Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Analyse stochastique de phénomènes irréguliers non-markoviens
No full textThis thesis focuses on some particular stochastic analysis aspects of non-Markovian irregular phenomena. It formulates existence and uniqueness for some martingale problems involving two types of irregulat drifts perturbed by path-dependant functionals: the first one is related to the case which is the derivative of continuous function and it models irregular path-dependent media; the second one concerns the case when the drift is of Bessel type in low dimension. Finally the thesis also focuses on rough paths techniques and its relation with the stochastic calculus via regularization.Cette thèse se concentre sur certains aspects d'analyse stochastique de modèles non-markoviens irréguliers. On formule existence et unicité pour certains problèmes de martingales impliquant deux types de dérive irrégulière perturbée par des fonctionnelles dépendant de la trajectoire. Dans le premier cas, on considère le cas où la dérive est la dérivée d'une fonction continue: le modèle correspondant est celui de milieux aléatoires irréguliers dépendant de la trajectoire. Le second concerne le cas où la dérive est celui d'un processus de Bessel en basse dimension: dans ce cas il est bien connu qu'en général les processus ne sont pas des semimartingales. Enfin la thèse explore également des relations et des analogies entre la théorie des chemins rugueux et le calcul stochastique via régularisation
La coopétition en R&D : une étude à partir de données de co-inventions
No full textInternational audienc
Planification optimisée du déploiement d'un réseau de télécommunication multitechnologie par dispositifs aéroportés sur un théâtre d'opérations extérieures
No full textInternational audienc
Real & Simulated QPSK Up-Converted Signals by a Sampling Method Using a Cascaded MZMs Link
No full textInternational audienceThis study focuses on a novel concept of transmitting of a quadrature phase shift keying (QPSK) modulation by an electro-optical frequency up-conversion using a cascaded Mach–Zehnder modulators (MZMs) link. Furthermore, we conduct and compare the results obtained by simulations using the Virtual Photonics Inc. (VPI) (Berlin, Germany) simulator and real-world experiments. The design and operating regime peculiarities of the MZM used as a sampling up-converter mixer in a radio over fiber (RoF) system are also analyzed. Besides, the simulation and experimental results of static and dynamic characteristics of the MZM have approximately the same behavior. The conversion gain of the cascaded MZMs link is simulated over many mixing frequencies and it can decrease from 17.5 dB at 8.3 GHz to −4.5 dB at 39.5 GHz. However, in real world settings, it may decrease from 15.5 dB at 8.3 GHz to −6 dB at 39.5 GHz. The maximum frequency range is attained at 78.5 GHz for up-conversion through simulations. Error vector magnitude (EVM) values have been done to evaluate the performance of our system. An EVM of 16% at a mixing frequency of 39.5 GHz with a bit rate of 12.5 Gbit/s was observed with the considering sampling technique, while it reached 19% in real-world settings with a sampling frequency of 39.5 GHz and a bit rate of 12.5 Gbit/s
Lateral buckling of submarine pipelines under high temperature and high pressure - A literature review
No full textInternational audienceThis paper consists of a literature review on the latest research progress about the lateral buckling of submarine pipelines under high temperature (HT) and high pressure (HP). First, the main general assumptions and simplifications made in the context of pipe lateral buckling are summarized in order to better understand the practical behavior of submarine pipelines. The governing equations of pipelines under uniaxial compression are then derived. Next, the controversial but widely deployed concept of effective axial force under complex sea environment is introduced. Influential parameters including initial imperfections, pipe-seabed interactions and residual stresses are elaborated and discussed. Furthermore, numerical simulation methods and experimental tests dealing with the lateral buckling of pipelines are presented as well. Controlled methods which are practically used for buckle initiation are described. This paper also reveals the remaining challenges and new tendencies, such as the use of data-driven methods for the smart prediction of pipe buckling. Finally, a specific case study is numerically conducted. It is found that the effect of axial friction variation can be generally ignored in practical calculations. This paper may provide a guidance for the design and research on pipelines in the future
Global optimization for sparse solution of least squares problems
No full textInternational audienceFinding solutions to least-squares problems with low cardinality has found many applications, including portfolio optimization, subset selection in statistics, and inverse problems in signal processing. Although most works consider local approaches that scale with high-dimensional problems, some others address its global optimization via mixed integer programming (MIP) reformulations. We propose dedicated branch-and-bound methods for the exact resolution of moderate-size, yet difficult, sparse optimization problems, through three possible formulations: cardinality-constrained and cardinality-penalized least-squares, and cardinality minimization under quadratic constraints. A specific tree exploration strategy is built. Continuous relaxation problems involved at each node are reformulated as (Formula presented.) -norm-based optimization problems, for which a dedicated algorithm is designed. The obtained certified solutions are shown to better estimate sparsity patterns than standard methods on simulated variable selection problems involving highly correlated variables. Problem instances selecting up to 24 components among 100 variables, and up to 15 components among 1000 variables, can be solved in less than 1000 s. Unguaranteed solutions obtained by limiting the computing time to 1s are also shown to provide competitive estimates. Our algorithms strongly outperform the CPLEX MIP solver as the dimension increases, especially for quadratically-constrained problems. The source codes are made freely available online
Assistances numériques domiciliaires pour les personnes âgées fragiles : Etudes de conception et d’évaluation pilote d’une technologie ambiante d’assistance domiciliaire basée sur l’orchestration d’objets connectés.
No full textLes Technologies d’Assistance numériques (TAn), visant à soutenir l’autonomie et la participation sociale des personnes âgées, sont un domaine en pleine expansion, comme en témoignent l’offre commerciale et les recherches actuelles. En effet, les avancées technologiques observées permettent d’envisager cette voie comme prometteuse et porteuse de progrès médico-social pour les personnes concernées. Pour dépasser les limites de l’approche techno-centrée, générant des produits souvent inadaptés aux besoins des futurs utilisateurs âgés et clamant des allégations de santé non justifiées scientifiquement, une approche intégrée est présentée alliant les modèles ergonomiques de conception centrée-utilisateur et les méthodes expérimentales de validation clinique, avec une emphase donnée aux mécanismes de motivation intrinsèque liés à l’auto-détermination. Pour illustration, est exposée une série de travaux portant sur une plateforme d’objets connectés pour l’assistance domiciliaire de personnes âgées fragiles, appelée DomAssist. Ces travaux ont été menés depuis la conception jusqu’au déploiement de la plateforme sur le terrain, avec des validations scientifiques amont (ergonomie et motivation suscitée) et aval (étude pilote des gains cliniques auprès des personnes âgées fragiles et/ou leurs aidants). Au final, des résultats prometteurs sont obtenus : la solution DomAssist chez la personne âgée fragile génère une bonne expérience utilisateur, améliore le sentiment d’auto-détermination, retarde les dégradations fonctionnelles, et enfin réduit le fardeau des aidants dans l’aide qu’ils apportent pour le fonctionnement quotidien de la personne âgée
On the justification of topological derivative for wave-based qualitative imaging of finite-sized defects in bounded media
No full textInternational audienceThe concept of topological derivative (TD) is known to provide, through its heuristic interpretation involving its sign and its spatial decay away from the true anomaly, a basis for the qualitative imaging of finite-sized anomalies. The TD imaging heuristic is currently partially backed by conditional mathematical justifications. Continuing earlier efforts towards the justification of TD-based identification, this work investigates the acoustic wave-based imaging of finite-sized (i.e. not necessarily small) medium anomalies embedded in bounded domains and affecting the leading-order term of the acoustic field equation. Both the probing excitation and the measurement are assumed to take place on the domain boundary. We extend to this setting the analysis approach previously used for unbounded media with either refraction-index anomalies and far-field measurements (Bellis et al., \emph{Inverse Problems} \textbf{29}:075012, 2013) or mass-density anomalies and meaurements at finite distance (Bonnet, Cakoni, \emph{Inverse Problems} \textbf{35}:104007, 2019). Like in the latter work, TD-based imaging functionals are reformulated for analysis using a suitable factorization of the acoustic fields, facilitated by a volume integral formulation. Our results, which echo corresponding results of our earlier investigations, conditionally validate the TD imaging heuristic. Moreover, we show on a geometrically simple configuration that the spatial behavior of the TD associated with standard cost functionals is degraded by ``echoes'' of the true anomaly, an aspect specific to the present bounded-domain framework. This undesirable effect is removed by a combination of (i) post-processing the measurements by application of a suitable integral operator (a treatment introduced by Ammari et al., 2011, for the analysis of TD-based imaging involving true flaws modelled using small-anomaly asymptotics), and (ii) expressing the background field as an incoming single-layer potential defined in the full space (after an idea used in Bonnet, Cakoni, 2019). Finally, we also show that selecting eigenfunctions of the source-to-measurement operator as excitations enhances the spatial decay properties of the TD functional
One Versus all for deep Neural Network Incertitude (OVNNI) quantification
No full textInternational audienceDeep neural networks (DNNs) are powerful learning models yet their results are not always reliable. This is due to the fact that modern DNNs are usually uncalibrated and we cannot characterize their epistemic uncertainty. In this work, we propose a new technique to quantify the epistemic uncertainty of data easily. This method consists in mixing the predictions of an ensemble of DNNs trained to classify One class vs All the other classes (OVA) with predictions from a standard DNN trained to perform All vs All (AVA) classification. On the one hand, the adjustment provided by the AVA DNN to the score of the base classifiers allows for a more fine-grained inter-class separation. On the other hand, the two types of classifiers enforce mutually their detection of out-of-distribution (OOD) samples, circumventing entirely the requirement of using such samples during training. Our method achieves state of the art performance in quantifying OOD data across multiple datasets and architectures while requiring little hyper-parameter tuning
Practical multiverse debugging through user-defined reductions
No full textInternational audienceMultiverse debugging is an extension of classical debugging methods, particularly adapted to non-deterministic systems. Recently, a language-independent formalization was proposed. Moreover, multiverse debugging is particularly beneficial for specification and design languages, such as UML. However, this method suffers from scalability issues during breakpoint lookup. This problem arises due to the exhaustive exploration performed on the potentially infinite state-space of the system. In this paper, we tackle this problem by introducing Reduced Multiverse Debugging, an extension proposing a way for the user to define reduction policies used during breakpoint lookup. We enrich the formalization of multiverse debugging with a modular breakpoint lookup strategy, which allows the integration of the reduction policy. We validate our approach by implementing a practical UML Statechart debugger in the AnimUML web framework. We show several ways the reduction can be applied, using methods such as predicate abstraction for breakpoint lookup on an infinite state-space, removing irrelevant variables, or creating classes of equivalent values. Moreover, we show the possibility to integrate probabilistic reduction strategies. Relying on hash collisions, these strategies can be iteratively refined to increase precision