Dissertations / Theses on the topic 'Apprentissage par modèle'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Apprentissage par modèle.'
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.
Liang, Ke. "Oculométrie Numérique Economique : modèle d'apparence et apprentissage par variétés." Thesis, Paris, EPHE, 2015. http://www.theses.fr/2015EPHE3020/document.
Gaze tracker offers a powerful tool for diverse study fields, in particular eye movement analysis. In this thesis, we present a new appearance-based real-time gaze tracking system with only a remote webcam and without infra-red illumination. Our proposed gaze tracking model has four components: eye localization, eye feature extraction, eye manifold learning and gaze estimation. Our research focuses on the development of methods on each component of the system. Firstly, we propose a hybrid method to localize in real time the eye region in the frames captured by the webcam. The eye can be detected by Active Shape Model and EyeMap in the first frame where eye occurs. Then the eye can be tracked through a stochastic method, particle filter. Secondly, we employ the Center-Symmetric Local Binary Patterns for the detected eye region, which has been divided into blocs, in order to get the eye features. Thirdly, we introduce manifold learning technique, such as Laplacian Eigen-maps, to learn different eye movements by a set of eye images collected. This unsupervised learning helps to construct an automatic and correct calibration phase. In the end, as for the gaze estimation, we propose two models: a semi-supervised Gaussian Process Regression prediction model to estimate the coordinates of eye direction; and a prediction model by spectral clustering to classify different eye movements. Our system with 5-points calibration can not only reduce the run-time cost, but also estimate the gaze accurately. Our experimental results show that our gaze tracking model has less constraints from the hardware settings and it can be applied efficiently in different real-time applications
Israilov, Sardor. "De l'identification basée apprentissage profond à la commande basée modèle." Electronic Thesis or Diss., Université Côte d'Azur, 2024. http://www.theses.fr/2024COAZ4003.
Fish swimming remains a complex subject that is not yet fully understood due to the inter-section of biology and fluid dynamics. Through years of evolution, organisms in nature have perfected their biological mechanisms to navigate efficiently in their environment and adaptto particular situations. Throughout history, mankind has been inspired by nature to innovateand develop nature-like systems. Biomimetic robotic fish, in particular, has a number of appli-cations in the real world and its control is yet to be optimized. Deep Reinforcement Learning showed excellent results in control of robotic systems, where dynamics is too complex to befully modeled and analyzed. In this thesis, we explored new venues of control of a biomimetic fish via reinforcement learning to effectively maximize the thrust and speed. However, to fully comprehend the newly-emerged data-based algorithms, we first studied the application of these methods on a standard benchmark of a control theory, the inverted pendulum with a cart. We demonstrated that deep Reinforcement Learning could control the system without any prior knowledge of the system, achieving performance comparable to traditional model-based con-trol theory methods. In the third chapter, we focus on the undulatory swimming of a roboticfish, exploring various objectives and information sources for control. Our studies indicate that the thrust force of a robotic fish can be optimized using inputs from both force sensors and cameras as feedback for control. Our findings demonstrate that a square wave control with a particular frequency maximizes the thrust and we rationalize it using Pontryagin Maximum Principle. An appropriate model is established that shows an excellent agreement between simulation and experimental results. Subsequently, we concentrate on the speed maximization of a robotic fish both in several virtual environments and experiments using visual data. Once again, we find that deep Reinforcement Learning can find an excellent swimming gait with a square wave control that maximizes the swimming speed
Goëau, Hervé. "Structuration de collections d'images par apprentissage actif crédibiliste." Phd thesis, Université Joseph Fourier (Grenoble), 2009. http://tel.archives-ouvertes.fr/tel-00410380.
Goëau, Hervé. "Structuration de collections d'images par apprentissage actif crédibiliste." Phd thesis, Grenoble 1, 2009. http://www.theses.fr/2009GRE10070.
Image annotation is an essential task in professional archives exploitation. Archivsits must describe every image in order to make easier future retrieval tasks. The main difficulties are how to interpret the visual contents, how to bring together images whitch can be associated in same categories, and how to deal with the user's subjectivity. In this thesis, we use the principle of active learning in order to help a user who wants organize with accuracy image collections. From the visual content analysis, complementary active learning strategies are proposed to the user to help him to identify and put together images in relevant categories according to his oppinion. We choose to express this image classification problem with active learning by using the Transferable Belief Model (TBM), an elaboration on the Dempster-Shafer theory of evidence. The TBM allows the combination, the revision and the representation of the knowledge which can be extracted from the visual contents and the previously identified categories. Our method proposed in this theoritical framework gives a detailed modeling of the knowledge by representing explicitly cases of multi-labeling, while quantifying uncertainty (related to the semantic gap) and conflict induced by the analysis of the visual content in different modalities (colors, textures). A human-machine interface was developed in order to validate our approach on reference tests, personal images collections and professional photos from the National Audiovisual Institute. An evaluation was driven with professional users and showed very positive results in terms of utility, of usability and satisfaction
Delcourt, Alexandre. "Amélioration des détecteurs CdZnTe pour l'imagerie gamma par apprentissage." Electronic Thesis or Diss., Université Grenoble Alpes, 2023. http://www.theses.fr/2023GRALM056.
Since a few years, the wide spread use of CZT-based detectors in gamma imaging drives their performance optimization to stay competitive at the industrial level. However, the presence of structural defect in the CZT crystal deteriorates the output signals quality and holds back the higher volume detectors development.The purpose of this thesis is the use of optimization and artificial intelligence algorithms using realistic simulations to override the impact of the defects and improve the localization performances of gamma interactions in the detector. We will develop a mathematical-based method in three steps as an alternative to common characterization and correction methods.First, we develop 3D CZT detector simulations enabling to implement defects with different natures to observe their impact on output signals. Then we build a simple neural network, which can be introduced in the electronics to localize the gamma interactions in the detector from simulation results. A second network based on a gradient computation method will allow determining the electric field and collection performance of a detector.The addition of these three steps will be used to learn through simulation the intern parameters of a determined detector such as the electric field. This simulation will serve to train the simple neural network and finally be used on experimental data to improve the localization performance of the detector.The development of this mathematical approach will help us having a better understanding of the intern structure of a CZT crystal being able to reproduce its behavior in simulation. In addition, the better performance of the detector might be sufficient to decrease the radiotracer dose for medical imaging or limit the exposition time of operators in a nuclear power plant
Zhang, Jian. "Modèles de Mobilité de Véhicules par Apprentissage Profond dans les Systèmes de Tranport Intelligents." Thesis, Ecole centrale de Lille, 2018. http://www.theses.fr/2018ECLI0015/document.
The intelligent transportation systems gain great research interests in recent years. Although the realistic traffic simulation plays an important role, it has not received enough attention. This thesis is devoted to studying the traffic simulation in microscopic level, and proposes corresponding vehicular mobility models. Using deep learning methods, these mobility models have been proven with a promising credibility to represent the vehicles in real-world. Firstly, a data-driven neural network based mobility model is proposed. This model comes from real-world trajectory data and allows mimicking local vehicle behaviors. By analyzing the performance of this basic learning based mobility model, we indicate that an improvement is possible and we propose its specification. An HMM is then introduced. The preparation of this integration is necessary, which includes an examination of traditional dynamics based mobility models and the adaptation method of “classical” models to our situation. At last, the enhanced model is presented, and a sophisticated scenario simulation is built with it to validate the theoretical results. The performance of our mobility model is promising and implementation issues have also been discussed
Badets, Arnaud. "Apprentissage moteur par observation d'un sujet modèle : rôles de la connaissance du résultat et de l'intention." Poitiers, 2004. http://www.theses.fr/2004POIT2263.
This thesis focus on the role of observational learning in motor learning and includes four studies on the role of the knowledge of results (KR) and one study on the role of the intention to reproduce the observed behavior. More precisely, the three first studies assess the role of the relative KR frequency during observational learning. Experiments 1 and 2 indicate that a low relative KR frequency (e. G. , 33%) enhances the learning of the parametric components of the task and the ability to detect errors in the model and in the participant own performance. The third experiment shows that a low KR frequency during observation enhances the learning of the invariant components of the task. The fourth experiment shows a positive effect of bandwidth KR during observational learning when compared to a condition that control the KR frequency. Finally, the last experiment indicates that intention to reproduce the observed behavior enhances the learning of the invariant components of the task. Results of those five studies on physical practice and on observation are discussed in light of recent suggestions from behavioral and neurophysiological studies (Shea et al. , 2000; Jeannerod, 2003)
Martinet, Louis-Emmanuel. "Apprentissage spatial et planification de l'action dans un modèle de réseau neuronal inspiré du cortex préfrontal." Phd thesis, Université Pierre et Marie Curie - Paris VI, 2010. http://tel.archives-ouvertes.fr/tel-00646738.
Rebière, Thérèse. "Apprentissage par l'expérience, formation initiale et recherche d'emploi des salariés." Le Havre, 2009. http://www.theses.fr/2009LEHA0014.
This thesis studies the consequences, on the labor market, of the on-the-job search process of workers, in a matching model capturing workers’ career path. The labor market is hierarchically segmented in two sectors : the low-skill sector, in which workers learn by doing, and the better-paying high-skill sector that workers wish to reach. The addressed topics are the followings : the efficiency of the labor market, a self-financed tax and subsidy policy, the minimum wage effects, educational choices
Brenon, Alexis. "Modèle profond pour le contrôle vocal adaptatif d'un habitat intelligent." Thesis, Université Grenoble Alpes (ComUE), 2017. http://www.theses.fr/2017GREAM057/document.
Smart-homes, resulting of the merger of home-automation, ubiquitous computing and artificial intelligence, support inhabitants in their activity of daily living to improve their quality of life.Allowing dependent and aged people to live at home longer, these homes provide a first answer to society problems as the dependency tied to the aging population.In voice controlled home, the home has to answer to user's requests covering a range of automated actions (lights, blinds, multimedia control, etc.).To achieve this, the control system of the home need to be aware of the context in which a request has been done, but also to know user habits and preferences.Thus, the system must be able to aggregate information from a heterogeneous home-automation sensors network and take the (variable) user behavior into account.The development of smart home control systems is hard due to the huge variability regarding the home topology and the user habits.Furthermore, the whole set of contextual information need to be represented in a common space in order to be able to reason about them and make decisions.To address these problems, we propose to develop a system which updates continuously its model to adapt itself to the user and which uses raw data from the sensors through a graphical representation.This new method is particularly interesting because it does not require any prior inference step to extract the context.Thus, our system uses deep reinforcement learning; a convolutional neural network allowing to extract contextual information and reinforcement learning used for decision-making.Then, this memoir presents two systems, a first one only based on reinforcement learning showing limits of this approach against real environment with thousands of possible states.Introduction of deep learning allowed to develop the second one, ARCADES, which gives good performances proving that this approach is relevant and opening many ways to improve it
Tmar, Mohamed. "Modèle auto-adaptatif de filtrage d'information : apprentissage incrémental du profil et de la fonction de décision." Toulouse 3, 2002. http://www.theses.fr/2002TOU30081.
Nikoulina, Vassilina. "Modèle de traduction statistique à fragments enrichi par la syntaxe." Phd thesis, Grenoble, 2010. http://www.theses.fr/2010GRENM008.
Traditional Statistical Machine Translation models are not aware of linguistic structure. Thus, target lexical choices and word order are controlled only by surface-based statistics learned from the training corpus. However, knowledge of linguistic structure can be beneficial since it provides generic information compensating data sparsity. The purpose of our work is to study the impact of syntactic information while preserving the general framework of Phrase-Based SMT. First, we study the integration of syntactic information using a reranking approach. We define features measuring the similarity between the dependency structures of source and target sentences, as well as features of linguistic coherence of the target sentences. The importance of each feature is assessed by learning their weights through a Structured Perceptron Algorithm. The evaluation of several reranking models shows that these features often improve the quality of translations produced by the basic model, in terms of manual evaluations as opposed to automatic measures. Then, we propose different models in order to increase the quality and diversity of the search graph produced by the decoder, through filtering out uninteresting hypotheses based on the source syntactic structure. This is done either by learning limits on the phrase recordering, or by decomposing the source sentence in order to simplify the translation process. The initial evaluations of these models look promising
Nikoulina, Vassilina. "Modèle de traduction statistique à fragments enrichi par la syntaxe." Phd thesis, Université de Grenoble, 2010. http://tel.archives-ouvertes.fr/tel-00996317.
Dollé, Laurent. "Contribution d'un modèle computationnel de sélection de stratégies de navigation aux hypothèses relatives à l'apprentissage spatial." Phd thesis, Paris 6, 2010. http://www.theses.fr/2010PA066407.
Monari, Gaétan. "Sélection de modèles non linéaires par "leave-one-out": étude théorique et application des réseaux de neurones au procédé de soudage par points." Phd thesis, Université Pierre et Marie Curie - Paris VI, 1999. http://pastel.archives-ouvertes.fr/pastel-00000676.
Mountassir, Mahjoub El. "Surveillance d'intégrité des structures par apprentissage statistique : application aux structures tubulaires." Thesis, Université de Lorraine, 2019. http://www.theses.fr/2019LORR0047.
To ensure better working conditions of civil and engineering structures, inspections must be made on a regular basis. However, these inspections could be labor-intensive and cost-consuming. In this context, structural health monitoring (SHM) systems using permanently attached transducers were proposed to ensure continuous damage diagnostic of these structures. In SHM, damage detection is generally based on comparison between the healthy state signals and the current signals. Nevertheless, the environmental and operational conditions will have an effect on the healthy state signals. If these effects are not taken into account they would result in false indication of damage (false alarm). In this thesis, classical machine learning methods used for damage detection have been applied in the case of pipelines. The effects of some measurements parameters on the robustness of these methods have been investigated. Afterthat, two approaches were proposed for damage diagnostic depending on the database of reference signals. If this database contains large variation of these EOCs, a sparse estimation of the current signal is calculated. Then, the estimation error is used as an indication of the presence of damage. Otherwise, if this database is acquired at limited range of EOCs, moving window PCA can be applied to update the model of the healthy state provided that the EOCs show slow and continuous variation. In both approaches, damage localization was ensured using a sliding window over the damaged pipe signal
Silberzahn, Philippe. "La détermination par la firme entrepreneuriale de ses produits et marchés : un modèle socio-cognitif." Phd thesis, Ecole Polytechnique X, 2009. http://pastel.archives-ouvertes.fr/pastel-00005103.
Mountassir, Mahjoub El. "Surveillance d'intégrité des structures par apprentissage statistique : application aux structures tubulaires." Electronic Thesis or Diss., Université de Lorraine, 2019. http://docnum.univ-lorraine.fr/ulprive/DDOC_T_2019_0047_EL_MOUNTASSIR.pdf.
To ensure better working conditions of civil and engineering structures, inspections must be made on a regular basis. However, these inspections could be labor-intensive and cost-consuming. In this context, structural health monitoring (SHM) systems using permanently attached transducers were proposed to ensure continuous damage diagnostic of these structures. In SHM, damage detection is generally based on comparison between the healthy state signals and the current signals. Nevertheless, the environmental and operational conditions will have an effect on the healthy state signals. If these effects are not taken into account they would result in false indication of damage (false alarm). In this thesis, classical machine learning methods used for damage detection have been applied in the case of pipelines. The effects of some measurements parameters on the robustness of these methods have been investigated. Afterthat, two approaches were proposed for damage diagnostic depending on the database of reference signals. If this database contains large variation of these EOCs, a sparse estimation of the current signal is calculated. Then, the estimation error is used as an indication of the presence of damage. Otherwise, if this database is acquired at limited range of EOCs, moving window PCA can be applied to update the model of the healthy state provided that the EOCs show slow and continuous variation. In both approaches, damage localization was ensured using a sliding window over the damaged pipe signal
Gaillard, Pierre. "Apprentissage statistique de la connexité d'un nuage de points par modèle génératif : application à l'analyse exploratoire et la classification semi-supervisée." Compiègne, 2008. http://www.theses.fr/2008COMP1767.
In this work, we propose a statistical model to learn the connectedness of a set of points. This model combine geometrical and statistical approaches by defining a mixture model based on a graph. From this generative graph, we propose and evaluate methods and algorithms to analyse the set of points and to realize semi-supervised learning
Decoux, Benoît. "Un modèle connexionniste de vision 3-D : imagettes rétiniennes, convergence stéréoscopique, et apprentissage auto-supervisé de la fusion." Rouen, 1995. http://www.theses.fr/1995ROUES056.
Cottret, Maxime. "Exploration visuelle d'environnement intérieur par détection et modélisation d'objets saillants." Phd thesis, Institut National Polytechnique de Toulouse - INPT, 2007. http://tel.archives-ouvertes.fr/tel-00289380.
Philippeau, Jérémy. "Apprentissage de similarités pour l'aide à l'organisation de contenus audiovisuels." Toulouse 3, 2009. http://thesesups.ups-tlse.fr/564/.
In the perspective of new usages in the field of the access to audiovisual archives, we have created a semi-automatic system that helps a user to organize audiovisual contents while performing tasks of classification, characterization, identification and ranking. To do so, we propose to use a new vocabulary, different from the one already available in INA documentary notices, to answer needs which can not be easily defined with words. We have conceived a graphical interface based on graph formalism designed to express an organisational task. The digital similarity is a good tool in respect with the handled elements which are informational objects shown on the computer screen and the automatically extracted audio and video low-level features. We have made the choice to estimate the similarity between those elements with a predictive process through a statistical model. Among the numerous existing models, the statistical prediction based on the univaried regression and on support vectors has been chosen. H)
Amour, Lamine. "Mise en oeuvre d’un modèle adaptatif pour l’estimation de la qualité réelle perçue par l’usager (QoE) : application aux services mobiles." Thesis, Paris Est, 2018. http://www.theses.fr/2018PESC0043.
The general context of the thesis is the development of adaptive mechanisms that uses information from a given environment and context (Quality of Experience: QoE) and adapting it in order to initiate a specific corrective actions following the occurrence of undesirable events such as an unsatisfactory Quality of Service (QoS), negative feedback or malfunctions of network elements. To do this, the main objective of this thesis is to develop a new network management by moving from a "network centred" view, where only parameters from the network itself were used, to a "user centred" view. This view consists in considering all the factors that can impact the end-user's experience. The work developed as part of this thesis focused on studying these factors on all components of the processing and transport chain (user, peripheral, application, network elements) and measuring their impact in estimating user QoE. The use case considered concerns approaches to the dissemination of adaptive bit rate streaming (ABR). The thesis work led to the proposal of a new solution for the adaptive bit rate streaming operating in a closed loop and sensitive to QoE. Such a solution allows to optimize in real time the quality of the video according to the instantaneous measurement of the estimated QoE
Motta, Jesus Antonio. "VENCE : un modèle performant d'extraction de résumés basé sur une approche d'apprentissage automatique renforcée par de la connaissance ontologique." Doctoral thesis, Université Laval, 2014. http://hdl.handle.net/20.500.11794/26076.
Several methods and techniques of artificial intelligence for information extraction, pattern recognition and data mining are used for extraction of summaries. More particularly, new machine learning models with the introduction of ontological knowledge allow the extraction of the sentences containing the greatest amount of information from a corpus. This corpus is considered as a set of sentences on which different optimization methods are applied to identify the most important attributes. They will provide a training set from which a machine learning algorithm will can abduce a classification function able to discriminate the sentences of new corpus according their information content. Currently, even though the results are interesting, the effectiveness of models based on this approach is still low, especially in the discriminating power of classification functions. In this thesis, a new model based on this approach is proposed and its effectiveness is improved by inserting ontological knowledge to the training set. The originality of this model is described through three papers. The first paper aims to show how linear techniques could be applied in an original way to optimize workspace in the context of extractive summary. The second article explains how to insert ontological knowledge to significantly improve the performance of classification functions. This introduction is performed by inserting lexical chains of ontological knowledge based in the training set. The third article describes VENCE , the new machine learning model to extract sentences with the most information content in order to produce summaries. An assessment of the VENCE performance is achieved comparing the results with those produced by current commercial and public software as well as those published in very recent scientific articles. The use of usual metrics recall, precision and F_measure and the ROUGE toolkit showed the superiority of VENCE. This model could benefit other contexts of information extraction as for instance to define models for sentiment analysis.
Dollé, Laurent. "Contribution d'un modèle computationnel de sélection de stratégies de navigation aux hypothèses relatives à l'apprentissage spatial." Phd thesis, Université Pierre et Marie Curie - Paris VI, 2010. http://tel.archives-ouvertes.fr/tel-00647199.
Ngandeu, Joseph blaise. "Apprentissage du français dans une université anglophone au Cameroun : de l’expérience du quasi-synchrone à un nouveau modèle d’intégration des TIC." Thesis, Clermont-Ferrand 2, 2015. http://www.theses.fr/2015CLF20023/document.
Information and Communication Technologies are now part of the life of many institutions of higher education, secondary education and even primary education. The benefits for these institutions, as well as for teachers and learners are established. However, there are very few cases of experiments carried out in learning Institutions with positive results. This thesis is about integrating ICTs into a French course in the Anglo-Saxon university of Buea in Cameroon. The context in which the course is taught in classrooms makes it difficult for the objectives to be met. Because of the large class size, the little time allocated to the course and the heterogeneous nature of students’ language levels, speaking, writing and interactional skills are not worked out in class. This action research thesis sets out to propose a blended learning set-up that hinges around the university IT Centre. Online classes are articulated with classroom sessions. Thus, students have the opportunity to involve in small groups, in communicative activities and in quasi synchronous mode. These communication modality, very often less studied by CMC researchers, is an efficient alternative to synchronous communication and environments that are very much reliable as far as internet quality is concerned in that part of the world. The general question that has driven the research is: How can a technology instrumented approach help in overcoming difficulties and ease learning? To answer that question, two technology based experiences were carried out. They were guided by a research protocol. Data collected enable me to analyse interactions and determine traces of language acquisition. From the circumstances surrounding the experiments, it is discovered that there are a number of obstacles to the integration of ICT in “low tech context” like that which is studied. The technology-based stet-up pattern is then questioned. This research work goes further to propose mobile learning as an alternative to the ICT integration model that concentrates all the resources in a single location
Schwint, Didier. "Le savoir artisan de fabrication et le modèle de la mètis : exemple des tourneurs et tabletiers sur bois du Jura." Poitiers, 1999. http://www.theses.fr/1999POIT5001.
Maza, Elie. "Prévision de trafic routier par des méthodes statistiques : espérance structurelle d’une fonction aléatoire." Toulouse 3, 2004. http://www.theses.fr/2004TOU30238.
In the first part of this thesis, we describe a travel time forecasting method on the Parisian motorway network. This method is based on a mixture model. Parameters are estimated by an automatic classification method and a training concept. The second part is devoted to the study of a semi-parametric curve translation model. Estimates are carried out by an M-estimation method. We show the consistency and the asymptotic normality of the estimators. In the third part, we widen the function warping model by considering that the warping functions result from a random process. That enables us to define, in an intrinsic way, a concept of structural expectation and thus to get round the non identifiability of the model. We propose an empirical estimator of this structural expectation and we show consistency and asymptotic normality
Bonidal, Rémi. "Sélection de modèle par chemin de régularisation pour les machines à vecteurs support à coût quadratique." Electronic Thesis or Diss., Université de Lorraine, 2013. http://www.theses.fr/2013LORR0066.
Model selection is of major interest in statistical learning. In this document, we introduce model selection methods for bi-class and multi-class support vector machines. We focus on quadratic loss machines, i.e., machines for which the empirical term of the objective function of the learning problem is a quadratic form. For SVMs, model selection consists in finding the optimal value of the regularization coefficient and choosing an appropriate kernel (or the values of its parameters). The proposed methods use path-following techniques in combination with new model selection criteria. This document is structured around three main contributions. The first one is a method performing model selection through the use of the regularization path for the l2-SVM. In this framework, we introduce new approximations of the generalization error. The second main contribution is the extension of the first one to the multi-category setting, more precisely the M-SVM². This study led us to derive a new M-SVM, the least squares M-SVM. Additionally, we present new model selection criteria for the M-SVM introduced by Lee, Lin and Wahba (and thus the M-SVM²). The third main contribution deals with the optimization of the values of the kernel parameters. Our method makes use of the principle of kernel-target alignment with centered kernels. It extends it through the introduction of a regularization term. Experimental validation of these methods was performed on classical benchmark data, toy data and real-world data
Hanna, Elias. "Improving Novelty Search Sample Efficiency with Models or Archive Bootstrapping." Electronic Thesis or Diss., Sorbonne université, 2024. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2024SORUS038.pdf.
In the context of policy search for robotic systems, being sample-efficient is a most. Evolutionary Algorithms have been used in the past ten years to achieve significant results in the robotics domain as their Darwinist approach to optimization allows them to bypass problems often encountered by gradient-based optimization methods like Reinforcement Learning. Nevertheless, such methods remain sample greedy and almost impossible to transfer directly onto a real robotic system. This Ph.D thesis takes interest in solving that sample-efficiency problem, especially for the Novelty Search algorithm, a novelty-driven Evolutionary Algorithm. Incorporating learned models in the optimization process has been a solution towards sample-efficiency for many years, but few works address this within the Novelty Search framework. Three research axis within that framework are explored in this manuscript. Firstly, the impact of pre-training the learned model with data gathered using random processes of varying time-correlation is evaluated. It is shown that the impact is negligible on a state-of-the-art model-based Evolutionary Algorithm, but that it is significant on a model-based Reinforcement Learning algorithm with returns with up to ten-fold differences between the best and worst random process used. Secondly, a preliminary study is made on a new approach aiming at biasing the initial population of the Novelty Search algorithm towards a more behavioraly diverse one using random dynamics models ensembles. It is shown that this approach successfully reduces the number of evaluations required by Novelty Search to cover the environment of a two-wheeled robot. It is also shown that this approach fails on a more complex locomotion environment of an hexapod robot, and the lack of diversity captured by the random models ensembles used is determined as the cause. Finally, a new model-based Evolutionary Algorithm, dubbed Model-Based Novelty Search, is proposed, with the aim of preserving the strong exploration capabilities of Novelty Search while reducing the numbers of evaluations needed to reach the same coverage of the Behavioral Space. Results on three robotic tasks show a reduction in sample usage compared to Novelty Search of 30% up to 75% depending on the considered task, a significant advance towards a more sample-efficient Novelty Search algorithm
Bonidal, Rémi. "Sélection de modèle par chemin de régularisation pour les machines à vecteurs support à coût quadratique." Thesis, Université de Lorraine, 2013. http://www.theses.fr/2013LORR0066/document.
Model selection is of major interest in statistical learning. In this document, we introduce model selection methods for bi-class and multi-class support vector machines. We focus on quadratic loss machines, i.e., machines for which the empirical term of the objective function of the learning problem is a quadratic form. For SVMs, model selection consists in finding the optimal value of the regularization coefficient and choosing an appropriate kernel (or the values of its parameters). The proposed methods use path-following techniques in combination with new model selection criteria. This document is structured around three main contributions. The first one is a method performing model selection through the use of the regularization path for the l2-SVM. In this framework, we introduce new approximations of the generalization error. The second main contribution is the extension of the first one to the multi-category setting, more precisely the M-SVM². This study led us to derive a new M-SVM, the least squares M-SVM. Additionally, we present new model selection criteria for the M-SVM introduced by Lee, Lin and Wahba (and thus the M-SVM²). The third main contribution deals with the optimization of the values of the kernel parameters. Our method makes use of the principle of kernel-target alignment with centered kernels. It extends it through the introduction of a regularization term. Experimental validation of these methods was performed on classical benchmark data, toy data and real-world data
Pinguet, Jérémy. "Contribution à la synthèse de contrôleurs neuronaux robustes par imitation." Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASG002.
This thesis focuses on developing control systems by imitating behaviors or decisions meeting complex requirements. The objective is to perform the learning of a neural controller efficiently and robustly on a database containing these behaviors.The chosen approach unifies robust control tools with those of neural network modeling. Methods for identifying dynamic systems are first developed according to neural structures in cohesion with the representations of linear systems with varying parameters. Access to this field of study opens the way to stability and performance analysis of these neural models.The work then proposes to exploit these properties to address the robustness issues inherent to the learning of control laws. The proposed method of identifying robust controllers is based on evaluating the stability margins of the neural feedback loop. It is then possible to consolidate the robustness of the controllers through a learning strategy with stability optimization by a multi-objective formulation. In addition, the deployment of the controllers is performed using a multi-model adaptive control method.The approach is finally applied to aircraft autopilots via a co-simulation with a flight simulator characterized by its high modeling reliability. The control issues addressed are, in the first step, to guide the aircraft according to a given heading and altitude, while a second experiment focuses on following a flight path consisting of a series of waypoints. The neural autopilots are developed by imitating an existing autopilot and then by imitating a pilot
Ottonelli, Claudio. "Apprentissage statistique de modèles réduits non-linéaires par approche expérimentale et design de contrôleurs robustes: le cas de la cavité ouverte." Phd thesis, Ecole Polytechnique X, 2014. http://pastel.archives-ouvertes.fr/pastel-01065782.
Aguirre, Cervantes José Luis. "Construction automatique de taxonomies à partir d'exemples dans un modèle de connaissances par objets." Grenoble INPG, 1989. http://www.theses.fr/1989INPG0067.
Madi, wamba Gilles. "Combiner la programmation par contraintes et l’apprentissage machine pour construire un modèle éco-énergétique pour petits et moyens data centers." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2017. http://www.theses.fr/2017IMTA0045/document.
Over the last decade, cloud computing technologies have considerably grown, this translates into a surge in data center power consumption. The magnitude of the problem has motivated numerous research studies around static or dynamic solutions to reduce the overall energy consumption of a data center. The aim of this thesis is to integrate renewable energy sources into dynamic energy optimization models in a data center. For this we use constraint programming as well as machine learning techniques. First, we propose a global constraint for tasks intersection that takes into account a ressource with variable cost. Second, we propose two learning models for the prediction of the work load of a data center and for the generation of such curves. Finally, we formalize the green energy aware scheduling problem (GEASP) and propose a global model based on constraint programming as well as a search heuristic to solve it efficiently. The proposed model integrates the various aspects inherent to the dynamic planning problem in a data center : heterogeneous physical machines, various application types (i.e., ractive applications and batch applications), actions and energetic costs of turning ON/OFF physical machine, interrupting/resuming batch applications, CPU and RAM ressource consumption of applications, migration of tasks and energy costs related to the migrations, prediction of green energy availability, variable energy consumption of physical machines
Verzat, Caroline. "Les logiques d'apprentissage collectif en recherche industrielle. Modèle de compréhension et de pilotage par les situations-type : recherche-action à la direction de la recherche de PSA-Peugeot-Citroën." Paris 9, 2000. https://portail.bu.dauphine.fr/fileviewer/index.php?doc=2000PA090080.
Tencé, Fabien. "Modèle probabiliste de comportement et algorithme d'apprentissage par imitation pour les personnages crédibles dans les jeux vidéo." Phd thesis, Université de Bretagne occidentale - Brest, 2011. http://tel.archives-ouvertes.fr/tel-00667072.
Ji, Hyungsuk. "Étude d'un modèle computationnel pour la représentation du sens des mots par intégration des relations de contexte." Phd thesis, Grenoble INPG, 2004. http://tel.archives-ouvertes.fr/tel-00008384.
Kambale, Bernard. "Modèle de m-learning et conception d'applications mobiles comme outils de support pour l'enseignement à distance en informatique et génie logiciel." Master's thesis, Université Laval, 2018. http://hdl.handle.net/20.500.11794/31325.
Many studies undertaken in the field of education have revealed that m-learning is emerging more and more as an effective learning method with the use of smartphones. Always turned on and easily transported, smartphones can be used anywhere, at any given time and in any context. Considering this potential of smartphones in our current society, in this thesis we present an m-learning model designed as a learning support tool in computer science and software engineering. To achieve this goal, we first show the existing efforts to integrate smartphones into programming tools. Following the same idea, we show examples of using mobile applications for programming. We then demonstrate that smartphones have limitations that make the programming exercise ineffective. Given these limitations, we show that, especially for programming courses, m-learning plays its role fully when it provides access to useful information on an ongoing basis to support the learning process. Thus, we come up with the design of m-learning as a learning support tool in computer science and software engineering. In this sense, m-learning provides an environment to receive updates on the planned work, notifications, comments, task schedules, new tasks to be performed, etc. Then, after having studied different architectural styles and different types of applications, we present possibilities of implementation of this m-learning model. We identify the hybrid architecture as the ideal architecture for designing and developing m-learning tools. We show how this hybrid architecture works by using the Apache Cordova Framework to produce m-learning tools that are both customizable and portable.
Lefebvre, Germain. "Sélection d'un modèle d'apprentissage pour rendre compte de la spéculation dans un paradigme de prospection monétaire." Thesis, Paris 2, 2018. http://www.theses.fr/2018PA020010.
This dissertation proposes to analyze empirically the microfoundations of the macroeconomic use of money, more particularly the human learning processes and cognitive abilities requiredfor a monetary equilibrium to emerge in an experimental economy implementing a search theoretical paradigm of money emergence. To achieve this, we operationalized the original Kiyotaki and Wright search model and fitted real subjects' behaviors with different reinforcement learning algorithms. We show that reinforcement learning better explains behavioral datain comparison to theoretical equilibria predictions, and highlight the importance of opportunity costs to implement a speculative use of money. Our results constitute a new step towards the understanding of learning processes at work in multi-step economic decision making and ofthe cognitive microfoundations of the macro-economic use of money. In parallel, this dissertation also compounds in-depth analyses of one of the core components of reinforcement learning,namely the update process. In two studies, we gradually show that the latter is biased positively towards confirmatory information. Indeed, we found that subjects performing different probability learning tasks preferentially took into account information that confirme dtheir initial thoughts in contrast to information that contradicted them. These results constitute a step towards the understanding of the genesis of optimism and confirmation biases at the neurocomputational level
Tencé, Fabien. "Probabilistic Behaviour Model and Imitation Learning Algorithm for Believable Characters in Video Games." Brest, 2011. http://www.theses.fr/2011BRES2032.
Ce manuscrit cherche à concevoir un modèle de comportement pour le contrôle de personnages crédibles dans les jeux vidéo. Nous définissons un personnage crédible comme un programme informatique capable de contrôler une représentation virtuelle de façon à ce que des observateurs dans l’environnement virtuel pensent que la représentation est contrôlée par un humain. Nous établissons 10 critères plus précis pour établir notre thèse. Pour répondre à ces 10 critères nous avons étudié des modèles développés à la fois dans le domaine académique et de l’industrie, L’évolution étant un des critères, nous avons aussi étudié les algorithmes d’apprentissages existants, notamment ceux basés sur l’imitation étant le mieux adaptés à la crédibilité. De ces études nous avons conclu que le modèle de Le Hy était une excellente base pour de futurs développements. Nous utilisons l’approche de Le Hy mais nous avons effectué des choix différents en vue d’une meilleur crédibilité. Nous proposons un raffinement sémantique et un mécanisme d’attention pour réduire le nombre de paramètres dans le modèle et améliorer le comportement. Un algorithme est ajouté pour permettre au personnage de s’orienter dans l’environnement et un l’algorithme d’apprentissage des paramètres du modèles a été repensé. Ces propositions permettent au modèle d’apprendre rapidement des associations stimuli-actions qui ressemblent à des comportement humains. Cependant de mauvaise associations sont aussi faites rendant le comportement non crédible. Selon nos mesures, notre modèle donne de meilleurs résultats en terme de crédibilité que le modèle de Le Hy, mais des améliorations restent encore à faire pour atteindre notre objectif
Caillon, Antoine. "Hierarchical temporal learning for multi-instrument and orchestral audio synthesis." Electronic Thesis or Diss., Sorbonne université, 2023. http://www.theses.fr/2023SORUS115.
Recent advances in deep learning have offered new ways to build models addressing a wide variety of tasks through the optimization of a set of parameters based on minimizing a cost function. Amongst these techniques, probabilistic generative models have yielded impressive advances in text, image and sound generation. However, musical audio signal generation remains a challenging problem. This comes from the complexity of audio signals themselves, since a single second of raw audio spans tens of thousands of individual samples. Modeling musical signals is even more challenging as important information are structured across different time scales, from micro (e.g. timbre, transient, phase) to macro (e.g. genre, tempo, structure) information. Modeling every scale at once would require large architectures, precluding the use of resulting models in real time setups for computational complexity reasons.In this thesis, we study how a hierarchical approach to audio modeling can address the musical signal modeling task, while offering different levels of control to the user. Our main hypothesis is that extracting different representation levels of an audio signal allows to abstract the complexity of lower levels for each modeling stage. This would eventually allow the use of lightweight architectures, each modeling a single audio scale. We start by addressing raw audio modeling by proposing an audio model combining Variational Auto Encoders and Generative Adversarial Networks, yielding high-quality 48kHz neural audio synthesis, while being 20 times faster than real time on CPU. Then, we study how autoregressive models can be used to understand the temporal behavior of the representation yielded by this low-level audio model, using optional additional conditioning signals such as acoustic descriptors or tempo. Finally, we propose a method for using all the proposed models directly on audio streams, allowing their use in realtime applications that we developed during this thesis. We conclude by presenting various creative collaborations led in parallel of this work with several composers and musicians, directly integrating the current state of the proposed technologies inside musical pieces
Fernandes, Hilaire. "iSTOA, modèle notionnel de guidage macroscopique de l'apprentissage." Phd thesis, Université des Sciences et Technologie de Lille - Lille I, 2010. http://tel.archives-ouvertes.fr/tel-00498599.
Nguyen, Huu Phuc. "Développement d'une commande à modèle partiel appris : analyse théorique et étude pratique." Thesis, Compiègne, 2016. http://www.theses.fr/2016COMP2323/document.
In classical control theory, the control law is generally built, based on the theoretical model of the system. That means that the mathematical equations representing the system dynamics are used to stabilize the closed loop. But in practice, the actual system differs from the theory, for example, the nonlinearity, the varied parameters and the unknown disturbances of the system. The proposed approach in this work is based on the knowledge of the plant system by using not only the analytical model but also the experimental data. The input values stabilizing the system on open loop, that minimize a cost function, for example, the distance between the desired output and the predicted output, or maximize a reward function are calculated by an optimal algorithm. The key idea of this approach is to use a numerical behavior model of the system as a prediction function on the joint state and input spaces or input-output spaces to find the controller’s output. To do this, a new non-linear control concept is proposed, based on an existing controller that uses a prediction map built on the state-space. The prediction model is initialized by using the best knowledge a priori of the system. It is then improved by using a learning algorithm based on the sensors’ data. Two types of prediction map are employed: the first one is based on the state-space model; the second one is represented by an input-output model. The output of the controller, that minimizes the error between the predicted output from the prediction model and the desired output, will be found using optimal algorithm. The application of the proposed controller has been made on various systems. Some real experiments for quadricopter, some actual tests for the electrical vehicle Zoé show its ability and efficiency to complex and fast systems. Other the results in simulation are tested in order to investigate and study the performance of the proposed controller. This approach is also used to estimate the rotor speed of the induction machine by considering the rotor speed as the input of the system
Malki, Noureddine. "Contribution au diagnostic des Systèmes à Evénements Discrets par modèles temporels et distributions de probabilité." Thesis, Reims, 2013. http://www.theses.fr/2013REIMS016/document.
The work presented in this thesis represents a contribution to the problem of diagnosis in discrete event systems (DES). The objective of our work consists in a proposition for a diagnostic approach by exploiting the temporal aspect which characterizing the occurrence of events. For this, the system is modeled by temporal graphs belonging to the timed automata formwork. The approach is designed according to the decentralized architecture to avoid any combinatorial explosion in the construction of the models. It has allowed the detection and isolation of abrupt faults occurring on equipment by combining the enablement conditions of events and the Boolean functions for the non-occurrence of events.Secondly, gradual faults coming from the process its self are considerate. For this, time constraints expressing the dates of occurrence of events in the Templates and Chronicles are modeled by probability distributions (PDs). These are used to characterize normal, degraded or failed functioning of each subsystem with a degree of certainty. Identification of this functioning mode is represented by the value of a degradation indicator
Ouarrak, Bouazza. "Les misconceptions dans la microgenèse de l’objet technique." Thesis, Paris, CNAM, 2011. http://www.theses.fr/2011CNAM0756/document.
This thesis investigates the cognitive resources that pupils engineers in a PBL (Problem based Learning) in a task of conception of a technical object mobilize. The situation-problem with which these pupils are confronted is constituted by an unpublished technical system of refrigeration without outside contribution of energy. In this learning, the pupils have to conceive the technical object and learn concepts in thermodynamics. Two groups of pupils are compared: the first one has an analogical model of a situation known to approach the new situation; the second has only the text. The questions of researches: what build these pupils as knowledge? What bring these two types of learning (the learning by a known situation and the learning by the text)? What are the obstacles which meet these pupils? The hypotheses: a learning by a known situation leads to the construction of operational knowledge (concepts tools). A learning by the text leads to the construction of knowledge out of context (concepts objects). A learning by the situations in a didactic device leads later to the construction of category-specific concepts. These two types of learning involve the epistemological obstacle in the construction of the concepts in them two functions: tool and object
Montandon, Véronique. "Un modèle neuronal pour la simulation opérationnelle des radiances observées par l'interféromètre spatial à haute résolution spectrale IASI." Paris 6, 2002. http://www.theses.fr/2002PA066442.
Dawoua, Kaoutoing Maxime. "Contributions à la modélisation et la simulation de la coupe des métaux : vers un outil d'aide à la surveillance par apprentissage." Thesis, Toulouse, INPT, 2020. http://www.theses.fr/2020INPT0013.
Shaping processes by material removal, also known as machining, are the manufacturing processes most commonly used for the production of mechanical parts, particularly in industrial sectors such as aeronautics, automotive, railways, etc. Although these processes are widely used in industry, the prediction of the characteristic sizes of the machining process is not always accurate, and a poor choice of cutting conditions can lead to abnormal tool wear or even to a deterioration in the quality of the machined part. The fine simulation of machining parameters, aiming at detecting anomalies, is a good example of this problem, as it represents the general problem of optimizing metal cutting to obtain cutting accuracy and anticipate rapid tool wear. This thesis is a contribution to the modelling and simulation of metal cutting, with a view to assisting mechanical parts manufacturing companies in their decision-making, based on knowledge extraction from simulated data. An efficient implementation of an analytical model of orthogonal cutting of metals, able to predict cutting parameters in a reduced time was proposed. The performance of this model was studied by comparing its predictions with the 1045 and carbon steel machining data that are available in the literature. By using the high speed resolution obtained from the proposed implementation, a large quantity of data simulating real cutting conditions was generated, and allowed the elaboration of a machining monitoring approach, based on a deep unsupervised learning method. The implementation with the simulated data highlighted the ability of the proposed detection approach to identify combinations of input parameter values (from the analytical cutting model) that could generate an abnormally high internal temperature; this was considered in the thesis as an indicator of the health of the machining system. Implementation of the proposed learning model gave an accuracy of 99,96 % and a precision of 96%, reflecting its ability to effectively predict the outcome
Bauvin, Pierre. "Modélisation de la stéatose hépatique (NAFLD) et de ses facteurs de risque par apprentissage sur des données de santé." Thesis, Lille 2, 2020. http://www.theses.fr/2020LIL2S028.
Non-alcoholic fatty liver disease (NAFLD) is a chronic liver disease which is a combination of simple, slowly progressing steatosis, and non-alcoholic steatohepatitis (NASH), an inflammatory form which accelerates its progression. It is estimated that one in four people in the world is affected by NAFLD, and its prevalence is increasing rapidly, in parallel with the prevalence of its main risk factors: overweight, obesity and type 2 diabetes.This pathology is asymptomatic up to the complications, cirrhosis and liver cancer (hepatocellular carcinoma, HCC), which leads to late diagnosis and a negative impact on the associated morbidity and mortality. Furthermore, the reference diagnosis requires a liver biopsy, an invasive examination that cannot be performed routinely. As a result, the progression of the disease is poorly known and its estimation may suffer from a selection bias, towards patients with significant risk factors, who require a biopsy in the first place. A better understanding would allow the implementation of strategies to reduce its burden.The modelling approach is appropriate to take into account all susceptible patients, without having to carry out a large-scale follow-up study using liver biopsies in patients who are mostly asymptomatic. The objectives of this thesis are to describe and quantify the progression of NAFLD, to predict the associated morbidity and mortality, and to identify the population at risk, using Markov models. To do this, it is necessary to fill in some of the progression parameters via a literature review, to characterise the initial states (population likely to develop NAFLD) and the final states (mortality due to NAFLD), in order to deduce the missing progression parameters between the onset of the disease and mortality, by back-calculation.To exhaustively characterise NAFLD mortality, we identified all patients with cirrhosis or HCC from national hospital databases, representing more than 380,000 patients. We then developed an identification algorithm to determine the etiology underlying the hepatic complication, based on all the stays of the identified patients. This algorithm requires the identification of patients with cirrhosis or HCC of alcoholic or viral origin, to obtain by elimination only NAFLD patients. Once the specific mortality data had been obtained, we estimated the population likely to develop NAFLD, defined as all individuals with overweight or type 2 diabetes, excluding the population of excessive drinkers. We estimated the prevalence and incidence of this population, and modelled its evolution with age and years, based on individual data from surveys representative of the French population.Finally, we quantified the progression of NAFLD, and the impact of risk factors, using two approaches: from the literature, and from biopsy data from more than 1,800 obese patients who were candidates for bariatric surgery, resulting in a tool for predicting the progression of NAFLD in this population. We chose to back-calculate the progression parameters corresponding to the asymptomatic states, which are the most susceptible to selection bias.We obtained a model of the progression of NAFLD, taking into account the dynamic distribution of the population among weight classes and diabetes status, and resulting in the observed statistics of NAFLD deaths. The model takes into account gender, age, year, BMI (body mass index) class, diabetes status and the presence of a genetic polymorphism (PNPLA3 rs738409, C→G) as covariates of progression. It is a tool for assessing the impact of a possible treatment or public health policy on morbidity and mortality
Fragne, Didier. "Proposition de l'architecture de l'agent gestionnaire du modèle de l'apprenant dans un système tuteur multi-agents en apprentissage de la lecture : contribution au projet AMICAL." Phd thesis, Université Blaise Pascal - Clermont-Ferrand II, 2009. http://tel.archives-ouvertes.fr/tel-00449160.