HAL-INSA Toulouse
Not a member yet
34325 research outputs found
Sort by
Optimizing excited states in quantum Monte Carlo: A reassessment of double excitations
No full text16 pages, 2 figuresInternational audienceQuantum Monte Carlo (QMC) methods have proven to be highly accurate for computing excited states, but the choice of optimization strategies for multiple states remains an active topic of investigation. In this work, we revisit the calculation of double excitation energies in nitroxyl, glyoxal, tetrazine, and cyclopentadienone, exploring different objective functionals and their impact on the accuracy and robustness of QMC. A previous study for these systems employed a penalty functional to enforce orthogonality among the states, but the chosen prefactors did not strictly ensure convergence to the target states. Here, we confirm the reliability of previous results by comparing excitation energies obtained with different functionals and analyzing their consistency. Additionally, we investigate the performance of different functionals when starting from a pre-collapsed excited state, providing insight into their ability to recover the target wave functions
Ground states of quasi-two-dimensional correlated systems via energy expansion
No full textInternational audienceWe introduce a generic method for computing groundstates that is applicable to a wide range of spatially anisotropic 2D many-body quantum systems. By representing the 2D system using a low-energy 1D basis set, we obtain an effective 1D Hamiltonian that only has quasi-local interactions, at the price of a large local Hilbert space. We apply our new method to three specific 2D systems of weakly coupled chains: hardcore bosons, a spin- Heisenberg Hamiltonian, and spinful fermions with repulsive interactions. In particular, we showcase a non-trivial application of the energy expansion framework, to the anisotropic triangular Heisenberg lattice, a highly challenging model related to 2D spin liquids. Treating lattices of unprecedented size, we provide evidence for the existence of a quasi-1D gapless spin liquid state in this system. We also demonstrate the energy expansion-framework to perform well where external validation is possible. For the fermionic benchmark in particular, we showcase the energy expansion-framework's ability to provide results of comparable quality at a small fraction of the resources required for previous computational efforts
Le rôle de l'auto-interaction dans la détection des défauts dans les semi-conducteurs irradiés
No full textInternational audienc
A machine learning and explainability-driven methodology for identifying winning strategies in Rugby Union
No full textInternational audienceInterest in predicting sports match outcomes has grown significantly, driven by advancements in machine learning techniques and widespread adoption. However, the utilization of these predictive models in enhancing tactical team performance remains relatively limited. We propose a methodology that combines machine learning and algorithm explainability techniques, which were demonstrated through a case study on Rugby Union. Our study unfolds in two phases: first, we identify the most suitable modeling approach for our data by establishing a prediction model based on performance indicators observed during games. Subsequently, we applied an analysis based on SHapley Additive exPlanations (SHAP) values to interpret the predictions of this model. Our findings serve three primary purposes: (i) from a global standpoint, identifying performance indicators that primarily determine match outcomes; (ii) from an aggregated point of view highlighting strengths and weaknesses of any given team; and (iii) from a local perspective, offering technical staff diagnostic analyses of past games
Thermodynamics of self-gravitating fermions as a robust theory for dark matter halos: Stability analysis applied to the Milky Way
No full textInternational audienceWe present a framework for dark matter (DM) halo formation based on a kinetic theory of self-gravitating fermions together with a solid connection to thermodynamics. Based on maximum entropy arguments, this approach predicts a most likely phase-space distribution which takes into account the Pauli exclusion principle, relativistic effects, and particle evaporation. The most general equilibrium configurations depend on the particle mass and develop a degenerate compact core embedded in a diluted halo, both linked by their fermionic nature. By applying such a theory to the Milky Way we analyze the stability of different families of equilibrium solutions with implications on the DM distribution and the mass of the DM candidate. We find that stable core-halo profiles, which explain the DM distribution in the Galaxy, exist only in the range . The lower bound is a consequence of imposing thermodynamical stability on the core-halo solutions having a quantum core mass alternative to the black hole hypothesis at the Galaxy center. The upper bound is solely an outcome of general relativity when the quantum core reaches the Oppenheimer-Volkoff limit and undergoes gravitational collapse towards a black hole. We demonstrate that there exists a set of stable core-halo profiles which are astrophysically relevant in the sense that their total mass is finite, do not suffer from the gravothermal catastrophe, and agree with observations. The morphology of the outer halo tail is described by a polytrope of index , developing a sharp decline of the density beyond in excellent agreement with the latest Gaia DR3 rotation curve data. Moreover, we obtain a total mass of about including baryons and a local DM density of about in line with recent independent estimates
Lien procédé-microstructure-propriétés de composites oxyde/oxyde élaborés par imprégnation de mèches en continu
No full textNational audienc
Oxygen Ingress in Titanium and Its Alloys After High-Temperature Oxidation: A Competition Between Strengthening and Embrittlement
No full textInternational audienceHigh-temperature oxidation of titanium leads to the formation of an external oxide scale and oxygen ingress into the metallic titanium material. Oxygen ingress can be significant due to the high solubility of O within Ti. An oxygen-rich layer (ORL) thus forms beneath the external oxide scale, exhibiting a brittle behavior. Microtensile specimens were used in order to exacerbate surface effects, i.e., surface reactivity in the case of the oxidation of titanium. Playing with the specimen thickness and pre-oxidation durations, it was possible to evaluate the evolution of tensile strength as well as the reduction in ductility for deep extensions of ORL relative to the specimen thickness (high fraction of ORL). In addition, ultrathin specimen extraction at different locations within the ORL depth aimed at better identifying the gradient of properties within the ORL. This micromechanical approach was applied to a commercially pure titanium (CP-Ti grade 2) and to a structural titanium alloy (Ti6242s). Both strengthening and loss of mechanical properties (yield strength and ductility) were observed depending on the material and oxygen ingress. While CP-Ti demonstrated an increase in mechanical strength up to ORL representing 80 pct of the gage section, Ti6242s experienced a loss of mechanical resistance even for the shortest exposure times (the ORL representing 10 pct of the gage section)
Optimisation bayésienne d'un réseau neuronal léger et précis pour la prédiction des performances aérodynamiques
No full textInternational audienceEnsuring high accuracy and efficiency of predictive models is paramount in the aerospace industry, particularly in the context of multidisciplinary design and optimization processes. These processes often require numerous evaluations of complex objective functions, which can be computationally expensive and time-consuming. To build efficient and accurate predictive models, we propose a new approach that leverages Bayesian Optimization (BO) to optimize the hyper-parameters of a lightweight and accurate Neural Network (NN) for aerodynamic performance prediction. To clearly describe the interplay between design variables, hierarchical and categorical kernels are used in the BO formulation. We demonstrate the efficiency of our approach through two comprehensive case studies, where the optimized NN significantly outperforms baseline models and other publicly available NNs in terms of accuracy and parameter efficiency. For the drag coefficient prediction task, the Mean Absolute Percentage Error (MAPE) of our optimized model drops from 0.1433% to 0.0163%, which is nearly an order of magnitude improvement over the baseline model. Additionally, our model achieves a MAPE of 0.82% on a benchmark aircraft self-noise prediction problem, significantly outperforming existing models (where their MAPE values are around 2 to 3%) while requiring less computational resources. The results highlight the potential of our framework to enhance the scalability and performance of NNs in large-scale MDO problems, offering a promising solution for the aerospace industry.Garantir la précision et l'efficacité des modèles prédictifs est primordial dans l'industrie aérospatiale, en particulier dans le contexte des processus de conception et d'optimisation multidisciplinaires. Ces processus nécessitent souvent de nombreuses évaluations de fonctions objectives complexes, ce qui peut être coûteux en temps et en argent. Pour construire des modèles prédictifs efficaces et précis, nous proposons une nouvelle approche qui s'appuie sur l'optimisation bayésienne (BO) pour optimiser les hyperparamètres d'un réseau neuronal (NN) léger et précis pour la prédiction des performances aérodynamiques. Pour décrire clairement l'interaction entre les variables de conception, des noyaux hiérarchiques et catégoriels sont utilisés dans la formulation BO. Nous démontrons l'efficacité de notre approche à travers deux études de cas complètes, où le réseau neuronal optimisé surpasse de manière significative les modèles de base et d'autres réseaux neuronaux publiquement disponibles en termes de précision et d'efficacité des paramètres. Pour la tâche de prédiction du coefficient de traînée, l'erreur absolue moyenne en pourcentage (MAPE) de notre modèle optimisé passe de 0,1433 % à 0,0163 %, ce qui représente une amélioration de près d'un ordre de grandeur par rapport au modèle de base. En outre, notre modèle atteint un MAPE de 0,82 % sur un problème de référence de prédiction du bruit propre d'un avion, ce qui est nettement supérieur aux modèles existants (dont les valeurs MAPE sont de l'ordre de 2 à 3 %) tout en nécessitant moins de ressources informatiques. Les résultats soulignent le potentiel de notre cadre pour améliorer l'évolutivité et la performance des NN dans les MDO à grande échelle
Porcine milk small extracellular vesicles modulate peripheral blood mononuclear cell proteome in vitro
No full textInternational audiencemall extracellular vesicles (EVs) are a subtype of nano-sized extracellular vesicles that mediate intercellular communication. EVs can be found in different body fluids, including milk. Monocytes internalize porcine milk EVs and modulate immune functions in vitro by decreasing their phagocytosis and chemotaxis while increasing their oxidative burst. This study aimed to assess the impact of porcine milk EVs on the porcine peripheral blood mononuclear cells (PBMC) proteome. Porcine PBMC were incubated with porcine milk EVs or medium as a control. Extracted proteins were then analyzed using nano-LC-MS/MS. A total of 1584 proteins were identified. The supervised multivariate statistical analysis, sparse variant partial least squares - discriminant analysis (sPLS-DA) for paired data identified discriminant proteins (DP) that contributed to a clear separation between the porcine milk EVs treated cells and control groups. A total of 384 DP from both components were selected. Gene Ontology (GO) enrichment analysis with ProteINSIDE provided the evidence that the DP with a higher abundance in porcine milk EVs, like TLR2, APOE, CD36, MFGE8, were mainly involved in innate immunity and the process of EVs uptake processes. These results provide a proteomics background to the immunomodulatory activity of porcine milk EVs and to the potential mechanisms used by immune cells to internalize them
Simulation du LiDAR, des Véhicules Terrestres Autonomes et des Réseaux Neuronaux pour L'estimation de la Surface Foliaire dans les Vergers
No full textInternational audienceThe leaf area index (LAI) is vital for assessing plant photosynthetic activity, crucial for optimising orchard management. This study presents a method to estimate leaf area density (LAD) variations and tree LAI using LiDAR data from unmanned ground vehicles (UGVs). Combining 3D tree reconstruction with neural network-based analysis of LiDAR penetration descriptors, the approach effectively estimates canopy parameters. The method was validated through simulation using diverse 3D canopy models, achieving performance metrics: RMSE: 0.2 m²/m³, R²: 0.95 for LAD and RMSE: 0.17 m²/m², R²: 0.84 for LAI. Results confirm the potential of LiDAR-based systems for precise orchard canopy monitoring.El índice de área foliar (LAI) es fundamental para evaluar la actividad fotosintética de las plantas, un parámetro clave para la optimización de la gestión de huertos. Este estudio presenta un método para estimar las variaciones de la densidad de área foliar (LAD) y el LAI de los árboles a partir de datos LiDAR adquiridos por vehículos terrestres autónomos (UGV). Al combinar la reconstrucción 3D de los árboles con un análisis de descriptores de penetración LiDAR mediante redes neuronales, el enfoque permite una estimación eficaz de los parámetros del dosel. La metodología fue validada mediante simulaciones con diversos modelos 3D de copas, obteniendo métricas de rendimiento: RMSE: 0,2 m²/m³, R²: 0,95 para LAD y RMSE: 0,17 m²/m², R²: 0,84 para LAI. Los resultados confirman el potencial de los sistemas basados en LiDAR para el monitoreo preciso del dosel en huertos.L'indice de surface foliaire (LAI) est essentiel pour évaluer l'activité photosynthétique des plantes, un paramètre clé pour l'optimisation de la gestion des vergers. Cette étude propose une méthode d'estimation des variations de la densité de surface foliaire (LAD) et du LAI des arbres à partir de données LiDAR acquises par des véhicules terrestres autonomes (UGV). En combinant la reconstruction 3D des arbres avec une analyse des descripteurs de pénétration LiDAR par réseaux neuronaux, l'approche permet une estimation efficace des paramètres du couvert végétal. La méthode a été validée par simulation sur divers modèles 3D de canopées, obtenant des performances de RMSE : 0,2 m²/m³, R² : 0,95 pour le LAD et RMSE : 0,17 m²/m², R² : 0,84 pour le LAI. Les résultats confirment le potentiel des systèmes basés sur le LiDAR pour un suivi précis du couvert végétal des vergers