization or re-crystallization of the modified matter [1]. The extreme conditions can even allow the observation of previously unreachable crystalline structures [2]
Apprendre à respecter les limites planétaires
No full textFestival des initiatives locales et positivesConférence publique donnée dans le cadre du « Festival des initiatives locales et positives » 2025 organisé par l'association "Groseille et ciboulette"
Adaptive finite-dimensional approximation of constrained Gaussian processes for large datasets
No full textIn the context of Bayesian shape-restricted function estimation, finite-dimensional Gaussian process (GP) approximations, such as those proposed in Maatouk and Bay (2017), have gained significant attention due to their flexibility and their ability to handle a wide range of shape constraints. However, these approaches have limitations in higher dimensions, as sampling relies on a regular discretization of the input space. Additionally, their computationalcomplexity grows cubically with the number of observations. In this paper, we propose a new adaptive low-rank methodology that enables finite-dimensional GP approximations to handle multidimensional input spaces, accommodate large datasets, and address multiple and complex shape constraints, even in extreme cases. This approach is based on iteratively adding optimal, non-uniform, one-dimensional discretization points and ordering the active input variables. An efficient Markov chain Monte Carlo (MCMC) method for sampling the full posterior distribution, with particular attention paid to large datasets n >> 10,000 and hyperparameter selection under shape constraints, is also developed. The proposed method significantly reduces computational complexity and mitigates the mass-shifting phenomenon observed in the posterior distribution, thereby substantially improving both prediction accuracy and uncertainty quantification. Its strong empirical performance is demonstrated on several synthetic and real-world datasets
L’aide à la décision pour les transitions énergétiques et environnementales des territoires et organisations
No full textEn effet, face aux enjeux environnementaux majeurs — crise climatique, raréfaction des ressources, érosion de la biodiversité — les transitions énergétiques et environnementales imposent une transformation des pratiques et appellent à repenser les processus de décision.Deux axes de recherche principaux structurent ces travaux de recherche :Les Meilleures Techniques Disponibles (MTD) : évaluation et amélioration des méthodes d’analyse des performances environnementales des sites industriels, avec une portée locale et internationale.La rénovation énergétique des bâtiments : développement d’outils, notamment l’Observatoire National des Bâtiments, pour mieux connaître le parc immobilier et identifier les gisements d’économie d’énergie à l’échelle territoriale.L’ensemble de ces recherches vise à outiller la décision pour accompagner les transitions, en tenant compte de la complexité des enjeux et de la diversité des acteurs impliqués