HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Knowledge Sharing and Cumulative Innovation in Business Networks
No full textInternational audienceHow can we explain the success of cooperative networks of firms which share innovations, such as Silicon Valley or the Open Source community? This paper shows that if innovations are cumulative, making an invention publicly available to a network of firms may be valuable if the firm expects to benefit from future improvements made by other firms. A cooperative equilibrium where all innovations are made public is shown to exist under certain conditions. Furthermore, such an equilibrium does not rest on punishment strategies being followed after a deviation: it is optimal not to deviate regardless of another firm’s actions following a deviation. A cooperative equilibrium is more likely to arise the greater the number of firms in the network. When R&D effort is endogenous, cooperative equilibria are associated with strategic complementarities between firms’ research effort, which may lead to multiple equilibria
Is OpenAlex a revolution or a challenge for bibliometrics/bibliometricians?
No full textInternational audienceThere has been one announcement after another about OpenAlex, a huge bibliographic database of 248 million scientific works whose metadata is distributed under a CC0 licence, accessible via the online interface and an API. At the end of 2023, Sorbonne Université decided not to renew its subscription to the Web of Science, stating that it was not just a question of saving money but of "resolutely freeing itself from proprietary bibliometric products" and using OpenAlex. In January 2024, the CNRS announced that it would be unsubscribing from Scopus (while maintaining its subscription to WoS) and would be "gradually switching to free bibliographic tools that are compatible with the organisation's open science policy", again mentioning OpenAlex as one of the possible alternatives. Finally, in February 2024, the MESR announced a multi-year partnership with OpenAlex, with financial support and no doubt human resources to contribute "to the improvement of OpenAlex's general data and by helping to enrich in particular the data linked to French research".In view of this enthusiasm, and before signing up for the sole reason of opening up science, we felt it necessary to assess the relevance of using OpenAlex exclusively to carry out a count of publications at the level of an institution, a classic procedure when you then want to carry out a more detailed bibliometric analysis on the basis of advanced indicators.This post presents the 2 tests we carried out to answer 2 questions.Question 1: What proportion of publications already listed (in WoS, Scopus or HAL) by Ecole des Ponts can be found in OpenAlex? Question 2: What proportion of publications is correctly retrieved by OpenAlex for a query on Ecole des Ponts
Promouvoir des modèles d’intelligence artificielle frugale pour et par les politiques publiques
No full textThe development of artificial intelligence (AI) is a priority for the French government who has enforced a national strategy for AI since 2018. Developing AI tackles various challenges in terms of research, economy, modernisation of public administration, regulation and ethics.The necessary ecological and energy transitions make us question our uses, our resource consumption and more generally the environmental impact of our activities in different areas, including AI. Over the past few years, AI has soared and spawned bigger and bigger models, fed with larger and larger datasets, capable of running increasingly complex calculations to achieve high levels of performance, prediction and precision. However, we wish to quit this ever growing tendency as we are aware of the climate emergency and we share growing concerns regarding the environment. AI development should be balanced between increased performances and environmental impact. In France, recent AI developments pave the way to integrating “frugality” into AI systems. It consists in optimising the volume of data sets used for training and inference as well as the internal architecture of the AI algorithms, choosing hardware and its source of energy in order to reduce the total environmental impact of AI, while maintaining an acceptable level of performance.Our study is based on a bibliographic research (scientific works and administrative documents mostly) and on interviews with AI stakeholders. We address various features of frugal AI, including its definition, different interpretations by the interviewees, how public administration is embracing this new concept and eventually how public research could play a part in disseminating new practices.Le développement de l’intelligence artificielle (IA) est une priorité pour la France qui a instauré une stratégie nationale pour l’IA depuis 2018. Ce développement correspond à divers enjeux sur le plan de la recherche, de l’économie, de la modernisation de l’action publique, de la régulation et de l’éthique.Le contexte de la transition écologique et de la transition énergétique nous interroge sur nos usages, sur la consommation des ressources et de manière globale sur l’impact environnemental de nos activités dans différents domaines, dont l’IA. Le domaine de l’IA s’est notamment développé ces dernières années sur un modèle de données toujours plus volumineuses et de calculs de plus en plus complexes pour atteindre des niveaux de performance élevés. Ce modèle montre ses limites dans un contexte où l’urgence climatique et les préoccupations environnementales grandissantes nous conduisent à rechercher des compromis entre performance et poids environnemental. C’est ainsi que les développements actuels en France conduisent à intégrer une dimension de frugalité dans le domaine de l’IA, en termes d’optimisation des quantités de données utilisées, d’architecture des algorithmes, de choix de matériel, ou de source d’énergie utilisée, afin de réduire l’impact environnemental de l’IA tout en conservant un niveau de performance acceptable.Sur la base d’une analyse bibliographie (publications, appels à projets, littérature grise) et d’entretiens avec les acteurs du domaine de l’IA, cette étude aborde divers aspects de l’IA frugale, dont sa définition, les différentes interprétations observées, la manière dont l’administration s’approprie ce concept et enfin le rôle que la recherche publique peut tenir pour participer à l’intégration de l’IA frugale dans nos usages
Concrete printing defects dataset for quality process monitoring using computer vision
No full textPublic datasetThe dataset contains images of concrete printing process monitoring camera. The dataset images are completed by the annotations dedicated to the detection of defects and anomalies that can be used in the close-loop control as feedback to an automated quality assessment system. It could stem from a rule-based expert system, vector-to-action dictionaries, supervised machine learning models, etc
Maximal Martingale Wasserstein Inequality
No full textInternational audienceIn this note, we complete the analysis of the Martingale Wasserstein Inequality started in [5] by checking that this inequality fails in dimension d ≥ 2 when the integrability parameter ρ belongs to [1, 2) while a stronger Maximal Martingale Wasserstein Inequality holds whatever the dimension d when ρ ≥ 2
A benchmark study on reactive two-phase flow in porous media : Part II -results and discussion
No full textInternational audienceThis paper presents and discusses the results obtained by the participants to the benchmark described in deHoop et al, Comput. Geosci. (2023). The benchmark uses a model for CO geological storage and focuses on the coupling between two-phase flow and geochemistry. Several test cases of various levels of difficulty are proposed, both in one and two spatial dimensions. Six teams participated in the benchmark, each with their own simulation code, though not all teams attempted all the cases. The codes used by the participants are described, and the results obtained on the various test cases are compared, as well as the performance of the codes. It is shown that the results obtained are widely consistent, giving a good level of confidence in the outcome of the benchmark. The general complexity of two-phase flow coupled with chemical reactions altering porous media means that some differences between the codes remain. Besides, from the convergence study, it is clear that the two-dimensional problem has a relatively high sensitivity to a spatial resolution which adds to the complexity
Non-convex coercive Hamilton-Jacobi equations: Guerand's relaxation revisited
No full textInternational audienceThis work is concerned with Hamilton-Jacobi equations of evolution type posed in domains and supplemented with boundary conditions. Hamiltonians are coercive but are neither convex nor quasiconvex. We analyse boundary conditions when understood in the sense of viscosity solutions. This analysis is based on the study of boundary conditions of evolution type. More precisely, we give a new formula for the relaxed boundary conditions derived by J. Guerand (J. Differ. Equations, 2017). This new point of view unveils a connection between the relaxation operator and the classical Godunov flux from the theory of conservation laws. We apply our methods to two classical boundary value problems. It is shown that the relaxed Neumann boundary condition is expressed in terms of Godunov's flux while the relaxed Dirichlet boundary condition reduces to an obstacle problem at the boundary associated with the lower non-increasing envelope of the Hamiltonian
Load-finding: another look at Maxwell's theorem and gridshell form-finding integrating construction constraints
No full textInternational audienceGridshells and shells are geometrical objects that must comply with mechanical and manufacturing constraints. Geometry and mechanics are related by the Maxwell's theorem. The form diagram can be "lifted" by various methods by interpreting the dual force diagram as the balance of horizontal forces under vertical loading. Several existing methods start with a planar grid and a given load to achieve balanced shapes. It is proposed in this paper to reverse the procedure in a kind of "load-finding strategy". Considering indeed that the behaviour of lightweight structures is not dominated by a single loading, one can afford to approximate the load used for the design. We hence use lifting methods with strict control over geometrical properties and we prescribe heights and forces at the boundary to calculate the load under which the structure operates without bending, i.e. the funicular load associated with the structure. We can then optimise the lift to find the load which will be the closest to a target distribution: null load will give self-stressed structures while uniform distributed vertical load will give Gaudi like structures. We hence get a space of planar faced polyhedron in equilibrium under a certain load that projects onto the input mesh, and which is controlled by heights and forces defined on guide curves. Moreover we demonstrate that the difference between the funicular load and the applied load is a powerful indicator for structural design: it helps the structural analysis and the development of a design strategy
