HAL Descartes
Not a member yet
3165288 research outputs found
Sort by
À la rentrée 2022, la taille des classes continue de diminuer dans le premier degré
No full textChaque note d’information présente les résultats les plus récents issus des exploitations d’enquêtes et d’études statistiques. Cette publication apporte l’éclairage de la direction de l’évaluation, de la prospective et de la performance (DEPP) à la compréhension et à l’analyse du système éducatif français. Elle est disponible en ligne, sous la forme de fichiers imprimables à télécharger, accompagnés de tableaux de données chiffrées et de graphiques.À la rentrée 2022, la taille des classes du premier degré continue de diminuer dans les écoles publiques, que ce soit dans les classes de niveau préélémentaire ou élémentaire.Cette tendance est particulièrement marquée depuis 2017, suite à la mise en oeuvre des dédoublements de classes en éducation prioritaire.Ces mesures, qui ont d’abord concerné les classes de CP-CE1, et depuis la rentrée 2020 les grandes sections de maternelle, n’ont pas retiré de moyens aux autres niveaux.Elles n’ont pas non plus été réalisées aux dépens des écoles rurales ou urbaines hors éducation prioritaire, dont le nombre d’élèves par classe diminue également.Par ailleurs, le plafonnement à 24 du nombre d’élèves par classe en grande section, CP et CE1 sur l’ensemble du territoire, amorcé en 2020, contribue aussi à alléger la taille des classes
Buddhist economics as a return to rational model of economic management
No full textInternational audienceThe concept of Buddhist economics is gaining increased appeal in a world where external factors are once again becoming more of a threat than a salvation. Buddhist economy is a return to the values of agricultural production, but taking into account the experience and achievements of the industrial and post-industrial economy. Care for the environment, personal development, community development, especially spiritual development – these are the priorities of the Buddhist economy. In particular, agricultural production appears as only the most convenient means for achieving these goals. However, Buddhist economics is not a rejection of the achievements of modern and postmodern society – it is an attempt to use these experiences and achievements for a more intelligent and effective implementation of the goals of the economy, which were defined by Aristotle. The rational model of economic management according to these views consists in thrifty but full consumption and restrained production with environmentally friendly aims
Benchmarking Bayesian neural networks and evaluation metrics for regression tasks
No full textDue to the growing adoption of deep neural networks in many fields of science and engineering, modeling and estimating their uncertainties has become of primary importance. Despite the growing literature about uncertainty quantification in deep learning, the quality of the uncertainty estimates remains an open question. In this work, we assess for the first time the performance of several approximation methods for Bayesian neural networks on regression tasks by evaluating the quality of the confidence regions with several coverage metrics. The selected algorithms are also compared in terms of predictivity, kernelized Stein discrepancy and maximum mean discrepancy with respect to a reference posterior in both weight and function space. Our findings show that (i) some algorithms have excellent predictive performance but tend to largely over or underestimate uncertainties (ii) it is possible to achieve good accuracy and a given target coverage with finely tuned hyperparameters and (iii) the promising kernel Stein discrepancy cannot be exclusively relied on to assess the posterior approximation. As a by-product of this benchmark, we also compute and visualize the similarity of all algorithms and corresponding hyperparameters: interestingly we identify a few clusters of algorithms with similar behavior in weight space, giving new insights on how they explore the posterior distribution
Attribution des défaillances de livraison des achats en ligne : le transporteur toujours fautif aux yeux du consommateur ?
No full textInternational audienceAs e-commerce grows, consumers make massive use of home delivery and regularly experience delivery incidents. How do they react? In the absence of information on the real causes of the failure, who do they spontaneously blame: the e-merchant or the carrier? From whom do they expect redress? Are their representations influenced by their level of familiarity with these situations of purchase and service failure, or by the perceived typicality and severity of the incident? Rooted in attribution theory, a questionnaire survey of 118 people is conducted to answer these questions. It provides results that invite last-mile carriers to assume greater visibility in e-commerce supply chains and to engage in relationship marketing approaches to consumers.Alors que le e-commerce se développe de plus en plus, les consommateurs recourent massivement à la livraison à domicile et subissent régulièrement des incidents de livraison. Comment y réagissent-ils ? En l’absence d’information sur les causes réelles de la défaillance, qui incriminent-ils spontanément : le e-commerçant ou le transporteur ? De qui attendent-ils réparation ? Leurs représentations sont-elles influencées par leur niveau de familiarité avec ces situations d’achat et d’échec de service, ou par le niveau de typicité et de sévérité perçues de l’incident ? Ancrée dans la théorie de l’attribution, une enquête par questionnaire, réalisée auprès de 118 personnes, est menée pour répondre à ces questions. Elle fournit des résultats qui invitent notamment les transporteurs du dernier kilomètre à assumer une plus grande visibilité dans les chaînes logistiques du e-commerce et à s’engager dans des démarches de marketing relationnel vis-à-vis des consommateurs
Baseline Method for the Sport Task of MediaEval 2022 with 3D CNNs using Attention Mechanisms
No full textInternational audienceThis paper presents the baseline method proposed for the Sports Video task part of the MediaEval 2022 benchmark. This task proposes two subtasks: stroke classification from trimmed videos, and stroke detection from untrimmed videos. This baseline addresses both subtasks. We propose two types of 3D-CNN architectures to solve the two subtasks. Both 3D-CNNs use Spatio-temporal convolutions and attention mechanisms. The architectures and the training process are tailored to solve the addressed subtask. This baseline method is shared publicly online to help the participants in their investigation and alleviate eventually some aspects of the task such as video processing, training method, evaluation and submission routine. The baseline method reaches 86.4% of accuracy with our v2 model for the classification subtask. For the detection subtask, the baseline reaches a mAP of 0.131 and IoU of 0.515 with our v1 model
Quantitative Morphometric 3D Terrain Analysis of Japan Using Scripts of GMT and R
No full textInternational audienceIn this paper, we describe two related scripting methods of cartographic data processing and visualization that provide 2D and 3D mapping of Japan with different algorithm complexity. The first algorithm utilizes Generic Mapping Toolset (GMT), which is known as an advanced console-based program for spatial data processing. The modules of GMT combine the functionality of scripting with the aspects of geoinformatics, which is especially effective for the rapid analysis of large geospatial datasets, multi-format data processing, and mapping in 2D and 3D modes. The second algorithm presents the use of the R programming language for cartographic visualization and spatial analysis. This R method utilizes the packages ‘tmap’, ‘raster’, ‘maps’, and ‘mapdata’ to model the morphometric elements of the Japanese archipelago, such as slope, aspect, hillshade and elevation. The general purpose graphical package ‘ggplot2’ of R was used for mapping the prefectures of Japan. The two scripting approaches demonstrated an established correspondence between the programming languages and cartography determined with the use of scripts for data processing. They outperform several well-known and state-of-the-art GIS methods for mapping due to their high automation of data processing. Cartography has largely reflected recent advances in data science, the rapid development of scripting languages, and transfer in the approaches of data processing. This extends to the shift from the traditional GIS to programming languages. As a response to these new challenges, we demonstrated in this paper the advantages of using scripts in mapping, which consist of repeatability and the flexible applicability of scripts in similar works
Light Can Always Escape Black Holes and This Is Why They Are Observed to Be Bright
No full textThe Schwarzschild metric in general relativity predicts that black holes exist in which the mass and energy inside the Schwarzschild event-horizon is collapsing into a singularity in their very center. We have not observed the inside of any black holes directly; all information about them is from observations from the outside. There is, however, one exception, as the Hubble sphere fulfills several of the mathematical properties related to a black hole, as pointed out in the literature. The Hubble radius is close to, or identical to, its Schwarzschild radius. We are on the inside of the Hubble sphere, so it has even been suggested in the literature that we live inside a black hole. However, clearly the universe we observe around us is nothing like the predictions from the Schwarzschild metric on black holes; namely, that everything inside collapses to a singularity. We show that if one includes Lorenz relativistic mass in the gravitational model, then one gets predictions suggesting that inside "black-holes" the mass (energy) has to be distributed around and cannot have collapsed into the center singularity. Incorporating Lorentz relativistic mass leads to a quite di↵erent interpretation of black holes, both from the inside and outside, that seems to be closer to what we have actually observed. We also show that photons can likely escape black holes, but nothing with rest-mass can escape these objects. This seems to be in line with observations of directly observed object where we can expect to find black holes, such as in quasars and in galactic centers, but which are very bright and not black as predicted by standard theory. We also demonstrate the close connection between the Hubble sphere and black holes. We conjecture that the big bang theory mainly developed from a misintepretation of comsological redshift, which could be simply a special type of of gravitational redshift
Intelligence artificielle et intuition: Les algorithmes d’apprentissage profondcomme occasion de décrire l’intuition
No full textSince the years 2010, and thanks to a number of feats achieved by connectionist artificial intelligence (like AlphaGo’s victory), the notion of intuition, which had been put aside by the symbolic approach, has recovered a new legitimacy. According to their inventors, deep learning algorithms manifest intuition. In order to assess this claim, this thesis describes deep learning – its history, operating, and usage –, highlights how it simulates some human abilities (perception, induction, imagination) and provides a definition of artificial intelligence. As it pretends to materialize intelligence, and more specifically intuition and creativity, the project of artificial intelligence steps on the toes of philosophers and constrains them to reopen the question of mind genesis : would gathering the adequate material conditions be enough in order for intelligence to arise, or are there some missing « ingredients », of which intuition is the name ? Drawing on the history of philosophy, like the debate between Descartes and Leibniz about formalism, as well as discussions of contemporary thinkers (Bachelard, Bateson, Caillois, Derrida, Latour, Rosset), AI founders (McCarthy, Turing), and deep learning inventors (LeCun, Hinton), this thesis offers a description of intuition and of some features (comprehension, reflexivity, invention) that seem reluctant to simulation. Finally, it examines the notion of randomness in order to assess the possibility, for a machine, to originate new forms.Grâce aux succès techniques remportés par l’école connexionniste dans les années 2010, dont la victoire d’AlphaGo, la notion d’intuition, longtemps mise à l’écart par l’école symbolique, a repris droit de cité dans le champ de l’intelligence artificielle (IA). Selon leurs inventeurs, les algorithmes dits « d’apprentissage profond » (deep learning) font preuve d’intuition. Afin d’apprécier la pertinence de cette affirmation, cette thèse présente les réseaux de neurones d’apprentissage profond – leur histoire, leur fonctionnement, leur usage –, et s’arrête sur leur manière de simuler certaines de nos facultés (perception, induction, imagination), puis propose une définition du projet d’intelligence artificielle. En aspirant à matérialiser l’intelligence, et plus précisément l’intuition et la créativité, le projet d’intelligence artificielle s’immisce dans le champ de la philosophie et invite à rouvrir la question de la genèse de la pensée : suffit-il de réunir certaines conditions matérielles pour susciter l’intelligence, ou bien d’autres « ingrédients », dont l’intuition serait le nom, sont nécessaires ? En s’appuyant sur l’histoire de la philosophie (le différend entre Descartes et Leibniz autour de la « pensée aveugle »), sur des textes de penseurs plus contemporains (Bachelard, Bateson, Caillois, Derrida, Latour, Rosset), des fondateurs du projet d’IA (McCarthy, Turing) et de chercheurs actuels (Hinton, LeCun) cette thèse offre une description de l’intuition, de phénomènes récalcitrants à la simulation (compréhension, réflexivité, invention), et mène une réflexion sur le hasard afin de statuer sur la possibilité, pour les machines, d’être à l’origine de formes nouvelles
La didactique des arts plastiques et la mémoire agissante du dessin
No full textInternational audienceFifty years after its introduction as a school discipline, thevisual arts (AP) are still the object of recurrent questioning,particularly by certain supporters of the learning of “drawing”in the sense of strictly technical knowledge. From a cultural-historical perspective (Vygotski, 1930/1985), the purposeof this article is, on the one hand, to review the origin of thisold antagonism, a survival of a quarrel between “ancients”and “moderns”, and on the other hand, to reaffirm the foundationsof a AP didactics that does not deny drawing. Whatis the source of such regularly reactivated questioning? Thecurrent teaching programmes are bearers of memory : whatchanges, ruptures and continuities can be identified, forwhat effects on the norms and frameworks of the work of educationprofessionals? The first part of this article proposesan approach to the prescriptions that will allow us to bringto light certain links of filiations and nodes of tension. Thesecond part questions the foundations of a didactic of AP astaught today in France which continues to visit the functionand place of drawing.Cinquante ans après son instauration en tant que disciplinescolaire, les arts plastiques (AP) sont toujours l’objet de remisesen cause récurrentes, notamment par certains partisansde l’apprentissage du « dessin » dans une acception desavoirs strictement techniques. Dans une perspective historico-culturelle (Vygotski, 1930/1985), l’objet de cet article est,d’une part, de revenir sur l’origine de ce vieil antagonisme,survivance d’une querelle entre « anciens » et « modernes »et, d’autre part, de réaffirmer les fondements d’une didactiquedes AP qui ne renie en rien le dessin. D’où viennent detelles remises en question réactivées à intervalles réguliers ?Les programmes d’enseignement actuels sont porteurs demémoire : quels infléchissements, ruptures et continuitéspeuvent être identifiés ? Pour quels effets sur les normes etles cadres du travail des professionnels de l’éducation ? Lapremière partie de cet article propose une approche desprescriptions permettant de mettre au jour certains liens defiliation et des noeuds de tension. La deuxième partie revientsur les fondements d’une didactique des AP, telle qu’enseignéeaujourd’hui en France, qui continue d’interroger la fonctionet la place du dessin
