Dissertationen zum Thema „Réseaux tensoriels“
Geben Sie eine Quelle nach APA, MLA, Chicago, Harvard und anderen Zitierweisen an
Machen Sie sich mit Top-17 Dissertationen für die Forschung zum Thema "Réseaux tensoriels" bekannt.
Neben jedem Werk im Literaturverzeichnis ist die Option "Zur Bibliographie hinzufügen" verfügbar. Nutzen Sie sie, wird Ihre bibliographische Angabe des gewählten Werkes nach der nötigen Zitierweise (APA, MLA, Harvard, Chicago, Vancouver usw.) automatisch gestaltet.
Sie können auch den vollen Text der wissenschaftlichen Publikation im PDF-Format herunterladen und eine Online-Annotation der Arbeit lesen, wenn die relevanten Parameter in den Metadaten verfügbar sind.
Sehen Sie die Dissertationen für verschiedene Spezialgebieten durch und erstellen Sie Ihre Bibliographie auf korrekte Weise.
Pawlowski, Filip igor. „High-performance dense tensor and sparse matrix kernels for machine learning“. Thesis, Lyon, 2020. http://www.theses.fr/2020LYSEN081.
Der volle Inhalt der QuelleIn this thesis, we develop high performance algorithms for certain computations involving dense tensors and sparse matrices. We address kernel operations that are useful for machine learning tasks, such as inference with deep neural networks (DNNs). We develop data structures and techniques to reduce memory use, to improve data locality and hence to improve cache reuse of the kernel operations. We design both sequential and shared-memory parallel algorithms. In the first part of the thesis we focus on dense tensors kernels. Tensor kernels include the tensor--vector multiplication (TVM), tensor--matrix multiplication (TMM), and tensor--tensor multiplication (TTM). Among these, TVM is the most bandwidth-bound and constitutes a building block for many algorithms. We focus on this operation and develop a data structure and sequential and parallel algorithms for it. We propose a novel data structure which stores the tensor as blocks, which are ordered using the space-filling curve known as the Morton curve (or Z-curve). The key idea consists of dividing the tensor into blocks small enough to fit cache, and storing them according to the Morton order, while keeping a simple, multi-dimensional order on the individual elements within them. Thus, high performance BLAS routines can be used as microkernels for each block. We evaluate our techniques on a set of experiments. The results not only demonstrate superior performance of the proposed approach over the state-of-the-art variants by up to 18%, but also show that the proposed approach induces 71% less sample standard deviation for the TVM across the d possible modes. Finally, we show that our data structure naturally expands to other tensor kernels by demonstrating that it yields up to 38% higher performance for the higher-order power method. Finally, we investigate shared-memory parallel TVM algorithms which use the proposed data structure. Several alternative parallel algorithms were characterized theoretically and implemented using OpenMP to compare them experimentally. Our results on up to 8 socket systems show near peak performance for the proposed algorithm for 2, 3, 4, and 5-dimensional tensors. In the second part of the thesis, we explore the sparse computations in neural networks focusing on the high-performance sparse deep inference problem. The sparse DNN inference is the task of using sparse DNN networks to classify a batch of data elements forming, in our case, a sparse feature matrix. The performance of sparse inference hinges on efficient parallelization of the sparse matrix--sparse matrix multiplication (SpGEMM) repeated for each layer in the inference function. We first characterize efficient sequential SpGEMM algorithms for our use case. We then introduce the model-parallel inference, which uses a two-dimensional partitioning of the weight matrices obtained using the hypergraph partitioning software. The model-parallel variant uses barriers to synchronize at layers. Finally, we introduce tiling model-parallel and tiling hybrid algorithms, which increase cache reuse between the layers, and use a weak synchronization module to hide load imbalance and synchronization costs. We evaluate our techniques on the large network data from the IEEE HPEC 2019 Graph Challenge on shared-memory systems and report up to 2x times speed-up versus the baseline
Moreaux, Patrice. „Structuration des chaines de Markov des réseaux de Petri stochastiques : décomposition tensorielle et agrégation“. Paris 9, 1996. https://portail.bu.dauphine.fr/fileviewer/index.php?doc=1996PA090041.
Der volle Inhalt der QuelleChbihi, Abdelouahed. „Analyse scalaire et tensorielle de la refermeture des porosités en mise forme“. Thesis, Paris Sciences et Lettres (ComUE), 2018. http://www.theses.fr/2018PSLEM047/document.
Der volle Inhalt der QuelleThe presence of voids in ingots is a major issue in the casting industry. These voids decrease materials properties (in particular ductility) and may induce premature failure during metal forming or service life. Hot metal forming processes are therefore used to close these voids and obtain a sound product. However, the amount of deformation required to close these voids is difficult to estimate.Numerical modeling is an interesting tool to study the influence of process parameters on void closure rate. In this work, an optimization-based strategy has been developed to identify the parameters of a mean-field model based on a database of 800 full-field REV simulations with various loading conditions and voids geometry and orientations. The first void closure model is a scalar model that gets rid of the axisymmetric loading hypothesis considered in most models in the literature. The Lode angle, coupled with the stress triaxiality ratio enables to identify the stress state in a unique way. Comparisons of this new model with three other models fromthe literature show the accuracy increase for general loading conditions. In order to address multistages processes, a second model is defined in a tensor version. The ellipsoid void inertia matrix is used to define void’s morphology, orientation and volume. The tensor model predicts the evolution of the inertia terms and its calibration is based on the full-field REV database and on a new Artificial Neural Networks approach. Comparisons were carried out between this tensor model, the scalar model and full-field simulations for multi-stages configurations. These comparisons showed up to 35% accuracy improvement with the tensor model. It is worth mentioning that this is the first attempt to define a void closure tensor model in the literature
Catalano, Alberto Giuseppe. „Understanding and exploiting non-local effects in quantum spin chains“. Electronic Thesis or Diss., Strasbourg, 2024. http://www.theses.fr/2024STRAF022.
Der volle Inhalt der QuelleAt the verge of the second quantum revolution, understanding and exploiting the phenomena resulting from the interplay between the intrinsic non-locality of quantum mechanics and purely non-local interactions is of crucial importance for the development of novel quantum technologies. In this thesis, we will mostly focus on the non-local effects introduced by topological frustration (TF), a form of weak frustration that was first introduced in the context of antiferromagnetic quantum spin chains by applying the so called frustrated boundary conditions, realized as a combination of periodic boundary conditions and odd number of spins. Our goal is double. From one side, we will further improve the theoretical understanding of topologically frustrated phases. Beyond these theoretical implications, this work will demonstrate that TF spin chains exhibit compelling technological potential, proposing them as competitive candidates for the development of robust and efficient quantum batteries
Correa, de Sales Afonso Henrique. „Réseaux d'Automates Stochastiques : Génération de l'espace d'états atteignables et Multiplication vecteur-descripteur pour une sémantique en temps discret“. Grenoble INPG, 2009. http://www.theses.fr/2009INPG0141.
Der volle Inhalt der QuelleThis thesis presents methods and algorithms for the performance evaluation of large state space models described by Stochastic Automata Networks (SAN) formalism. We propose reachable state space generation methods of SAN models which use transitions that depend on model's global states. We also propose a multiplication method of a vector by the transition matrix of a discret time SAN model which is represented by a descriptor. The method executes a series of operations that manipulate data on the size of an automaton. (and for ail automata) instead of executing the multiplication on the product space of the'-' mode!
Borsoi, Ricardo Augusto. „Variabilité spectrale en démélange de données hyperspectrales : Stratégies multi-échelles, tensorielles et basées sur des réseaux neuronaux“. Thesis, Université Côte d'Azur, 2021. http://www.theses.fr/2021COAZ4012.
Der volle Inhalt der QuelleThe spectral signatures of the materials contained in hyperspectral images, also called endmembers (EMs), can be significantly affected by variations in atmospheric, illumination or environmental conditions typically occurring within an image. Traditional spectral unmixing (SU) algorithms neglect the spectral variability of the endmembers, what propagates significant mismodeling errors throughout the whole unmixing process and compromises the quality of the estimated abundances. Therefore, significant effort have been recently dedicated to mitigate the effects of spectral variability in SU. However, many challenges still remain in how to best explore a priori information about the problem in order to improve the quality, the robustness and the efficiency of SU algorithms that account for spectral variability. In this thesis, new strategies are developed to address spectral variability in SU. First, an (over)-segmentation-based multiscale regularization strategy is proposed to explore spatial information about the abundance maps more effectively. New algorithms are then proposed for both semi-supervised and blind SU, leading to improved abundance reconstruction performance at a small computational complexity. Afterwards, three new models are proposed to represent spectral variability of the EMs in SU, using parametric, tensor, and neural network-based representations for EM spectra at each image pixel. The parametric model introduces pixel-dependent scaling factors over a reference EM matrix to model arbitrary spectral variability, while the tensor-based representation allows one to exploit the high-dimensional nature of the data by means of its underlying low-rank structure. Generative neural networks (such as variational autoencoders or generative adversarial networks) finally allow one to model the low-dimensional manifold of the spectral signatures of the materials more effectively. The proposed models are used to devise three new blind SU algorithms, and to perform data augmentation in library-based SU. Finally, we provide a brief overview of work which extends the proposed strategies to new problems in SU and in hyperspectral image analysis. This includes the use of the multiscale abundance regularization in nonlinear SU, modeling spectral variability and accounting for sudden changes when performing SU and change detection of multitemporal hyperspectral images, and also accounting for spectral variability and changes in the multimodal (i.e., hyperspectral and multispectral) image fusion problem
Carrozza, Sylvain. „Methodes tensorielles et renormalization appliquées aux théories GFT“. Phd thesis, Université Paris Sud - Paris XI, 2013. http://tel.archives-ouvertes.fr/tel-00872784.
Der volle Inhalt der QuelleCarrozza, Sylvain. „Tensorial methods and renormalization in Group Field Theories“. Thesis, Paris 11, 2013. http://www.theses.fr/2013PA112147/document.
Der volle Inhalt der QuelleIn this thesis, we study the structure of Group Field Theories (GFTs) from the point of view of renormalization theory.Such quantum field theories are found in approaches to quantum gravity related to Loop Quantum Gravity (LQG) on the one hand,and to matrix models and tensor models on the other hand. They model quantum space-time, in the sense that their Feynman amplitudes label triangulations, which can be understood as transition amplitudes between LQG spin network states. The question of renormalizability is crucial if one wants to establish interesting GFTs as well-defined (perturbative) quantum field theories, and in a second step connect them to known infrared gravitational physics. Relying on recently developed tensorial tools, this thesis explores the GFT formalism in two complementary directions. First, new results on the large cut-off expansion of the colored Boulatov-Ooguri models allow to explore further a non-perturbative regime in which infinitely many degrees of freedom contribute. The second set of results provide a new rigorous framework for the renormalization of so-called Tensorial GFTs (TGFTs) with gauge invariance condition. In particular, a non-trivial 3d TGFT with gauge group SU(2) is proven just-renormalizable at the perturbative level, hence opening the way to applications of the formalism to (3d Euclidean) quantum gravity
Maurice, Olivier. „Introduction d’une théorie des jeux dans des topologies dynamiques“. Limoges, 2013. http://aurore.unilim.fr/theses/nxfile/default/77d23cbe-e698-42fd-acea-866f63d382a6/blobholder:0/2013LIMO4048.pdf.
Der volle Inhalt der QuelleThis thesis presents a method for modeling complexity. Starting from tensorial analysis of networks, we show that this technique allows to model any physical process. It gives in a common formalism all the tools to integrate equations coming from various physics. The purpose is not to develop an unique method rather than having one able to embed developments coming from any kind of physic material. The formalism embed quantum mechanics, relativity, etc. Once the physical part of the system take in charge, we use game theory to take the psychical part. Both methods linked by special mathematical objects like "tenfolds" or gamma matrices makes a global technique for complexity. A tree cross talking the two theories models the complex system evolution. A special representation in a "choices-utility" space gives a comprehensible image of the system evolution
Frusque, Gaëtan. „Inférence et décomposition modale de réseaux dynamiques en neurosciences“. Thesis, Lyon, 2020. http://www.theses.fr/2020LYSEN080.
Der volle Inhalt der QuelleDynamic graphs make it possible to understand the evolution of complex systems evolving over time. This type of graph has recently received considerable attention. However, there is no consensus on how to infer and study these graphs. In this thesis, we propose specific methods for dynamical graph analysis. A dynamical graph can be seen as a succession of complete graphs sharing the same nodes, but with the weights associated with each link changing over time. The proposed methods can have applications in neuroscience or in the study of social networks such as Twitter and Facebook for example. The issue of this thesis is epilepsy, one of the most common neurological diseases in the world affecting around 1% of the population.The first part concerns the inference of dynamical graph from neurophysiological signals. To assess the similarity between each pairs of signals, in order to make the graph, we use measures of functional connectivity. The comparison of these measurements is therefore of great interest to understand the characteristics of the resulting graphs. We then compare functional connectivity measurements involving the instantaneous phase and amplitude of the signals. We are particularly interested in a measure called Phase-Locking-Value (PLV) which quantifies the phase synchrony between two signals. We then propose, in order to infer robust and interpretable dynamic graphs, two new indexes that are conditioned and regularized PLV. The second part concerns tools for dynamical graphs decompositions. The objective is to propose a semi-automatic method in order to characterize the most important patterns in the pathological network from several seizures of the same patient. First, we consider seizures that have similar durations and temporal evolutions. In this case the data can be conveniently represented as a tensor. A specific tensor decomposition is then applied. Secondly, we consider seizures that have heterogeneous durations. Several strategies are proposed and compared. These are methods which, in addition to extracting the characteristic subgraphs common to all the seizures, make it possible to observe their temporal activation profiles specific to each seizures. Finally, the selected method is used for a clinical application. The obtained decompositions are compared to the visual interpretation of the clinician. As a whole, we found that activated subgraphs corresponded to brain regions involved during the course of the seizures and their time course were highly consistent with classical visual interpretation
Rabusseau, Guillaume. „A tensor perspective on weighted automata, low-rank regression and algebraic mixtures“. Thesis, Aix-Marseille, 2016. http://www.theses.fr/2016AIXM4062.
Der volle Inhalt der QuelleThis thesis tackles several problems exploring connections between tensors and machine learning. In the first chapter, we propose an extension of the classical notion of recognizable function on strings and trees to graphs. We first show that the computations of weighted automata on strings and trees can be interpreted in a natural and unifying way using tensor networks, which naturally leads us to define a computational model on graphs: graph weighted models; we then study fundamental properties of this model and present preliminary learning results. The second chapter tackles a model reduction problem for weighted tree automata. We propose a principled approach to the following problem: given a weighted tree automaton with n states, how can we find an automaton with m
Zniyed, Yassine. „Breaking the curse of dimensionality based on tensor train : models and algorithms“. Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLS330.
Der volle Inhalt der QuelleMassive and heterogeneous data processing and analysis have been clearly identified by the scientific community as key problems in several application areas. It was popularized under the generic terms of "data science" or "big data". Processing large volumes of data, extracting their hidden patterns, while preforming prediction and inference tasks has become crucial in economy, industry and science.Treating independently each set of measured data is clearly a reductiveapproach. By doing that, "hidden relationships" or inter-correlations between thedatasets may be totally missed. Tensor decompositions have received a particular attention recently due to their capability to handle a variety of mining tasks applied to massive datasets, being a pertinent framework taking into account the heterogeneity and multi-modality of the data. In this case, data can be arranged as a D-dimensional array, also referred to as a D-order tensor.In this context, the purpose of this work is that the following properties are present: (i) having a stable factorization algorithms (not suffering from convergence problems), (ii) having a low storage cost (i.e., the number of free parameters must be linear in D), and (iii) having a formalism in the form of a graph allowing a simple but rigorous mental visualization of tensor decompositions of tensors of high order, i.e., for D> 3.Therefore, we rely on the tensor train decomposition (TT) to develop new TT factorization algorithms, and new equivalences in terms of tensor modeling, allowing a new strategy of dimensionality reduction and criterion optimization of coupled least squares for the estimation of parameters named JIRAFE.This methodological work has had applications in the context of multidimensional spectral analysis and relay telecommunications systems
Benoit, Anne. „Méthodes et algorithmes pour l'évaluation des performances des systèmes“. Phd thesis, Grenoble INPG, 2003. http://tel.archives-ouvertes.fr/tel-00004361.
Der volle Inhalt der QuelleCholachue, Ngounou Christel. „Caractérisation des blindages électromagnétiques des câbles et faisceaux aéronautiques“. Thesis, Normandie, 2020. http://www.theses.fr/2020NORMR025.
Der volle Inhalt der QuelleDuring the last decade, the proliferation of on-board leisure activities in the new aircrafts have been growing exponentially. In the airplane like A380, each seat integrates several functions (video games, music, etc. ..) Additionally, each function must be connected using at least one electric cable. This system requires a significant number of kilometers of cables to establish all the on-board electrical connections. Furthermore, for reasons of safety and security related to mechanical, hydraulic or pneumatic functions, the wiring EMC requirements associated to the massive progressive electrification becomes considerably stricter. The coexistence of kilometer lengths of cables system in such a small space has increased the requirements in terms of electromagnetic (EM) shielding. Most of existing numerous methods for analyzing the shielding of cables and harnesses are limited in terms of computation speed, design process and in accuracies for the multiport systems analysis. Moreover, most of popular simulation and commercial tools are very expensive (for example with license cost can be more than 18K€). The use of commercial tools requires advanced skills and a lot of time to characterize the shielding of cables and harnesses. For example, with a simulation tool like HFSSR from ANSYSR , the computation time may cost approximately 3 hours to create a design model of a braided shields heath. Then, the computed results can be generated during an average simulation time of 20 minutes using a PC equipped with an Intel single-core processor RXeon RCPU E5-1620 v4 @ 3.50 GHz and 32 GB of physical RAM with 64-bit Windows 10. Most of available methods and techniques for characterizing the shielding of aeronautical cables and cable harnesses have shown their limits. For example, most of existing triaxial benches are particularly difficult to deploy for the transfer impedance measurements and they cannot operate beyond 100 MHz. The present PhD thesis aims to overcome these technical limits. Doing this, an original analytical method is developed for extracting S-parameters from multiport systems under fast computation speed and design process. An innovative methodology of EMC modelling was proposed. The knowledge of S-parameters is helpful to determine the broadband EM intrinsic parameters of the cabling as coaxial system. The developed analytical and semi-hybrid model is based on the unfamiliar formalism using tensorial analysis of networks (TAN) based on the Kron’s method. The model offers an outstanding possibility to analyze complex systems with deep knowledge of physical phenomenal behind the EM shielding. Thanks to the TAN formalism, an innovative method of circuit theory has been developed to determine the shielding efficiency (SE) of shielded cable. The feasibility of this multiport S-parameter approach was verified with the consideration of EM coupling between a nude cable constituting an internal conductor and a braided cable placed in parallel. More importantly, an advanced study of shielding efficiency (SE) with respect to the EM coupling configuration between a shielded coaxial cable and a loop probe is performed. Substantially, it was noteworthy that the TAN formalism provides an illuminating know-how on the theoretical, numerical and experimental analyses of cables and bundles EM shielding, and transfer impedances of the shielding sheath. Moreover, the TAN modelling effectiveness was confirmed with different applications with computation time which does not exceed milliseconds. Finally, the TAN model was also used to develop a SE characterization bench for tubular EM shielding structures up to 300 MHz. Emphatically, broadband SE and transfer impedance results in good correlation between 3D simulations and measurements were obtained
Brenner, Leonardo. „Réseaux d'Automates Stochastiques : Analyse transitoire en temps continu et algèbre tensorielle pour une sémantique en temps discret“. Phd thesis, 2009. http://tel.archives-ouvertes.fr/tel-00424652.
Der volle Inhalt der QuelleLa première partie de cette thèse s'intéresse au calcul des indices de performances transitoires pour des grands modèles. Lorsqu'on calcule des indices de performances transitoires, tel que la disponibilité ponctuelle, la méthode d'uniformisation est la plus souvent utilisée. Cependant le nombre d'itérations vecteur-matrice peut être très grand ce qui devient critique pour de très grands modèles. Des méthodes de détection du régime stationnaire peuvent réduire le coût de calcul en arrêtant les itérations lorsque le régime stationnaire est atteint. Dans cette thèse, nous proposons une adaptation et une comparaison de différentes méthodes de détection du régime stationnaire lorsque la matrice est stockée sous un format tensoriel. Les méthodes sont comparées selon deux critères : nombre d'itérations et précision des résultats.
Dans la deuxième partie, nous présentons le formalisme SAN à temps discret. La définition formelle du formalisme SAN présentée dans cette thèse nous permet de définir la sémantique des modèles en temps discret que nous souhaitons exploiter. Nous définissons une nouvelle algèbre tensorielle (appelée Algèbre Tensorielle compleXe - ATX) capable d'exprimer cette sémantique. Pour cela, trois opérateurs sont définis afin de décrire différents comportement d'un système, tels que la simultanéité, la concurrence et le choix. Enfin, le principal apport de cette thèse réside dans la définition d'une formule tensorielle (appelée Descripteur discret) qui utilise cette nouvelle algèbre pour représenter un modèle SAN à temps discret de façon compacte. Nous montrons que ce descripteur discret permet aisément de générer la chaîne de Markov représentée par le modèle SAN.
Correa, De Sales Afonso Henrique. „Réseaux d'Automates Stochastiques : Génération de l'espace d'états atteignables et Multiplication vecteur-descripteur pour une sémantique en temps discret“. Phd thesis, 2009. http://tel.archives-ouvertes.fr/tel-00436020.
Der volle Inhalt der QuelleFernandes, Paulo. „Méthodes numériques pour la solution de systèmes Markoviens à grand espace d'états“. Phd thesis, 1998. http://tel.archives-ouvertes.fr/tel-00004886.
Der volle Inhalt der Quelle