Scientific Publications of the University of Toulouse II Le Mirail
Not a member yet
92205 research outputs found
Sort by
Potentiel des satellites Pléiades Neo pour restituer la topographie fine des glaciers
No full textInternational audienceLa résolution de 30 cm offerte par les satellites Pléiades Neo représente un gain d’un facteur deux par rapport à leurs prédécesseurs, les Pléiades HR. Combinée avec des capacités stéréoscopiques, elle permet de construire des modèles numériques de terrain (MNTs) métriques pour caractériser finement les reliefs, en particulier pour les petits glaciers, comme ceux des Pyrénées. Dans cette étude, nous évaluons la précision des MNTs déduits d’images Pléiades Neo sur les glaciers de l’Aneto, d’Ossoue et du Monte Perdido en les comparant à des MNTs acquis par drone et aux couvertures Lidar nationales. Nous montrons que les MNTs déduits avec le corrélateur More Global Matching (MGM) sont plus précis que d’autres corrélateurs. Nous soulignons également la capacité des MNTs Pléiades Neo à restituer des détails fins de la morphologie glaciaire comme les moulins sur le glacier d’Ossoue, beaucoup plus difficiles à caractériser avec Pléiades HR
Admission Control of Quasi-Reversible Queueing Systems: Optimization and Reinforcement Learning
No full textIn this paper, we introduce a versatile scheme for optimizing the arrival rates of quasi-reversible queueing systems. We first propose an alternative definition of quasi-reversibility that encompasses reversibility and highlights the importance of the definition of customer classes. Then we introduce balanced arrival control policies, which generalize the notion of balanced arrival rates introduced in the context of Whittle networks, to the much broader class of quasi-reversible queueing systems. We prove that supplementing a quasi-reversible queueing system with a balanced arrival-control policy preserves the quasi-reversibility, and we specify the form of the stationary measures. We revisit two canonical examples of quasi-reversible queueing systems, Whittle networks and order-independent queues. Lastly, we focus on the problem of admission control and leverage our results in the frameworks of optimization and reinforcement learning
La recherche-intervention comme espace de supervision
No full textInternational audienceLes auteures présentent ici un dispositif de supervision construit dans le cadre d’une recherche-intervention pilotée par un laboratoire de recherche en collaboration avec des psychologues au sein d’organismes de formation par apprentissage. Elles témoignent des difficultés rencontrées par les intervenants dans ces établissements, mais aussi de la manière dont le partenariat avec l’université a permis de constituer un collectif identitaire par rapport aux autres professionnels. Cela a permis de maintenir un cadre solide, juste et contenant, favorable à la résolution des dilemmes déontologiques et éthiques
Martin Becker, Jaroslav Rudiš, Lost in Praha, traduction sous la direction d’Hélène Leclerc et Catherine Mazellier-Lajarrige, Toulouse, PUM
No full textpièce radiophonique inédite de Martin Becker et Jaroslav RudišLa pièce radiophonique Lost in Praha nous entraîne dans les rues labyrinthiques de Prague aux côtés de Tomas, enseignant rêveur, et de Tereza, son ancienne élève. Tombés amoureux depuis peu, ils ont quitté Berlin pour recommencer à zéro dans la mythique « Ville Dorée », théâtre de leurs fantasmes. Mais très vite, leurs chemins se séparent – au propre comme au figuré. Tomas, nourri par les écrits de Kafka et de Hrabal, ne voit en Prague qu’une vitrine figée, une cité-musée noyée sous les touristes. Tereza, au contraire, s’abandonne à l’effervescence de la Prague underground, celle des concerts, du rock alternatif, « là où le cœur de la vie bat vraiment » (M. Becker). Elle escalade le mur du cimetière juif avec Jára, un chanteur de rock, tandis que Tomas, victime de violents maux d’estomac, se retrouve hospitalisé à Bulovka. C’est là, entre deux souffrances et un réel vacillant, qu’il rencontre l’écrivain Hrabal – pourtant mort depuis des années –, qui lui offre une bouleversante leçon de vie
Capacité de charge
No full textLa capacité de charge touristique d’un site, d’une ville, d’un équipement, d’un site naturel peut être définie comme étant le nombre maximum de personnes qui peuvent s’y rendre et le visiter au même moment sans provoquer la destruction de son environnement physique, économique et socioculturel ou une gêne, une diminution de la satisfaction des visiteurs
Efficient Gaussian Process learning via subspace projections
No full textInternational audienceWe propose a novel training objective for GPs constructed using lower-dimensional linear projections of the data, referred to as projected likelihood (PL). We provide a closed-form expression for the information loss related to the PL and empirically show that it can be reduced with random projections on the unit sphere. We show the superiority of the PL, in terms of accuracy and computational efficiency, over the exact GP training and the variational free energy approach to sparse GPs over different optimisers, kernels and datasets of moderately large sizes
« Le fil d’araignée », Traduction de la nouvelle d’Akutagawa Ryūnosuke (japonais) par Christopher Chassagneux et Lucie Demesy.
No full textAkutagawa Ryūnosuke, « « Le fil d’araignée », Traduction de la nouvelled’Akutagawa Ryūnosuke (japonais) par Christopher Chassagneux et LucieDemesy. », La main de Thôt [En ligne], 13 | 2025, mis en ligne le 23 janvier 2026,consulté le 23 janvier 2026. URL : http://interfas.univ-tlse2.fr/lamaindethot/180
New nonlinear model for the prediction of components breakdown subject to transient disturbances
No full textDynamic models of electrostatic discharge (ESD)protection devices have many advantages in simulating the response of these components when they are subjected to very high transient pulses, especially in the cases of electromagnetic pulse residues (EMPs). However, a major problem with these models is that they do not consider the dynamics of variation of the component as a function of the energy flowing through it. In this paper, we will explore a way to improve dynamic simulation by building electrical models from frequency measurements, which can consider very fast current variations. Through a concrete case of a filter including a transient voltage suppressor (TVS) in parallel with a capacitor, we will demonstrate that the proposed modeling approach can predict the destruction of the latter
Two Schemes for Thining Knowledge Extraction
No full textInternational audienceTwo Schemes for Thinking Knowledge ExtractivismAlongside the debate on techno-feudalism—which contrasts two visions of digital capitalism, between the capture of rent through predation and the intensified exploitation of labor—I would like to foreground two schemes of knowledge or data capture and production across the history of modernity and capitalism.On the one hand, there is “reduction into art (réduction en art)” (Hélène Vérin), which consists in analytically describing trades or activities in order to systematize them and, where possible, mechanize them. The culmination of this process of reduction into art, initiated during the Renaissance and pursued through colonial questionnaires distributed in New Spain in the sixteenth century (the Relaciones Geográficas), is undoubtedly Diderot and d’Alembert’s Encyclopédie. Through its multiple figures and descriptions, we find—among many things— an article about pins which provided the model for the division of labor that Adam Smith would later mobilize in The Wealth of Nations. Practices of reduction into art continues with figures such as Babbage, Galton, and Taylor, each of whom systematized, in their own way, techniques for capturing knowledge through reduction into art, with an explicit political and economic agenda: the gathering of knowledge or data is always undertaken in order to secure a position of power or to intensify the rationalization of labor processes through their measurement and the increase of relative surplus value.On the other hand, where reduction into art fails, cybernetics has popularized the technique of “black-boxing”, which involves disregarding the internal details of a person or a system and focusing instead solely on the relationships between inputs and outputs for a given object. At first glance, this appears less as an act of extraction than as an admission of modesty: with the “excessively complex objects” (Stafford Beer) addressed by cybernetics, no knowledge can be extracted in the classical sense, and it is therefore preferable to black-box them. Cyberneticians then ask how such objects can be known and, above all, controlled—a question that leads in particular to the generalized computerization of their environment: in the absence of the ability to know and program individual entities, the aim becomes to master everything that surrounds them.Through these two schemes, whose complementarity rather than opposition we will examine, we will analyze the kinds of knowledge and power at stake between the objects and subjects of cognitive extractivism, as well as their connections to contemporary forms of capitalism
Finite convergence of the Moment-SOS hierarchy under hidden convexity
No full textInternational audienceOne considers polynomial optimization problems with compact feasible set Ω defined by SOS-concave polynomials gj, and with a globally non-convex polynomial objective f . We show that if f is strongly convex on Ω, or SOS-convex on Ω when the constraints gj are at most quadratic, then the Moment-SOS hierarchy converges in finitely many steps, without à priori knowledge of this hidden (local) convexity. In addition, in the latter case, the exact order for which the relaxation is exact is provided by the degree of a Putinar-like certificate of convexity. This demonstrates that a general-purpose hierarchy can adapt to favorable hidden properties of a specific instance without being informed of them, yielding certified global minimizers