HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Hybrid high-order methods for elasto-acoustic wave propagation in the time domain
No full textInternational audienceWe devise a Hybrid High-Order (HHO) method for the coupling between the acoustic and elastic wave equations in the time domain. A first-order formulation in time is considered. The HHO method can use equal-order and mixed-order settings with polynomial degree k >= 0 for the face unknowns, together with O(1)-and O(1/h)-stabilizations. An energy-error estimate is established in the time-continuous case. A numerical spectral analysis is performed, showing that O(1)-stabilization is required to avoid excessive CFL limitations for explicit time discretizations. Moreover, the spectral radius of the stiffness matrix is fairly independent of the geometry of the mesh cells. For analytical solutions on general meshes, optimal convergence rates of order (k + 1) are shown in both equal-and mixed-order settings using O(1)-stabilization, whereas order (k + 2) is achieved in the mixed-order setting using O(1/h)-stabilization. Test cases with a Ricker wavelet as an initial condition showcase the relevance of the proposed method for the simulation of elasto-acoustic wave propagation across media with contrasted material properties
UNEM: UNrolled Generalized EM for Transductive Few-Shot Learning
No full textInternational audienceTransductive few-shot learning has recently triggered wide attention in computer vision. Yet, current methods introduce key hyper-parameters, which control the prediction statistics of the test batches, such as the level of class balance, affecting performances significantly. Such hyper-parameters are empirically grid-searched over validation data, and their configurations may vary substantially with the target dataset and pre-training model, making such empirical searches both sub-optimal and computationally intractable. In this work, we advocate and introduce the unrolling paradigm, also referred to as "learning to optimize", in the context of few-shot learning, thereby learning efficiently and effectively a set of optimized hyperparameters. Specifically, we unroll a generalization of the ubiquitous Expectation-Maximization (EM) optimizer into a neural network architecture, mapping each of its iterates to a layer and learning a set of key hyper-parameters over validation data. Our unrolling approach covers various statistical feature distributions and pre-training paradigms, including recent foundational vision-language models and standard vision-only classifiers. We report comprehensive experiments, which cover a breadth of fine-grained downstream image classification tasks, showing significant gains brought by the proposed unrolled EM algorithm over iterative variants. The achieved improvements reach up to 10% and 7.5% on vision-only and vision-language benchmarks, respectively. The source code and learned parameters are available at https://github.com/ZhouLong0/ UNEM-Transductive
Long run convergence of discrete-time interacting particle systems of the McKean-Vlasov type
No full textInternational audienceWe consider a discrete-time system of n coupled random vectors, a.k.a. interacting particles. The dynamics involve a vanishing step size, some random centered perturbations, and a mean vector field which induces the coupling between the particles. We study the doubly asymptotic regime where both the number of iterations and the number n of particles tend to infinity, without any constraint on the relative rates of convergence of these two parameters. We establish that the empirical measure of the interpolated trajectories of the particles converges in probability, in an ergodic sense, to the set of recurrent Mc-Kean-Vlasov distributions. A first application example is the granular media equation, where the particles are shown to converge to a critical point of the Helmholtz energy. A second example is the convergence of stochastic gradient descent to the global minimizer of the risk, in a wide two-layer neural networks using random features
A standardized population-based approach for food consumption analysis and socio-metabolic research: insights from its application to France
No full textFood consumption has a significant urban environmental impact and is progressively being considered on par with mobility and housing, all three of which involve different types of population. Despite preliminary work, there is a lack of standardized methods to quantify different populations' contribution to urban metabolism. A comprehensive framework is provided in this paper to estimate the populations present in an area and their food consumption there, over a given period of time, based on broadly available census and statistical data. Drawing on population equivalents, a standardized and consistent approach is used to integrate the different types of population and their respective time spent in the area, as well as their share of daily food intake. A case study of France shows that, in 85% of municipalities and 75% of inter-municipal cooperation areas, the present population is smaller than the resident population. The eating population is usually 5 to 20% smaller than the resident population, because children, teenagers, and commuters consume less food overall. While metropolitan areas attract populations, this is compensated by the resident population leaving for vacation, leading to close to zero or negative changes compared to the resident population. Only touristic areas attract additional populations to such an extent that the present population becomes substantially larger than the resident population over a year. Previous approaches tend to overestimate the biomass directly associated with urban food consumption. Our population-based framework can help reduce uncertainties about biomass use in cities and improve the comparison of material consumption between different areas
Lessons for Interactive Theorem Proving Researchers from a Survey of Coq Users
No full textInternational audienceThe Coq Community Survey 2022 was an online public survey of users of the Coq proof assistant conducted during February 2022. Broadly, the survey asked about use of Coq features, user interfaces, libraries, plugins, and tools, views on renaming Coq and Coq improvements, and also demographic data such as education and experience with Coq and other proof assistants and programming languages. The survey received 466 submitted responses, making it the largest survey of users of an interactive theorem prover (ITP) so far. We present the design of the survey, a summary of key results, and analysis of answers relevant to ITP technology development and usage. In particular, we analyze user characteristics associated with adoption of tools and libraries and make comparisons to adjacent software communities. Notably, we find that experience has significant impact on Coq user behavior, including on usage of tools, libraries, and integrated development environments (IDEs)
Ergodic properties of concurrent systems
No full textInternational audienceA concurrent system is defined as a monoid action of a trace monoid on a finite set of states. Concurrent systems represent state models where the state is distributed and where state changes are local. Starting from a spectral property on the combinatorics of concurrent systems, we prove the existence and uniqueness of a Markov measure on the space of infinite trajectories relatively to any weight distributions. In turn, we obtain a combinatorial result by proving that the kernel of the associated Möbius matrix has dimension 1; the Möbius matrix extends in this context the Möbius polynomial of a trace monoid. We study ergodic properties of irreducible concurrent systems and we prove a Strong law of large numbers. It allows us to introduce the speedup as a measurement of the average amount of concurrency within infinite trajectories. Examples are studied
Assessing the thermohaline coherence of mesoscale eddies as described from in situ data
No full textInternational audienceAbstract. Mesoscale eddies are a ubiquitous feature of the global ocean. According to Lagrangian theory, these eddies often transport a distinct water mass within their cores, making them materially coherent. This study aims to determine if such a distinct water mass exists in eddy cores, thereby verifying their material coherence using in situ data, despite the lack of temporal continuity. We introduce the term “thermohaline coherence” to describe this approach. Identifying such a water mass would signal Lagrangian transport from the eddy formation region. We analyzed the water masses at the cores of various eddies sampled during eight research cruises using high-resolution data (approximately 20 km horizontally and 10 m vertically). We revisited coherence definitions and checked data accuracy. Comparing the horizontal positions of these core anomalies with eddy surface signatures revealed that surface data alone are insufficient for characterizing the eddy material coherence. To calculate eddy volumes, we compare thermohaline anomalies with other criteria, and we present two methods for extrapolating eddy volumes from a single hydrographic section. The results show that the outermost closed contour of the Brunt–Väisälä frequency anomaly at each depth provides a reliable approximation for the eddy boundary
State of the art and taxonomy survey on federated learning and blockchain integration in UAV applications
No full textInternational audienceUnmanned aerial vehicles (UAVs) have recently transformed the current landscape into a smart ecosystem, which requires higher autonomy and operational efficiency. However, technical and resource management issues present considerable challenges. Emerging technologies, such as federated learning and blockchain, provide promising solutions to security, privacy, and resource management issues. Numerous research projects have addressed the challenges faced by UAVs, including investigations into federated learning and blockchain technology. Our study outlines blockchain and federated learning solutions for UAVs with a complete taxonomy, including architecture, challenges, and applications of each technology. The combination of federated learning with blockchain technology in UAV applications harmonizes FL’s decentralized model training and blockchain’s secure ledger, addressing issues such as scalability and security. These technologies will benefit factories, cities, logistics, healthcare, and disaster monitoring. Future directions provide guidance for advancing further study
Matérialité et gouvernance du système sociotechnique très haut débit à Dakar (Sénégal) : vision stratégique de l’infrastructure numérique, planification métropolitaine et chantiers de construction
No full textThis thesis analyzes the development of a very high-speed socio-technical system (fiber optic network and data centers) and the urban transformations of a West African metropolis, questioning the conditions and modalities of a “digital urbanism” in Dakar (Senegal). The aim is to understand how digital infrastructure and urban spaces shape each other, and to propose a reading based on materiality rather than overhanging an inert materiality-support.Drawing on work from urban studies and social approaches to technology inspired by Science and Technology Studies (STS), the investigation focuses on the socio-political, socio-technical and territorial dimensions of very-high-speed infrastructure deployment, which it analyzes at three scales, corresponding to distinct spheres of action and temporalities, and in different types of intra-urban spaces. Using a qualitative methodology (interviews, participant observation, mapping), the thesis demonstrates the ambivalent relationship between the privatized development of the very-high-speed socio-technical system and socio-spatial inequalities. On a national and metropolitan scale, it highlights the links between urban planners and powerful international telecoms operators, and their respective influence on metropolitan digital planning in a context marked by neoliberal policies. At the street level, it analyzes the role of network engineers and technicians, highlighting the role of material contingencies, informality and socio-technical “bricolage”, as well as the role of “street-level” technical expertise in the material development of very high-speed infrastructure, through a study of construction sites and “on-the-spot” work. Analysis of the governance of deployment, coupled with that of urban materiality and metropolitan socio-spatial evolutions, has enabled us to qualify a digital urbanism that differs according to urban profiles. However, while insisting on the decisive weight of telecoms operators and the insufficient prescription and regulation capacity of public authorities, the thesis identifies self-regulation mechanisms that are sometimes beneficial to the general interest, and partial catch-up mechanisms in certain less “attractive” neighborhoods. These results lead us to qualify the effects of the “privatization” of digital development on urban fragmentation, and suggest elements of reflection for the public reengagement of a thought and practice of urban digital planning in DakarCette thèse analyse le développement d’un système sociotechnique très haut débit (réseau de fibre optique et data centers) et les transformations urbaines d’une métropole ouest-africaines en interrogeant les conditions et modalités d’une « urbanisme du numérique » à Dakar (Sénégal). L’ambition est de comprendre comment infrastructure numérique et espaces urbains se façonnent mutuellement et d’en proposer une lecture à partir de la matérialité plutôt que surplombant une matérialité-support inerte. S’appuyant sur des travaux issus des études urbaines et des approches sociales de la technique inspirées des Science and Technology Studies (STS), l’enquête porte sur les dimensions sociopolitiques, sociotechniques et territoriales du déploiement de l’infrastructure très haut débit qu’elle analyse à trois échelles, correspondant à des sphères d’action et des temporalités distinctes, et dans différents types d’espaces intra-urbains. Avec une méthodologique qualitative (entretiens, observation participante, cartographie), la thèse démontre les relations ambivalentes qu'entretiennent le développement privatisé du système sociotechnique très haut débit et les inégalités socio-spatiales. Aux échelles nationales et métropolitaines, elle met en évidence les liens entre les acteurs de l’urbanisme et de puissants opérateurs de télécommunication internationaux ainsi que leur influence respective dans l’aménagement numérique métropolitain dans un contexte marqué par des politiques néolibérales. A l’échelle de la rue, elle analyse le rôle des ingénieurs et techniciens réseaux en soulignant, à travers une étude des chantiers et du travail « en situation », le rôle des contingences matérielles, de l’informalité et du « bricolage » sociotechnique mais aussi celui d’une expertise technique « de la rue » essentielle dans le développement matériel de l’infrastructure très haut débit. L’analyse de la gouvernance du déploiement couplée à celle de la matérialité urbaine et des évolutions socio-spatiales métropolitaines a permis de qualifier un urbanisme numérique différencié selon les profils urbains. Cependant, tout en insistant sur le poids déterminant des opérateurs de télécommunication et l’insuffisante capacité de prescription et de régulation des pouvoirs publics, la thèse identifie des mécanismes d’autorégulation parfois bénéfiques à l’intérêt général et des mécanismes de rattrapage partiel dans certains quartiers moins « attractifs ». Ces résultats conduisent à nuancer les effets de la « privatisation » du développement du numérique sur la fragmentation urbaine et suggèrent des éléments de réflexion pour le réengagement public d’une pensée et d’une pratique de l’aménagement numérique urbain à Daka
Y a-t-il un progrès en art ?
No full textThis article reflects on major artistic movements throughout history, highlighting certain similarities and differences with philosophical and economic thinking on progress. We first focus on cases where art has developed alongside technical advances, using the example of perspective as discussed in the works of Erwin Panofsky. We then focus on two situations considered ‘unsurpassable’: classical Greek sculpture and its curious realism, and the modernism in architecture advocated by Le Corbusier in the 1920s, which continues to this day. This, of course, only partially illuminates the big question of technoscientific progress, but it does highlight the influence that progress has had on some of the most recognised artistic movements. The article seeks to highlight the relationship between art and cultural experience as a source of living meaning.Cet article est une réflexion sur les grands mouvements artistiques à l'échelle historique, pour faire ressortir certaines analogies et différences avec la pensée philosophique et économique sur le progrès. Nous portons l'attention tout d'abord sur le cas où un art côtoie un développement technique avec l'exemple de la perspective abordé dans les travaux d'Erwin Panofsky. Puis nous insistons sur deux situations pensées comme "indépassables" qui sont d'une part le cas de la sculpture en Grèce classique et son curieux réalisme, d'autre part le cas du modernisme en architecture prôné par Le Corbusier dans les années 1920 et qui perdure encore aujourd'hui. Ceci, naturellement, n'éclaire que partiellement la grande question du progrès technoscientifique, mais souligne l'influence que celui-ci a eu sur des courants artistiques parmi les plus reconnus. L'article tend à valoriser la relation de l'art avec le vécu culturel porteur de signifiés vivants.</div