Dissertations / Theses on the topic 'Réseaux neuronaux à deux couches'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 20 dissertations / theses for your research on the topic 'Réseaux neuronaux à deux couches.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Dabo, Issa-Mbenard. "Applications de la théorie des matrices aléatoires en grandes dimensions et des probabilités libres en apprentissage statistique par réseaux de neurones." Electronic Thesis or Diss., Bordeaux, 2025. http://www.theses.fr/2025BORD0021.
Full textThe functioning of machine learning algorithms relies heavily on the structure of the data they are given to study. Most research work in machine learning focuses on the study of homogeneous data, often modeled by independent and identically distributed random variables. However, data encountered in practice are often heterogeneous. In this thesis, we propose to consider heterogeneous data by endowing them with a variance profile. This notion, derived from random matrix theory, allows us in particular to study data arising from mixture models. We are particularly interested in the problem of ridge regression through two models: the linear ridge model and the random feature ridge model. In this thesis, we study the performance of these two models in the high-dimensional regime, i.e., when the size of the training sample and the dimension of the data tend to infinity at comparable rates. To this end, we propose asymptotic equivalents for the training error and the test error associated with the models of interest. The derivation of these equivalents relies heavily on spectral analysis from random matrix theory, free probability theory, and traffic theory. Indeed, the performance measurement of many learning models depends on the distribution of the eigenvalues of random matrices. Moreover, these results enabled us to observe phenomena specific to the high-dimensional regime, such as the double descent phenomenon. Our theoretical study is accompanied by numerical experiments illustrating the accuracy of the asymptotic equivalents we provide
Lorquet, Vincent. "Etude d'un codage semi-distribué adaptatif pour les réseaux multi-couches. Application au diagnostic, à la modélisation et à la commande." Paris, ENST, 1992. http://www.theses.fr/1992ENST0025.
Full textCosta, Pascale. "Contribution à l'utilisation des réseaux de neurones à couches en traitement du signal." Cachan, Ecole normale supérieure, 1996. http://www.theses.fr/1996DENS0030.
Full textClemens, Stephan. "Expression, modulation dynamique et interactions de deux réseaux neuronaux stomatogastriques des crustacés : études électrophysiologiques in vivo." Bordeaux 1, 1998. http://www.theses.fr/1998BOR10516.
Full textHugues, Emmanuel. "Modélisation des performances des avions par réseaux de neurones multi-couches : application à la maximisation de la masse au décollage." Toulouse, ENSAE, 1999. http://www.theses.fr/1999ESAE0003.
Full textAmontieva, Ianina. "Plasticité de la modulation serotoninergique de rythmes moteurs chez deux vertebrés inférieurs : la lamproie et l'urodele." Bordeaux 2, 2003. http://www.theses.fr/2003BOR21082.
Full textIt is now well established that the rythmic activities underlying motor behaviors in Vertebrates are generated by neural networks, which are under the control of several neuromodulators, among which serotonin (5-HT) plays a critical role because it is involved in the modulation and functional recovery after lesion of several motor networks. In the present study we have studied the serotonergic 5-HT modulation of the respiratory network in lamprey and that of the locomotor network after spinal cord injury in an adult salamander (Pleurodeles waltlii). Extracellular recordings from the ventral roots in lamprey showed that 5-HT modulates the activity of the respiratory rythm generator in different directions, depending on its concentration. The respiratory rythm was decreased at high concentration (>50 mu M) while it was increased at low (< mu M) concentration. We have investigated the effects of a local application (1-2 spinal segments) of 5-HT on the sublesional spinal cord in recovered (3-8 months post op. ) spinalized salamanders. The induced effects were compared to those induced by a similar drug application in intact animals. The locomotor pattern was documented using electromyographical recordings. Our results showed that an exogenous application of 5-HT (0. 5-205 mu M) induced a decrease of the swimming rythm in intact animal whereas it induced an increase in the swimming rythm in recovered spinal-transected animals. All these effects were mimicked by a local application of a specific 5-HT uptake blocker. This suggested that endogenously released 5-HT could also induce these changes. Pharmacological analyses using different specific 5-HT receptor agonists evidenced that reversal of 5-HT effects on the swimming frequency involved 5-HT3 like-receptors. . . .
Gereige, Issam. "Contribution des réseaux de neurones dans le domaine de l'ellipsométrie : application à la scatterométrie." Phd thesis, Université Jean Monnet - Saint-Etienne, 2008. http://tel.archives-ouvertes.fr/tel-00365631.
Full textPerez, Schuster Veronica. "the neural basis of motion after effect in zebrafish larvae." Thesis, Paris 6, 2014. http://www.theses.fr/2014PA066030.
Full textOne of the main goals in neuroscience is to understand how cognitive functions are encoded by the dynamics of large neuronal networks. The main stream to study sensory perception has mainly focused on sensory stimulation and neuronal recordings of the induced neural responses. An alternative approach is the use of sensory illusions, in which sensory perception take place in the absence of physical external stimulation. For this purpose, we have used a multidisciplinary approach combining the zebrafish larva as the experimental model, two-photon calcium imaging, motor behavior and optogenetics. We showed that the zebrafish larva is capable of perceiving motion after-effect (MAE). Using optogenetics (halorhodopsin) to prevent eye movements during the presentation of the conditioning stimulus (CS), we showed that pursuit eye movements during CS are not imperative for the induction of MAE, suggesting that neither muscular fatigue nor eye-muscle proprioception feedback play a role in the generation of MAE. Furthermore, we used two-photon microscopy in combination with transgenic fish expressing GcaMP3 . We first observed that during MAE, neurons in the optic tectum (the largest and highest visual brain center of the larva) were strongly habituated. This habituation later decayed with a temporal scale that matched that of the optomotor MAE-like behavior. In contrast, no significant habituation was observed in the retina, Thus, we suggested that the optic tectum but not the retina plays a role in generation of MAE. Our approach contributed to a more comprehensive view of the neuronal mechanisms underlying MAE, and shed light on the neuronal correlate of motion perception
Lefevre, Fabien. "Caractérisation de structures du type plaque par ondes guidées générées et détectées par laser." Valenciennes, 2010. http://ged.univ-valenciennes.fr/nuxeo/site/esupversions/24980ba6-f06c-4c75-988a-16e1228d2e42.
Full textThe deposition of thin layers on substrates is more and more required in many applications. For example, to reach high technical performance, bumpers or other parts are nickeled to improve their impermeability and resistance. Another example in microelectronics is the realization of transistors found in LCDs where they are associated with each pixel. The use of these layer/substrate structures is growing, so the importance of having non-destructive techniques to monitor and characterize them is well understood. The point in using ultrasonic waves for non-destructive testing and evaluation of various materials and structures is well known. In this work, the aim was to use guided waves to monitor and to characterize plaque-like structures. The main advantage of using these modes lies in their ability to test very large areas and inaccessible structures. For the generation and detection of guided waves, the laser ultrasonics technique was preferred. It is a broadband and non contact method which doesn't imply the use of coupling medium and which can be adapted to complex geometries. To take full advantage of this technique, it has been combined with neural networks in order to solve the inverse problem posed by the propagation of guided waves. As a result, an original, e cient and polyvalent characterization method has been obtained, which allowed us to determine the geometric properties and / or the elastic parameters of di erent plate-like structures. Structures made of silicon have been studied with this method. Finite element simulations and studies concerning the in uence of defects, including adhesion, on the waves propagation are also presented
El, Achkar Roger. "Contribution à l'étude et à la validation expérimentale de commandes neuronales d'un palier magnétique actif." Compiègne, 2008. http://www.theses.fr/2008COMP1747.
Full textThe active magnetic bearing (AMB) presents a solution for technical proble since it ensures the total levitation of a body in space eliminating any mechanical contact between the rotor and the stator. The goal of o work is to show that the control of the AMB by Multilayer perceptions (MLP) involves an improvement of the responses compared to the non linear control of the AMB by classical controllers. Two methods with MLP were developed to control the AMB. The first consists in adding the MLP in order to stabilize the system around the desired answers. This method is also used to reduce the value of the unbalance. In the second method, the MLP controls the parameters of the PID in order to minimize the oscillations of the answers obtained with the previous method. This tuning, by neural network, of the parameters of the PID controller reduces the consumption of the energy used by the AMB. The last section is devoted to the simulation of these two methods and the implementation of the MLP in real time application on an active magnetic bearing at the Heudiasyc laboratory of the UTC
Bedecarrats, Thomas. "Etude et intégration d’un circuit analogique, basse consommation et à faible surface d'empreinte, de neurone impulsionnel basé sur l’utilisation du BIMOS en technologie 28 nm FD-SOI." Thesis, Université Grenoble Alpes (ComUE), 2019. http://www.theses.fr/2019GREAT045.
Full textWhile Moore’s law reaches its limits, microelectronics actors are looking for new paradigms to ensure future developments of our information society. Inspired by biologic nervous systems, neuromorphic engineering is providing new perspectives which have already enabled breakthroughs in artificial intelligence. To achieve sufficient performances to allow their spread, neural processors have to integrate neuron circuits as small and as low power(ed) as possible so that artificial neural networks they implement reach a critical size. In this work, we show that it is possible to reduce the number of components necessary to design an analogue spiking neuron circuit thanks to the functionalisation of parasitic generation currents in a BIMOS transistor integrated in 28 nm FD-SOI technology and sized with the minimum dimensions allowed by this technology. After a systematic characterization of the FD-SOI BIMOS currents under several biases through quasi-static measurements at room temperature, a compact model of this component, adapted from the CEA-LETI UTSOI one, is proposed. The BIMOS-based leaky, integrate-and-fire spiking neuron (BB-LIF SN) circuit is described. Influence of the different design and bias parameters on its behaviour observed during measurements performed on a demonstrator fabricated in silicon is explained in detail. A simple analytic model of its operating boundaries is proposed. The coherence between measurement and compact simulation results and predictions coming from the simple analytic model attests to the relevance of the proposed analysis. In its most successful achievement, the BB-LIF SN circuit is 15 µm², consumes around 2 pJ/spike, triggers at a rate between 3 and 75 kHz for 600 pA to 25 nA synaptic currents under a 3 V power supply
Sicard, Pascal. "Nouvelles méthodes de synthèse logique." Phd thesis, Grenoble INPG, 1988. http://tel.archives-ouvertes.fr/tel-00327269.
Full textNadjar, Yann. "Susd2 et Susd4 sont deux nouveaux gènes codant pour des protéines avec domaines CCP (Complement Control Protein) jouant un rôle dans plusieurs étapes du développement des circuits neuronaux au sein de cultures d'hippocampe de rat." Thesis, Paris 6, 2014. http://www.theses.fr/2014PA066664/document.
Full textDuring brain development, several steps precisely coordinated lead to establishment of a functional neuronal network. Many molecules participate to this process, including adhesion proteins mediating interactions between neurons and their environment. Involvement of numerous genes coding for adhesion proteins in neuropsychiatric diseases such as autism argue for usefulness of identifying new ones. During my PhD, I characterized two new genes, Sud2 and Susd4, coding for proteins containing CCP domains (Complement Control Protein), classically described in proteins involved in Complement regulation system. Recently, in mammals, CCP containing proteins were shown to be involved in neuronal development. Identification of several predicted CCP containing proteins without a known function prompted me to characterize Susd2 and Susd4 which are part of them.Susd2 is expressed in neurons from hippocampal cell cultures. Its peak of expression takes place in early post natal period, suggesting a developmental function. Susd2 recombinant protein has a diffuse neuronal localization, but is particularly enriched in excitatory synapses. Decreased expression of Susd2 leads to decreased axonal growth, increased dendritic growth, and specific inhibition of excitatory synaptogenesis. Susd4 is also expressed in neurons, with a peak of expression during embryonic development, and seems to act as a regulator of dendritic growth
Dutto, Rémy. "Méthode à deux niveaux et préconditionnement géométrique en contrôle optimal. Application au problème de répartition de couple des véhicules hybrides électriques." Electronic Thesis or Diss., Université de Toulouse (2023-....), 2024. http://www.theses.fr/2024TLSEP088.
Full textMotivated by the torque split and gear shift industrial problem of hybrid electric vehicles, this work mainly proposes two new indirect optimal control problem methods. The first one is the Macro-Micro method, which is based on a bilevel decomposition of the optimal control problem and uses Bellman’s value functions at fixed times. These functions are known to be difficult to create. The main idea of this method is to approximate these functions by neural networks, which leads to a hierarchical resolution of a low dimensional optimization problem and a set of independent optimal control problems defined on smaller time intervals. The second one is a geometric preconditioning method, which allows a more efficient resolution of the optimal control problem. This method is based on a geometrical interpretation of the Pontryagin’s co-state and on the Mathieu transformation, and uses a linear diffeomorphism which transforms an ellipse into a circle. These two methods, presented separately, can be combined and lead together to a fast, robust and light resolution for the torque split and gear shift optimal control problem, closer to the embedded requirements
Falez, Pierre. "Improving spiking neural networks trained with spike timing dependent plasticity for image recognition." Thesis, Lille 1, 2019. http://www.theses.fr/2019LIL1I101.
Full textComputer vision is a strategic field, in consequence of its great number of potential applications which could have a high impact on society. This area has quickly improved over the last decades, especially thanks to the advances of artificial intelligence and more particularly thanks to the accession of deep learning. Nevertheless, these methods present two main drawbacks in contrast with biological brains: they are extremely energy intensive and they need large labeled training sets. Spiking neural networks are alternative models offering an answer to the energy consumption issue. One attribute of these models is that they can be implemented very efficiently on hardware, in order to build ultra low-power architectures. In return, these models impose certain limitations, such as the use of only local memory and computations. It prevents the use of traditional learning methods, for example the gradient back-propagation. STDP is a learning rule, observed in biology, which can be used in spiking neural networks. This rule reinforces the synapses in which local correlations of spike timing are detected. It also weakens the other synapses. The fact that it is local and unsupervised makes it possible to abide by the constraints of neuromorphic architectures, which means it can be implemented efficiently, but it also provides a solution to the data set labeling issue. However, spiking neural networks trained with the STDP rule are affected by lower performances in comparison to those following a deep learning process. The literature about STDP still uses simple data but the behavior of this rule has seldom been used with more complex data, such as sets made of a large variety of real-world images.The aim of this manuscript is to study the behavior of these spiking models, trained through the STDP rule, on image classification tasks. The main goal is to improve the performances of these models, while respecting as much as possible the constraints of neuromorphic architectures. The first contribution focuses on the software simulations of spiking neural networks. Hardware implementation being a long and costly process, using simulation is a good alternative in order to study more quickly the behavior of different models. Then, the contributions focus on the establishment of multi-layered spiking networks; networks made of several layers, such as those in deep learning methods, allow to process more complex data. One of the chapters revolves around the matter of frequency loss seen in several spiking neural networks. This issue prevents the stacking of multiple spiking layers. The center point then switches to a study of STDP behavior on more complex data, especially colored real-world image. Multiple measurements are used, such as the coherence of filters or the sparsity of activations, to better understand the reasons for the performance gap between STDP and the more traditional methods. Lastly, the manuscript describes the making of multi-layered networks. To this end, a new threshold adaptation mechanism is introduced, along with a multi-layer training protocol. It is proven that such networks can improve the state-of-the-art for STDP
Peltier, Marie-Agnès. "Un système adaptatif de diagnostic d'évolution basé sur la reconnaissance des formes floues : application au diagnostic du comportement d'un conducteur automobile." Compiègne, 1993. http://www.theses.fr/1993COMPD634.
Full textBouvier, Louis. "Apprentissage structuré et optimisation combinatoire : contributions méthodologiques et routage d'inventaire chez Renault." Electronic Thesis or Diss., Marne-la-vallée, ENPC, 2024. http://www.theses.fr/2024ENPC0046.
Full textThis thesis stems from operations research challenges faced by Renault supply chain. Toaddress them, we make methodological contributions to the architecture and training of neural networks with combinatorial optimization (CO) layers. We combine them with new matheuristics to solve Renault’s industrial inventory routing problems.In Part I, we detail applications of neural networks with CO layers in operations research. We notably introduce a methodology to approximate constraints. We also solve some off- policy learning issues that arise when using such layers to encode policies for Markov decision processes with large state and action spaces. While most studies on CO layers rely on supervised learning, we introduce a primal-dual alternating minimization scheme for empirical risk minimization. Our algorithm is deep learning-compatible, scalable to large combinatorial spaces, and generic. In Part II, we consider Renault European packaging return logistics. Our rolling-horizon policy for the operational-level decisions is based on a new large neighborhood search for the deterministic variant of the problem. We demonstrate its efficiency on large-scale industrialinstances, that we release publicly, together with our code and solutions. We combine historical data and experts’ predictions to improve performance. A version of our policy has been used daily in production since March 2023. We also consider the tactical-level route contracting process. The sheer scale of this industrial problem prevents the use of classic stochastic optimization approaches. We introduce a new algorithm based on methodological contributions of Part I for empirical risk minimization
Wei, Yan. "Planification et Suivi de Mouvement d’un Système de Manipulateur Mobile non-holonome à deux bras." Thesis, Ecole centrale de Lille, 2018. http://www.theses.fr/2018ECLI0004/document.
Full textThis thesis focuses on the motion planning and tracking of a dual-arm mobile humanoid. First, MDH is used for kinematic modeling. The co-simulation via Simulink-Adams on prototype is realized to validate the effectiveness of RBFNN controller. In order to overcome the shortcomings of Euler-Lagrange’s formulations that require calculating energy and energy derivatives, Kane’s method is used. In addition, physical stability is analyzed based on Kane’s method and a controller is designed using back-stepping technique. Secondly, an improved MaxiMin NSGA-II is proposed to design the mobile base’s (MB) optimal position-orientation and the upper manipulator’s (UM) optimal configuration given only the initial pose and end-effectors’ (EEs) desired positions-orientations. A direct connect algorithm combining BiRRT and gradient-descent is designed to plan the transition from initial pose to optimal pose, and a geometric optimization method is designed to optimize and cohere the path. In addition, forward motions are obtained by assigning orientations for MB thus indicating robot’s intention. In order to solve the failure problem of offline algorithm, an online algorithm is proposed while estimating dynamic obstacles’ motions. In addition, in order to optimize via-poses, an algorithm based on EEs’ via-points and MOGA is proposed by optimizing four via-pose-based objective functions. Finally, the motion tracking problem is studied given EEs’ motions in the task space. Instead of controlling the absolute motion, two relative motions are introduced to realize the coordination and cooperation between MB and UM. In addition, an modulated WLN technique is proposed to avoid joints’ limits
Ak, Ronay. "Neural Network Modeling for Prediction under Uncertainty in Energy System Applications." Thesis, Supélec, 2014. http://www.theses.fr/2014SUPL0015/document.
Full textThis Ph.D. work addresses the problem of prediction within energy systems design and operation problems, and particularly the adequacy assessment of renewable power generation systems. The general aim is to develop an empirical modeling framework for providing predictions with the associated uncertainties. Along this research direction, a non-parametric, empirical approach to estimate neural network (NN)-based prediction intervals (PIs) has been developed, accounting for the uncertainty in the predictions due to the variability in the input data and the system behavior (e.g. due to the stochastic behavior of the renewable sources and of the energy demand by the loads), and to model approximation errors. A novel multi-objective framework for estimating NN-based PIs, optimal in terms of both accuracy (coverage probability) and informativeness (interval width) is proposed. Ensembling of individual NNs via two novel approaches is proposed as a way to increase the performance of the models. Applications on real case studies demonstrate the power of the proposed framework
Van, Der Baan Mirko. "Deux méthodes d'inférence statistique appliquées aux données de sismique réflexion profonde : détection de signaux et localisation d'onde." Phd thesis, 1999. http://tel.archives-ouvertes.fr/tel-00745500.
Full text