Dissertations / Theses on the topic 'Machine Learning Bayésien'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 47 dissertations / theses for your research on the topic 'Machine Learning Bayésien.'
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.
Zecchin, Matteo. "Robust Machine Learning Approaches to Wireless Communication Networks." Electronic Thesis or Diss., Sorbonne université, 2022. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2022SORUS397.pdf.
Full textArtificial intelligence is widely viewed as a key enabler of sixth generation wireless systems. In this thesis we target fundamental problems arising from the interaction between these two technologies with the end goal of paving the way towards the adoption of reliable AI in future wireless networks. We develop of distributed training algorithms that allow collaborative learning at edge of wireless networks despite communication bottlenecks, unreliability of its workers and data heterogeneity. We then take a critical look at the application of the standard frequentist learning paradigm to wireless communication problems and propose an extension of the generalized Bayesian learning, that concurrently counteracts three prominent challenges arising in application domain: data scarcity, the presence of outliers and model misspecification
Huix, Tom. "Variational Inference : theory and large scale applications." Electronic Thesis or Diss., Institut polytechnique de Paris, 2024. http://www.theses.fr/2024IPPAX071.
Full textThis thesis explores Variational Inference methods for high-dimensional Bayesian learning. In Machine Learning, the Bayesian approach allows one to deal with epistemic uncertainty and provides and a better uncertainty quantification, which is necessary in many machine learning applications. However, Bayesian inference is often not feasible because the posterior distribution of the model parameters is generally untractable. Variational Inference (VI) allows to overcome this problem by approximating the posterior distribution with a simpler distribution called the variational distribution.In the first part of this thesis, we worked on the theoretical guarantees of Variational Inference. First, we studied VI when the Variational distribution is a Gaussian and in the overparameterized regime, i.e., when the models are high dimensional. Finally, we explore the Gaussian mixtures Variational distributions, as it is a more expressive distribution. We studied both the optimization error and the approximation error of this method.In the second part of the thesis, we studied the theoretical guarantees for contextual bandit problems using a Bayesian approach called Thompson Sampling. First, we explored the use of Variational Inference for Thompson Sampling algorithm. We notably showed that in the linear framework, this approach allows us to obtain the same theoretical guarantees as if we had access to the true posterior distribution. Finally, we consider a variant of Thompson Sampling called Feel-Good Thompson Sampling (FG-TS). This method allows to provide better theoretical guarantees than the classical algorithm. We then studied the use of a Monte Carlo Markov Chain method to approximate the posterior distribution. Specifically, we incorporated into FG-TS a Langevin Monte Carlo algorithm and a Metropolized Langevin Monte Carlo algorithm. Moreover, we obtained the same theoretical guarantees as for FG-TS when the posterior distribution is known
Jarraya, Siala Aida. "Nouvelles paramétrisations de réseaux bayésiens et leur estimation implicite : famille exponentielle naturelle et mélange infini de Gaussiennes." Phd thesis, Nantes, 2013. https://archive.bu.univ-nantes.fr/pollux/show/show?id=aef89743-c009-457d-8c27-a888655a4e58.
Full textLearning a Bayesian network consists in estimating the graph (structure) and the parameters of conditional probability distributions associated with this graph. Bayesian networks learning algorithms rely on classical Bayesian estimation approach whose a priori parameters are often determined by an expert or defined uniformly The core of this work concerns the application of several advances in the field of statistics as implicit estimation, Natural exponential families or infinite mixtures of Gaussian in order to (1) provide new parametric forms for Bayesian networks, (2) estimate the parameters of such models and (3) learn their structure
Jarraya, Siala Aida. "Nouvelles paramétrisations de réseaux Bayésiens et leur estimation implicite - Famille exponentielle naturelle et mélange infini de Gaussiennes." Phd thesis, Université de Nantes, 2013. http://tel.archives-ouvertes.fr/tel-00932447.
Full textSynnaeve, Gabriel. "Programmation et apprentissage bayésien pour les jeux vidéo multi-joueurs, application à l'intelligence artificielle de jeux de stratégies temps-réel." Phd thesis, Université de Grenoble, 2012. http://tel.archives-ouvertes.fr/tel-00780635.
Full textGrappin, Edwin. "Model Averaging in Large Scale Learning." Thesis, Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLG001/document.
Full textThis thesis explores properties of estimations procedures related to aggregation in the problem of high-dimensional regression in a sparse setting. The exponentially weighted aggregate (EWA) is well studied in the literature. It benefits from strong results in fixed and random designs with a PAC-Bayesian approach. However, little is known about the properties of the EWA with Laplace prior. Chapter 2 analyses the statistical behaviour of the prediction loss of the EWA with Laplace prior in the fixed design setting. Sharp oracle inequalities which generalize the properties of the Lasso to a larger family of estimators are established. These results also bridge the gap from the Lasso to the Bayesian Lasso. Chapter 3 introduces an adjusted Langevin Monte Carlo sampling method that approximates the EWA with Laplace prior in an explicit finite number of iterations for any targeted accuracy. Chapter 4 explores the statisctical behaviour of adjusted versions of the Lasso for the transductive and semi-supervised learning task in the random design setting
Grappin, Edwin. "Model Averaging in Large Scale Learning." Electronic Thesis or Diss., Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLG001.
Full textThis thesis explores properties of estimations procedures related to aggregation in the problem of high-dimensional regression in a sparse setting. The exponentially weighted aggregate (EWA) is well studied in the literature. It benefits from strong results in fixed and random designs with a PAC-Bayesian approach. However, little is known about the properties of the EWA with Laplace prior. Chapter 2 analyses the statistical behaviour of the prediction loss of the EWA with Laplace prior in the fixed design setting. Sharp oracle inequalities which generalize the properties of the Lasso to a larger family of estimators are established. These results also bridge the gap from the Lasso to the Bayesian Lasso. Chapter 3 introduces an adjusted Langevin Monte Carlo sampling method that approximates the EWA with Laplace prior in an explicit finite number of iterations for any targeted accuracy. Chapter 4 explores the statisctical behaviour of adjusted versions of the Lasso for the transductive and semi-supervised learning task in the random design setting
Araya-López, Mauricio. "Des algorithmes presque optimaux pour les problèmes de décision séquentielle à des fins de collecte d'information." Electronic Thesis or Diss., Université de Lorraine, 2013. http://www.theses.fr/2013LORR0002.
Full textThe purpose of this dissertation is to study sequential decision problems where acquiring information is an end in itself. More precisely, it first covers the question of how to modify the POMDP formalism to model information-gathering problems and which algorithms to use for solving them. This idea is then extended to reinforcement learning problems where the objective is to actively learn the model of the system. Also, this dissertation proposes a novel Bayesian reinforcement learning algorithm that uses optimistic local transitions to efficiently gather information while optimizing the expected return. Through bibliographic discussions, theoretical results and empirical studies, it is shown that these information-gathering problems are optimally solvable in theory, that the proposed methods are near-optimal solutions, and that these methods offer comparable or better results than reference approaches. Beyond these specific results, this dissertation paves the way (1) for understanding the relationship between information-gathering and optimal policies in sequential decision processes, and (2) for extending the large body of work about system state control to information-gathering problems
Araya-López, Mauricio. "Des algorithmes presque optimaux pour les problèmes de décision séquentielle à des fins de collecte d'information." Phd thesis, Université de Lorraine, 2013. http://tel.archives-ouvertes.fr/tel-00943513.
Full textRahier, Thibaud. "Réseaux Bayésiens pour fusion de données statiques et temporelles." Thesis, Université Grenoble Alpes (ComUE), 2018. http://www.theses.fr/2018GREAM083/document.
Full textPrediction and inference on temporal data is very frequently performed using timeseries data alone. We believe that these tasks could benefit from leveraging the contextual metadata associated to timeseries - such as location, type, etc. Conversely, tasks involving prediction and inference on metadata could benefit from information held within timeseries. However, there exists no standard way of jointly modeling both timeseries data and descriptive metadata. Moreover, metadata frequently contains highly correlated or redundant information, and may contain errors and missing values.We first consider the problem of learning the inherent probabilistic graphical structure of metadata as a Bayesian Network. This has two main benefits: (i) once structured as a graphical model, metadata is easier to use in order to improve tasks on temporal data and (ii) the learned model enables inference tasks on metadata alone, such as missing data imputation. However, Bayesian network structure learning is a tremendous mathematical challenge, that involves a NP-Hard optimization problem. We present a tailor-made structure learning algorithm, inspired from novel theoretical results, that exploits (quasi)-determinist dependencies that are typically present in descriptive metadata. This algorithm is tested on numerous benchmark datasets and some industrial metadatasets containing deterministic relationships. In both cases it proved to be significantly faster than state of the art, and even found more performant structures on industrial data. Moreover, learned Bayesian networks are consistently sparser and therefore more readable.We then focus on designing a model that includes both static (meta)data and dynamic data. Taking inspiration from state of the art probabilistic graphical models for temporal data (Dynamic Bayesian Networks) and from our previously described approach for metadata modeling, we present a general methodology to jointly model metadata and temporal data as a hybrid static-dynamic Bayesian network. We propose two main algorithms associated to this representation: (i) a learning algorithm, which while being optimized for industrial data, is still generalizable to any task of static and dynamic data fusion, and (ii) an inference algorithm, enabling both usual tasks on temporal or static data alone, and tasks using the two types of data.%We then provide results on diverse cross-field applications such as forecasting, metadata replenishment from timeseries and alarms dependency analysis using data from some of Schneider Electric’s challenging use-cases.Finally, we discuss some of the notions introduced during the thesis, including ways to measure the generalization performance of a Bayesian network by a score inspired from the cross-validation procedure from supervised machine learning. We also propose various extensions to the algorithms and theoretical results presented in the previous chapters, and formulate some research perspectives
Löser, Kevin. "Apprentissage non-supervisé de la morphologie des langues à l’aide de modèles bayésiens non-paramétriques." Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLS203/document.
Full textA crucial issue in statistical natural language processing is the issue of sparsity, namely the fact that in a given learning corpus, most linguistic events have low occurrence frequencies, and that an infinite number of structures allowed by a language will not be observed in the corpus. Neural models have already contributed to solving this issue by inferring continuous word representations. These continuous representations allow to structure the lexicon by inducing semantic or syntactic similarity between words. However, current neural models only partially solve the sparsity issue, due to the fact that they require a vectorial representation for every word in the lexicon, but are unable to infer sensible representations for unseen words. This issue is especially present in morphologically rich languages, where word formation processes yield a proliferation of possible word forms, and little overlap between the lexicon observed during model training, and the lexicon encountered during its use. Today, several languages are used on the Web besides English, and engineering translation systems that can handle morphologies that are very different from western European languages has become a major stake. The goal of this thesis is to develop new statistical models that are able to infer in an unsupervised fashion the word formation processes underlying an observed lexicon, in order to produce morphological analyses of new unseen word forms
Boutkhamouine, Brahim. "Stochastic modelling of flood phenomena based on the combination of mechanist and systemic approaches." Thesis, Toulouse, INPT, 2018. http://www.theses.fr/2018INPT0142/document.
Full textFlood forecasting describes the rainfall-runoff transformation using simplified representations. These representations are based on either empirical descriptions, or on equations of classical mechanics of the involved physical processes. The performances of the existing flood predictions are affected by several sources of uncertainties coming not only from the approximations involved but also from imperfect knowledge of input data, initial conditions of the river basin, and model parameters. Quantifying these uncertainties enables the decision maker to better interpret the predictions and constitute a valuable decision-making tool for flood risk management. Uncertainty analysis on existing rainfall-runoff models are often performed using Monte Carlo (MC)- simulations. The implementation of this type of techniques requires a large number of simulations and consequently a potentially important calculation time. Therefore, quantifying uncertainties of real-time hydrological models is challenging. In this project, we develop a methodology for flood prediction based on Bayesian networks (BNs). BNs are directed acyclic graphs where the nodes correspond to the variables characterizing the modelled system and the arcs represent the probabilistic dependencies between these variables. The presented methodology suggests to build the RBs from the main hydrological factors controlling the flood generation, using both the available observations of the system response and the deterministic equations describing the processes involved. It is, thus, designed to take into account the time variability of different involved variables. The conditional probability tables (parameters), can be specified using observed data, existing hydrological models or expert opinion. Thanks to their inference algorithms, BN are able to rapidly propagate, through the graph, different sources of uncertainty in order to estimate their effect on the model output (e.g. riverflow). Several case studies are tested. The first case study is the Salat river basin, located in the south-west of France, where a BN is used to simulate the discharge at a given station from the streamflow observations at 3 hydrometric stations located upstream. The model showed good performances estimating the discharge at the outlet. Used in a reverse way, the model showed also satisfactory results when characterising the discharges at an upstream station by propagating back discharge observations of some downstream stations. The second case study is the Sagelva basin, located in Norway, where a BN is used to simulate the accumulation of snow water equivalent (SWE) given available weather data observations. The performances of the model are affected by the learning dataset used to train the BN parameters. In the absence of relevant observation data for learning, a methodology for learning the BN-parameters from deterministic models is proposed and tested. The resulted BN can be used to perform uncertainty analysis without any MC-simulations to be performed in real-time. From these case studies, it appears that BNs are a relevant decisionsupport tool for flood risk management
Trabelsi, Ghada. "New structure learning algorithms and evaluation methods for large dynamic Bayesian networks." Phd thesis, Université de Nantes, 2013. http://tel.archives-ouvertes.fr/tel-00996061.
Full textCherief-Abdellatif, Badr-Eddine. "Contributions to the theoretical study of variational inference and robustness." Electronic Thesis or Diss., Institut polytechnique de Paris, 2020. http://www.theses.fr/2020IPPAG001.
Full textThis PhD thesis deals with variational inference and robustness. More precisely, it focuses on the statistical properties of variational approximations and the design of efficient algorithms for computing them in an online fashion, and investigates Maximum Mean Discrepancy based estimators as learning rules that are robust to model misspecification.In recent years, variational inference has been extensively studied from the computational viewpoint, but only little attention has been put in the literature towards theoretical properties of variational approximations until very recently. In this thesis, we investigate the consistency of variational approximations in various statistical models and the conditions that ensure the consistency of variational approximations. In particular, we tackle the special case of mixture models and deep neural networks. We also justify in theory the use of the ELBO maximization strategy, a model selection criterion that is widely used in the Variational Bayes community and is known to work well in practice.Moreover, Bayesian inference provides an attractive online-learning framework to analyze sequential data, and offers generalization guarantees which hold even under model mismatch and with adversaries. Unfortunately, exact Bayesian inference is rarely feasible in practice and approximation methods are usually employed, but do such methods preserve the generalization properties of Bayesian inference? In this thesis, we show that this is indeed the case for some variational inference algorithms. We propose new online, tempered variational algorithms and derive their generalization bounds. Our theoretical result relies on the convexity of the variational objective, but we argue that our result should hold more generally and present empirical evidence in support of this. Our work presents theoretical justifications in favor of online algorithms that rely on approximate Bayesian methods. Another point that is addressed in this thesis is the design of a universal estimation procedure. This question is of major interest, in particular because it leads to robust estimators, a very hot topic in statistics and machine learning. We tackle the problem of universal estimation using a minimum distance estimator based on the Maximum Mean Discrepancy. We show that the estimator is robust to both dependence and to the presence of outliers in the dataset. We also highlight the connections that may exist with minimum distance estimators using L2-distance. Finally, we provide a theoretical study of the stochastic gradient descent algorithm used to compute the estimator, and we support our findings with numerical simulations. We also propose a Bayesian version of our estimator, that we study from both a theoretical and a computational points of view
Drosouli, Ifigeneia. "Multimodal machine learning methods for pattern analysis in smart cities and transportation." Electronic Thesis or Diss., Limoges, 2024. http://www.theses.fr/2024LIMO0028.
Full textIn the context of modern, densely populated urban environments, the effective management of transportation and the structure of Intelligent Transportation Systems (ITSs) are paramount. The public transportation sector is currently undergoing a significant expansion and transformation with the objective of enhancing accessibility, accommodating larger passenger volumes without compromising travel quality, and embracing environmentally conscious and sustainable practices. Technological advancements, particularly in Artificial Intelligence (AI), Big Data Analytics (BDA), and Advanced Sensors (AS), have played a pivotal role in achieving these goals and contributing to the development, enhancement, and expansion of Intelligent Transportation Systems. This thesis addresses two critical challenges within the realm of smart cities, specifically focusing on the identification of transportation modes utilized by citizens at any given moment and the estimation and prediction of transportation flow within diverse transportation systems. In the context of the first challenge, two distinct approaches have been developed for Transportation Mode Detection. Firstly, a deep learning approach for the identification of eight transportation media is proposed, utilizing multimodal sensor data collected from user smartphones. This approach is based on a Long Short-Term Memory (LSTM) network and Bayesian optimization of model’s parameters. Through extensive experimental evaluation, the proposed approach demonstrates remarkably high recognition rates compared to a variety of machine learning approaches, including state-of-the-art methods. The thesis also delves into issues related to feature correlation and the impact of dimensionality reduction. The second approach involves a transformer-based model for transportation mode detection named TMD-BERT. This model processes the entire sequence of data, comprehends the importance of each part of the input sequence, and assigns weights accordingly using attention mechanisms to grasp global dependencies in the sequence. Experimental evaluations showcase the model's exceptional performance compared to state-of-the-art methods, highlighting its high prediction accuracy. In addressing the challenge of transportation flow estimation, a Spatial-Temporal Graph Convolutional Recurrent Network is proposed. This network learns from both the spatial stations network data and time-series of historical mobility changes to predict urban metro and bike sharing flow at a future time. The model combines Graph Convolutional Networks (GCN) and Long Short-Term Memory (LSTM) Networks to enhance estimation accuracy. Extensive experiments conducted on real-world datasets from the Hangzhou metro system and the NY City bike sharing system validate the effectiveness of the proposed model, showcasing its ability to identify dynamic spatial correlations between stations and make accurate long-term forecasts
Argouarc, h. Elouan. "Contributions to posterior learning for likelihood-free Bayesian inference." Electronic Thesis or Diss., Institut polytechnique de Paris, 2024. http://www.theses.fr/2024IPPAS021.
Full textBayesian posterior inference is used in many scientific applications and is a prevalent methodology for decision-making under uncertainty. It enables practitioners to confront real-world observations with relevant observation models, and in turn, infer the distribution over an explanatory variable. In many fields and practical applications, we consider ever more intricate observation models for their otherwise scientific relevance, but at the cost of intractable probability density functions. As a result, both the likelihood and the posterior are unavailable, making posterior inference using the usual Monte Carlo methods unfeasible.In this thesis, we suppose that the observation model provides a recorded dataset, and our aim is to bring together Bayesian inference and statistical learning methods to perform posterior inference in a likelihood-free setting. This problem, formulated as learning an approximation of a posterior distribution, includes the usual statistical learning tasks of regression and classification modeling, but it can also be an alternative to Approximate Bayesian Computation methods in the context of simulation-based inference, where the observation model is instead a simulation model with implicit density.The aim of this thesis is to propose methodological contributions for Bayesian posterior learning. More precisely, our main goal is to compare different learning methods under the scope of Monte Carlo sampling and uncertainty quantification.We first consider the posterior approximation based on the likelihood-to-evidence ratio, which has the main advantage that it turns a problem of conditional density learning into a problem of binary classification. In the context of Monte Carlo sampling, we propose a methodology for sampling from such a posterior approximation. We leverage the structure of the underlying model, which is conveniently compatible with the usual ratio-based sampling algorithms, to obtain straightforward, parameter-free, and density-free sampling procedures.We then turn to the problem of uncertainty quantification. On the one hand, normalized models such as the discriminative construction are easy to apply in the context of Bayesian uncertainty quantification. On the other hand, while unnormalized models, such as the likelihood-to-evidence-ratio, are not easily applied in uncertainty-aware learning tasks, a specific unnormalized construction, which we refer to as generative, is indeed compatible with Bayesian uncertainty quantification via the posterior predictive distribution. In this context, we explain how to carry out uncertainty quantification in both modeling techniques, and we then propose a comparison of the two constructions under the scope of Bayesian learning.We finally turn to the problem of parametric modeling with tractable density, which is indeed a requirement for epistemic uncertainty quantification in generative and discriminative modeling methods. We propose a new construction of a parametric model, which is an extension of both mixture models and normalizing flows. This model can be applied to many different types of statistical problems, such as variational inference, density estimation, and conditional density estimation, as it benefits from rapid and exact density evaluation, a straightforward sampling scheme, and a gradient reparameterization approach
Lesieur, Thibault. "Factorisation matricielle et tensorielle par une approche issue de la physique statistique." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLS345/document.
Full textIn this thesis we present the result on low rank matrix and tensor factorization. Matrices being such an ubiquitous mathematical object a lot of machine learning can be mapped to a low-rank matrix factorization problem. It is for example one of the basic methods used in data analysis for unsupervised learning of relevant features and other types of dimensionality reduction. The result presented in this thesis have been included in previous work [LKZ 201].The problem of low rank matrix becomes harder once one adds constraint to the problem like for instance the positivity of one of the factor of the factorization. We present a framework to study the constrained low-rank matrix estimation for a general prior on the factors, and a general output channel through which the matrix is observed. We draw a paralel with the study of vector-spin glass models -- presenting a unifying way to study a number of problems considered previously in separate statistical physics works. We present a number of applications for the problem in data analysis. We derive in detail ageneral form of the low-rank approximate message passing (Low-RAMP) algorithm that is known in statistical physics as the TAP equations. We thus unify the derivation of the TAP equations for models as different as the Sherrington-Kirkpatrick model, the restricted Boltzmann machine, the Hopfield model or vector (xy, Heisenberg and other) spin glasses. The state evolution of the Low-RAMP algorithm is also derived, and is equivalent to the replica symmetric solution for the large class of vector-spin glass models. In the section devoted to result we study in detail phase diagrams and phase transitions for the Bayes-optimal inference in low-rank matrix estimation. We present a typology of phase transitions and their relation to performance of algorithms such as the Low-RAMP or commonly used spectral methods
Rio, Maxime. "Modèles bayésiens pour la détection de synchronisations au sein de signaux électro-corticaux." Electronic Thesis or Diss., Université de Lorraine, 2013. http://www.theses.fr/2013LORR0090.
Full textThis thesis promotes new methods to analyze intracranial cerebral signals (local field potentials), which overcome limitations of the standard time-frequency method of event-related spectral perturbations analysis: averaging over the trials and relying on the activity in the pre-stimulus period. The first proposed method is based on the detection of sub-networks of electrodes whose activity presents cooccurring synchronisations at a same point of the time-frequency plan, using bayesian gaussian mixture models. The relevant sub-networks are validated with a stability measure computed over the results obtained from different trials. For the second proposed method, the fact that a white noise in the temporal domain is transformed into a rician noise in the amplitude domain of a time-frequency transform made possible the development of a segmentation of the signal in each frequency band of each trial into two possible levels, a high one and a low one, using bayesian rician mixture models with two components. From these two levels, a statistical analysis can detect time-frequency regions more or less active. To develop the bayesian rician mixture model, new algorithms of variational bayesian inference have been created for the Rice distribution and the rician mixture distribution. Performances of the new methods have been evaluated on artificial data and experimental data recorded on monkeys. It appears that the new methods generate less false positive results and are more robust to a lack of data in the pre-stimulus period
Burban, Ewen. "Approche génomique de la détection des barrières au flux de gènes." Electronic Thesis or Diss., Université de Rennes (2023-....), 2024. http://www.theses.fr/2024URENB007.
Full textCharacterizing the mechanisms that underlie reproductive isolation between diverging lineages is central in understanding the speciation process. As populations evolve, they gradually develop reproductive isolation (RI) by passing through intermediate steps, often referred to as the "gray zone of speciation". This isolation is marked by the emergence of genomic regions acting as barriers to local gene flow, distinct from the rest of the genome. Detecting these barrier loci involves identifying outlier loci with specific signatures. However, other processes can create similar patterns, which challenges barrier loci detection. In my thesis, I developed a new tool, RIDGE - Reproductive Isolation Detection using Genomic Polymorphisms, a novel free and portable tool tailored for this purpose in a comparative framework. RIDGE utilizes an Approximate Bayesian Computation model-averaging approach based on a random forest to accommodate diverse scenarios of lineage divergence. It considers heterogeneity in migration rate, linked selection, and recombination, estimates barrier proportion and conducts locus-scale tests for gene flow barriers. Simulations and analyses of published datasets in crow species pairs demonstrate RIDGE's efficacy in detecting ongoing migration and identifying barrier loci, even for recent divergence times. Furthermore, the contribution of summary statistics varies depending on the dataset, highlighting the complexity of gene flow barrier genomic signals and the interest of combining several statistics. Subsequently, I applied RIDGE to wild/domestic pairs in maize (an outcrosser), and foxtail millet (a selfer), both domesticated around 9,000 years ago. Gene flow between forms has been reported in these two systems. Consistently, models with ongoing migration and heterogeneity in migration rate were clearly dominant over other models. RIDGE also demonstrated its ability to distinguish between barrier loci and domestication loci (that experienced selective sweeps within the domestic forms). The perspectives of this work include applying RIDGE to multiple population/species pairs encompassing a large spectrum of divergence to determine the genomic pattern of RI during speciation, to test the snowball theory formulated by Orr in 1995 or to determine the nature of speciation genes
Cottet, Vincent R. "Theoretical study of some statistical procedures applied to complex data." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLG002/document.
Full textThe main part of this thesis aims at studying the theoretical and algorithmic aspects of three distinct statistical procedures. The first problem is the binary matrix completion. We propose an estimator based on a variational approximation of a pseudo-Bayesian estimator. We use a different loss function of the ones used in the literature. We are able to compute non asymptotic risk bounds. It is much faster to compute the estimator than a MCMC method and we show on examples that it is efficient in practice. In a second part we study the theoretical properties of the regularized empirical risk minimizer for Lipschitz loss functions. We are therefore able to apply it on the logistic regression with the SLOPE regularization and on the matrix completion as well. The third chapter develops an Expectation-Propagation approximation when the likelihood is not explicit. We then use an ABC approximation in a second stage. This procedure may be applied to many models and is more precise and faster than the classic ABC approximation. It is used in a spatial extremes model
Rio, Maxime. "Modèles bayésiens pour la détection de synchronisations au sein de signaux électro-corticaux." Phd thesis, Université de Lorraine, 2013. http://tel.archives-ouvertes.fr/tel-00859307.
Full textZambrano, Ramirez Oscar Daniel. "Bayesian statistics and modeling for the prediction of radiotherapy outcomes : an application to glioblastoma treatment." Thesis, Normandie, 2018. http://www.theses.fr/2018NORMC277/document.
Full textA Bayesian statistics framework was created in this thesis work for developing clinical based models in a continuous learning approach in which new data can be added. The objective of the models is to forecast radiation therapy effects based on clinical evidence. Machine learning concepts were used for solving the Bayesian framework. The models developed concern an aggressive brain cancer called glioblastoma. The medical data comprises a database of about 90 patients suffering glioblastoma; the database contains medical images and data entries such as age, gender, etc. Neurologic grade predictions models were constructed for illustrating the type of models that can be build with the methodology. Glioblastoma recurrence models, in the form of Generalized Linear Models (GLM) and decision tree models, were developed to explore the possibility of predicting the recurrence location using pre-radiation treatment imaging. Following, due to the lack of a sufficiently strong prediction obtained by the tree models, we decided to develop visual representation tools to directly observe the medical image intensity values concerning the recurrence and non-recurrence locations. Overall, the framework developed for modeling of radiation therapy clinical data provides a solid foundation for more complex models to be developed
Jallais, Maëliss. "Enabling cortical cell-specific sensitivity on diffusion MRI microstructure measurements using likelihood-free inference." Electronic Thesis or Diss., université Paris-Saclay, 2022. http://www.theses.fr/2022UPASG012.
Full textNeurodegenerative diseases, such as Alzheimer's or Huntington's disease, lead to the progressive and irreversible loss of mental functions. Dementia and cognitive deficits appear to be primarily related to neuronal and synaptic connectivity loss. Although these diseases' external impact and progression are readily observable, accessing microstructural changes in the brain remains a challenge, making it difficult to understand these diseases and develop treatments.With technological advances, diffusion Magnetic Resonance Imaging (dMRI) has emerged as a novel method to study the microstructure of the brain non-invasively and in-vivo. This medical imaging technique is based on the study of random microscopic movements of water molecules, known as Brownian movements. In the brain, the movements of the molecules are constrained by cell membranes, making diffusion anisotropic. Each tissue component, such as somas (cell bodies) or neurites, has a distinct shape. The characteristics of the tissue thus modulate the diffusion brain signal obtained during an MRI acquisition.My thesis aims to develop a method to infer a tissue microstructure from a dMRI acquisition in the grey matter (GM).The solution to this inverse problem of estimating brain microstructure from dMRI is threefold:1. The definition of a biological model describing the GM tissues. Existing microstructural models of white matter were proven not to hold in the GM. We adapted these models to take into account the abundance of somas in the GM.2. A mathematical modeling of the GM tissue. We modeled each compartment of the tissue model by simple geometrical shapes, for which the diffusion signal is known. We developed a signal processing algorithm to synthesize the key information contained in the diffusion signal and relate it to a set of parameters describing the tissue (notably the size and density of neurons). This algorithm is based on a study of the statistical moments of the signal at different MRI gradient strengths. Unlike existing methods, no biological parameters are arbitrarily fixed, which allows for the best possible description of the cortical tissue of each subject.3. An inversion algorithm to estimate the tissue parameters that generated the acquisition signal. Once the mathematical model relating tissue parameters to the diffusion signal is defined, the objective is to solve the inverse problem of estimating tissue microstructure from an observation. A limitation of current methods is their inability to identify all possible tissue configurations that can explain the same observed diffusion signal, making the interpretation of the proposed estimates difficult. We used a Bayesian deep-learning method called "likelihood-based inference" combined with neural networks to solve this problem. This method allows identifying and returning all possible tissue configurations along with their posterior distributions (probability given an observation), facilitating their interpretation.We first validated this approach on simulations. Based on a few acquisition constraints, we then applied the global resolution method to the HCP MGH and HCP1200 databases of the Human Connectome Project. We developed a python library to study those simulated or acquired data. The obtained results were then compared with histological and cognitive studies to verify their validity
Sarao, Mannelli Stefano. "On the dynamics of descent algorithms in high-dimensional planted models." Thesis, université Paris-Saclay, 2020. http://www.theses.fr/2020UPASP014.
Full textOptimization of high-dimensional non-convex models has always been a difficult and fascinating problem. Since our minds tend to apply notions that we experienced and naturally learned in low-dimension, our intuition is often led astray.Those problems appear naturally and become more and more relevant, in particular in an era where an increasingly large amount of data is available. Most of the information that we receive is useless and identifying what is relevant is an intricate problem.Machine learning problems and inference problems often fall in these settings.In both cases we have a cost function that depends on a large number of parameters that should be optimized. A rather simple, but common, choice is the use of local algorithms based on the gradient, that descend in the cost function trying to identify good solutions.If the cost function is convex, then under mild conditions on the descent rate, we are guaranteed to find the good solution. However often we do not have convex costs. To understand what happens in the dynamics of these non-convex high-dimensional problems is the main goal of this project.In the thesis we will space from Bayesian inference to machine learning in the attempt of building a theory that describes how the algorithmic dynamics evolve and when it is doomed to fail. Prototypical models of machine learning and inference problems are intimately related. Another interesting connection that is known since long time, is the link between inference problems and disordered systems studied by statistical physicists. The techniques and the results developed in the latter form the true skeleton of this work.In this dissertation we characterize the algorithmic limits of gradient descent and Langevin dynamics. We analyze the structure of the landscape and find the counter-intuitive result that in general an exponential number of spurious solutions do not prevent vanilla gradient descent initialized randomly to find the only good solution. Finally, we build a theory that explains quantitatively and qualitatively the underlying phenomenon
Guedj, Benjamin. "Agrégation d'estimateurs et de classificateurs : théorie et méthodes." Phd thesis, Université Pierre et Marie Curie - Paris VI, 2013. http://tel.archives-ouvertes.fr/tel-00922353.
Full textEgele, Romain. "Optimization of Learning Workflows at Large Scale on High-Performance Computing Systems." Electronic Thesis or Diss., université Paris-Saclay, 2024. http://www.theses.fr/2024UPASG025.
Full textIn the past decade, machine learning has experienced exponential growth, propelled by abundant datasets, algorithmic advancements, and increased computational power. Simultaneously, high-performance computing (HPC) has evolved to meet rising computational demands, offering resources to tackle complex scientific challenges.However, machine learning is often a sequential process, making it difficult to scale on HPC systems. Machine learning workflows are built from modules offering numerous configurable parameters, from data augmentation policies to training procedures and model architectures. This thesis focuses on the hyperparameter optimization of learning workflows on large-scale HPC systems, such as the Polaris at the Argonne Leadership Computing Facility.Key contributions include (1) asynchronous decentralized parallel Bayesian optimization, (2) extension to multi-objective, (3) integration of early discarding, and (4) uncertainty quantification of deep neural networks. Furthermore, an open-source software, DeepHyper, is provided, encapsulating the proposed algorithms to facilitate research and application. The thesis highlights the importance of scalable Bayesian optimization methods for the hyperparameter optimization of learning workflows, which is crucial for effectively harnessing the vast computational resources of modern HPC systems
Mattei, Pierre-Alexandre. "Sélection de modèles parcimonieux pour l’apprentissage statistique en grande dimension." Thesis, Sorbonne Paris Cité, 2017. http://www.theses.fr/2017USPCB051/document.
Full textThe numerical surge that characterizes the modern scientific era led to the rise of new kinds of data united in one common immoderation: the simultaneous acquisition of a large number of measurable quantities. Whether coming from DNA microarrays, mass spectrometers, or nuclear magnetic resonance, these data, usually called high-dimensional, are now ubiquitous in scientific and technological worlds. Processing these data calls for an important renewal of the traditional statistical toolset, unfit for such frameworks that involve a large number of variables. Indeed, when the number of variables exceeds the number of observations, most traditional statistics becomes inefficient. First, we give a brief overview of the statistical issues that arise with high-dimensional data. Several popular solutions are presented, and we present some arguments in favor of the method utilized and advocated in this thesis: Bayesian model uncertainty. This chosen framework is the subject of a detailed review that insists on several recent developments. After these surveys come three original contributions to high-dimensional model selection. A new algorithm for high-dimensional sparse regression called SpinyReg is presented. It compares favorably to state-of-the-art methods on both real and synthetic data sets. A new data set for high-dimensional regression is also described: it involves predicting the number of visitors in the Orsay museum in Paris using bike-sharing data. We focus next on model selection for high-dimensional principal component analysis (PCA). Using a new theoretical result, we derive the first closed-form expression of the marginal likelihood of a PCA model. This allows us to propose two algorithms for model selection in PCA. A first one called globally sparse probabilistic PCA (GSPPCA) that allows to perform scalable variable selection, and a second one called normal-gamma probabilistic PCA (NGPPCA) that estimates the intrinsic dimensionality of a high-dimensional data set. Both methods are competitive with other popular approaches. In particular, using unlabeled DNA microarray data, GSPPCA is able to select genes that are more biologically relevant than several popular approaches
Papanastasiou, Effrosyni. "Feasibility of Interactions and Network Inference of Online Social Networks." Electronic Thesis or Diss., Sorbonne université, 2024. http://www.theses.fr/2024SORUS173.
Full textThis thesis deals with the problem of network inference in the domain of Online So-cial Networks. The main premise of network inference problems is that the networkwe are observing is not the network that we really need. This is especially prevalentin today's digital space, where the abundance of information usually comes withcrucial unreliability, in the form of noise and missing points in the data. However, existing approaches either ignore or do not guarantee to infer networks in a waythat can explain the data we have at hand. As a result, there is an ambiguity around the meaning of the network that we are inferring, while also having little intuition or control over the inference itself. The goal of this thesis is to further explore this problem. To quantify how well an inferred network can explain a dataset, we introduce a novel quality criterion called feasibility. Our intuition is that if a dataset is feasible given an inferred network, we might also be closer to the ground truth. To verify this,we propose a novel network inference method in the form of a constrained, Maximum Likelihood-based optimization problem that guarantees 100% feasibility. It is tailored to inputs from Online Social Networks, which are well-known sources of un-reliable and restricted data. We provide extensive experiments on one synthetic andone real-world dataset coming from Twitter/X. We show that our proposed method generates a posterior distribution of graphs that guarantees to explain the dataset while also being closer to the true underlying structure when compared to other methods. As a final exploration, we look into the field of deep learning for more scalable and flexible alternatives, providing a preliminary framework based on Graph Neural Networks and contrastive learning that gives promising results
Granjal, Cruz Gonçalo Jorge. "Development and validation of a bayesian measurement technique for data-driven measurement reduction." Electronic Thesis or Diss., Ecully, Ecole centrale de Lyon, 2024. http://www.theses.fr/2024ECDL0012.
Full textThis work presents a complete hybrid testing methodology for assessing the flow in turbomachinery components. Focused on minimizing testing times and instrumentation requirements, the methodology strategically integrates standard experimental measurements with numerical simulations, specifically employing Multi-Fidelity Gaussian Processes, Sparse Variational Gaussian Processes, and adaptive Bayesian optimization.The methodology systematically reduces both instrumentation efforts and testing times, providing uncertainty metrics comparable to traditional methodologies. Applied initially to a benchmarked axial high-pressure compressor (H25) and afterwards to an ultra-high bypass ratio fan (ECL5 UHBR) in blind test conditions, the methodology demonstrates robustness, adaptability, and significant reductions in measurement points and testing times leading to a direct impact in experimental campaign costs.For the H25 axial compressor, the proposed framework proves capable of predicting flow fields, emphasizing the trade-off between high-fidelity measurements and mean flow prediction accuracy. The ECL5 UHBR fan blind test results validate the methodology's efficiency in aerodynamic assessments and demonstrates time savings of at least one hour per operating condition.The a priori Design of Experiments achieves at least a 50% reduction in measurements, outperforming random sampling, and effectively assists in experimental campaign planning. The In situ adaptive sampling outperforms random sampling by up to 44%, showcasing accurate detection of flow phenomena and promising applications in achieving high accuracy experimental demands. The modular and adaptable nature of the methodology positions it for broad application in both academic and industrial settings, while its exploitation opens paths to infer unmeasured flow quantities or improve performance evaluation measurements.This work introduces a paradigm shift in experimental campaign planning, optimizing measurement budgets strategically beforehand or enhancing accuracy dynamically during a campaign, emphasizing the potential of machine learning-driven trends in shaping new research paths
Morvant, Emilie. "Apprentissage de vote de majorité pour la classification supervisée et l'adaptation de domaine : approches PAC-Bayésiennes et combinaison de similarités." Phd thesis, Aix-Marseille Université, 2013. http://tel.archives-ouvertes.fr/tel-00879072.
Full textCharon, Clara. "Classification probabiliste pour la prédiction et l'explication d'événements de santé défavorables et évitables en EHPAD." Electronic Thesis or Diss., Sorbonne université, 2024. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2024SORUS200.pdf.
Full textNursing homes, which provide housing for dependent elderly people,are an option used by a large and growing population when, for a variety of reasons, including health, it is no longer possible for them to live at home.With the development of new information technologies in the health sector, an increasing number of health care facilities are equipped with information systems that group together administrative and medical data of patients as well as information on the care they receive. Among these systems, electronic health records (EHRs) have emerged as essential tools, providing quick and easy access to patient information in order to improve the quality and safety of care.We use the anonymized data of the EHRs from NETSoins, a software widely used in nursing homes in France, to propose and analyze classifiers capable of predicting several adverse health events in the elderly that are potentially modifiable by appropriate health interventions. Our approach focuses in particular on the use of methods that can provide explanations, such as probabilistic graphical models, including Bayesian networks.After a complex preprocessing step to adapt event-based data into data suitable for statistical learning while preserving their medical coherence, we have developed a learning method applied in three probabilistic classification experiments using Bayesian networks, targeting different events: the risk of occurrence of the first pressure ulcer, the risk of emergency hospitalization upon the resident's entry into the nursing home, and the risk of fracture in the first months of housing.For each target, we have compared the performance of our Bayesian network classifier according to various criteria with other machine learning methods as well as with the practices currently used in nursing homes to predict these risks. We have also compared the results of the Bayesian networks with clinical expertise.This study demonstrates the possibility of predicting these events from the data already collected in routine by caregivers, thus paving the way for new predictive tools that can be integrated directly into the software already used by these professionals
Rouillard, Louis. "Bridging Simulation-based Inference and Hierarchical Modeling : Applications in Neuroscience." Electronic Thesis or Diss., université Paris-Saclay, 2024. http://www.theses.fr/2024UPASG024.
Full textNeuroimaging investigates the brain's architecture and function using magnetic resonance (MRI). To make sense of the complex observed signal, Neuroscientists posit explanatory models, governed by interpretable parameters. This thesis tackles statistical inference : guessing which parameters could have yielded the signal through the model.Inference in Neuroimaging is complexified by at least three hurdles : a large dimensionality, a large uncertainty, and the hierarchcial structure of data. We look into variational inference (VI) as an optimization-based method to tackle this regime.Specifically, we conbine structured stochastic VI and normalizing flows (NFs) to design expressive yet scalable variational families. We apply those techniques in diffusion and functional MRI, on tasks including individual parcellation, microstructure inference and directional coupling estimation. Through these applications, we underline the interplay between the forward and reverse Kullback-Leibler (KL) divergences as complemen-tary tools for inference. We also demonstrate the ability of automatic VI (AVI) as a reliable and scalable inference method to tackle the challenges of model-driven Neuroscience
Fond, Antoine. "Localisation par l'image en milieu urbain : application à la réalité augmentée." Electronic Thesis or Diss., Université de Lorraine, 2018. http://www.theses.fr/2018LORR0028.
Full textThis thesis addresses the problem of localization in urban areas. Inferring accurate positioning in the city is important in many applications such as augmented reality or mobile robotics. However, systems based on inertial sensors (IMUs) are subject to significant drifts and GPS data can suffer from a valley effect that limits their accuracy. A natural solution is to rely on the camera pose estimation in computer vision. We notice that buildings are the main visual landmarks of human beings but also objects of interest for augmented reality applications. We therefore aim to compute the camera pose relatively to a database of known reference buildings from a single image. The problem is twofold : find the visible references in the current image (place recognition) and compute the camera pose relatively to them. Conventional approaches to these two sub-problems are challenged in urban environments due to strong perspective effects, frequent repetitions and visual similarity between facades. While specific approaches to these environments have been developed that exploit the high structural regularity of such environments, they still suffer from a number of limitations in terms of detection and recognition of facades as well as pose computation through model registration. The original method developed in this thesis is part of these specific approaches and aims to overcome these limitations in terms of effectiveness and robustness to clutter and changes of viewpoints and illumination. For do so, the main idea is to take advantage of recent advances in deep learning by convolutional neural networks to extract high-level information on which geometric models can be based. Our approach is thus mixed Bottom- Up/Top-Down and is divided into three key stages. We first propose a method to estimate the rotation of the camera pose. The 3 main vanishing points of the image of urban environnement, known as Manhattan vanishing points, are detected by a convolutional neural network (CNN) that estimates both these vanishing points and the image segmentation relative to them. A second refinement step uses this information and image segmentation in a Bayesian model to estimate these points effectively and more accurately. By estimating the camera’s rotation, the images can be rectified and thus free from perspective effects to find the translation. In a second contribution, we aim to detect the facades in these rectified images to recognize them among a database of known buildings and estimate a rough translation. For the sake of efficiency, a series of cues based on facade specific characteristics (repetitions, symmetry, semantics) have been proposed to enable the fast selection of facade proposals. Then they are classified as facade or non-facade according to a new contextual CNN descriptor. Finally, the matching of the detected facades to the references is done by a nearest neighbor search using a metric learned on these descriptors. Eventually we propose a method to refine the estimation of the translation relying on the semantic segmentation inferred by a CNN for its robustness to changes of illumination ans small deformations. If we can already estimate a rough translation from these detected facades, we choose to refine this result by relying on the se- mantic segmentation of the image inferred from a CNN for its robustness to changes of illuminations and small deformations. Since the facade is identified in the previous step, we adopt a model-based approach by registration. Since the problems of registration and segmentation are linked, a Bayesian model is proposed which enables both problems to be jointly solved. This joint processing improves the results of registration and segmentation while remaining efficient in terms of computation time. These three parts have been validated on consistent community data sets. The results show that our approach is fast and more robust to changes in shooting conditions than previous methods
Faour, Ahmad. "Une architecture semi-supervisée et adaptative pour le filtrage d'alarmes dans les systèmes de détection d'intrusions sur les réseaux." Phd thesis, INSA de Rouen, 2007. http://tel.archives-ouvertes.fr/tel-00917605.
Full textDulac, Adrien. "Etude des modèles à composition mixée pour l'analyse de réseaux complexes." Thesis, Université Grenoble Alpes (ComUE), 2018. http://www.theses.fr/2018GREAM080/document.
Full textRelational data are ubiquitous in the nature and their accessibility has not ceased to increase in recent years. Those data, see as a whole, form a network, which can be represented by a data structure called a graph, where each vertex of the graph is an entity and each edge a connection between pair of vertices. Complex networks in general, such as the Web, communication networks or social network, are known to exhibit common structural properties that emerge through their graphs. In this work we emphasize two important properties called *homophilly* and *preferential attachment* that arise on most of the real-world networks. We firstly study a class of powerful *random graph models* in a Bayesian nonparametric setting, called *mixed-membership model* and we focus on showing whether the models in this class comply with the mentioned properties, after giving formal definitions in a probabilistic context of the latter. Furthermore, we empirically evaluate our findings on synthetic and real-world network datasets. Secondly, we propose a new model, which extends the former Stochastic Mixed-Membership Model, for weighted networks and we develop an efficient inference algorithm able to scale to large-scale networks
Fond, Antoine. "Localisation par l'image en milieu urbain : application à la réalité augmentée." Thesis, Université de Lorraine, 2018. http://www.theses.fr/2018LORR0028/document.
Full textThis thesis addresses the problem of localization in urban areas. Inferring accurate positioning in the city is important in many applications such as augmented reality or mobile robotics. However, systems based on inertial sensors (IMUs) are subject to significant drifts and GPS data can suffer from a valley effect that limits their accuracy. A natural solution is to rely on the camera pose estimation in computer vision. We notice that buildings are the main visual landmarks of human beings but also objects of interest for augmented reality applications. We therefore aim to compute the camera pose relatively to a database of known reference buildings from a single image. The problem is twofold : find the visible references in the current image (place recognition) and compute the camera pose relatively to them. Conventional approaches to these two sub-problems are challenged in urban environments due to strong perspective effects, frequent repetitions and visual similarity between facades. While specific approaches to these environments have been developed that exploit the high structural regularity of such environments, they still suffer from a number of limitations in terms of detection and recognition of facades as well as pose computation through model registration. The original method developed in this thesis is part of these specific approaches and aims to overcome these limitations in terms of effectiveness and robustness to clutter and changes of viewpoints and illumination. For do so, the main idea is to take advantage of recent advances in deep learning by convolutional neural networks to extract high-level information on which geometric models can be based. Our approach is thus mixed Bottom- Up/Top-Down and is divided into three key stages. We first propose a method to estimate the rotation of the camera pose. The 3 main vanishing points of the image of urban environnement, known as Manhattan vanishing points, are detected by a convolutional neural network (CNN) that estimates both these vanishing points and the image segmentation relative to them. A second refinement step uses this information and image segmentation in a Bayesian model to estimate these points effectively and more accurately. By estimating the camera’s rotation, the images can be rectified and thus free from perspective effects to find the translation. In a second contribution, we aim to detect the facades in these rectified images to recognize them among a database of known buildings and estimate a rough translation. For the sake of efficiency, a series of cues based on facade specific characteristics (repetitions, symmetry, semantics) have been proposed to enable the fast selection of facade proposals. Then they are classified as facade or non-facade according to a new contextual CNN descriptor. Finally, the matching of the detected facades to the references is done by a nearest neighbor search using a metric learned on these descriptors. Eventually we propose a method to refine the estimation of the translation relying on the semantic segmentation inferred by a CNN for its robustness to changes of illumination ans small deformations. If we can already estimate a rough translation from these detected facades, we choose to refine this result by relying on the se- mantic segmentation of the image inferred from a CNN for its robustness to changes of illuminations and small deformations. Since the facade is identified in the previous step, we adopt a model-based approach by registration. Since the problems of registration and segmentation are linked, a Bayesian model is proposed which enables both problems to be jointly solved. This joint processing improves the results of registration and segmentation while remaining efficient in terms of computation time. These three parts have been validated on consistent community data sets. The results show that our approach is fast and more robust to changes in shooting conditions than previous methods
Jouffroy, Emma. "Développement de modèles non supervisés pour l'obtention de représentations latentes interprétables d'images." Electronic Thesis or Diss., Bordeaux, 2024. http://www.theses.fr/2024BORD0050.
Full textThe Laser Megajoule (LMJ) is a large research device that simulates pressure and temperature conditions similar to those found in stars. During experiments, diagnostics are guided into an experimental chamber for precise positioning. To minimize the risks associated with human error in such an experimental context, the automation of an anti-collision system is envisaged. This involves the design of machine learning tools offering reliable decision levels based on the interpretation of images from cameras positioned in the chamber. Our research focuses on probabilistic generative neural methods, in particular variational auto-encoders (VAEs). The choice of this class of models is linked to the fact that it potentially enables access to a latent space directly linked to the properties of the objects making up the observed scene. The major challenge is to study the design of deep network models that effectively enable access to such a fully informative and interpretable representation, with a view to system reliability. The probabilistic formalism intrinsic to VAE allows us, if we can trace back to such a representation, to access an analysis of the uncertainties of the encoded information
Cottet, Vincent R. "Theoretical study of some statistical procedures applied to complex data." Electronic Thesis or Diss., Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLG002.
Full textThe main part of this thesis aims at studying the theoretical and algorithmic aspects of three distinct statistical procedures. The first problem is the binary matrix completion. We propose an estimator based on a variational approximation of a pseudo-Bayesian estimator. We use a different loss function of the ones used in the literature. We are able to compute non asymptotic risk bounds. It is much faster to compute the estimator than a MCMC method and we show on examples that it is efficient in practice. In a second part we study the theoretical properties of the regularized empirical risk minimizer for Lipschitz loss functions. We are therefore able to apply it on the logistic regression with the SLOPE regularization and on the matrix completion as well. The third chapter develops an Expectation-Propagation approximation when the likelihood is not explicit. We then use an ABC approximation in a second stage. This procedure may be applied to many models and is more precise and faster than the classic ABC approximation. It is used in a spatial extremes model
Remy, Benjamin. "Generative modeling for weak lensing inverse problems." Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASP163.
Full textGravitational lensing, which is the effect of the distortion of distant galaxy images through the influence of massive matter densities in the line of sight, holds significant promise in addressing questions about dark matter and dark energy. It reflects the distribution of total matter of the Universe and is therefore a promising probe for cosmological models. In the case where these distortions are small, we call it the weak gravitational lensing regime and a straightforward mapping exists between the matter distribution projected in the line of sight, called mass-map, and the measured lensing effect. However, when attempting to reconstruct matter mass-maps under conditions involving missing data and high noise corruption, this linear inverse problem becomes ill-posed and may lack a meaningful solution without additional prior knowledge. The main objective of this thesis is to employ recent breakthroughs in the generative modeling literature that enable the modeling of complex distribution in high-dimensional spaces. We propose in particular a novel methodology to solve high-dimensional ill-posed inverse problems, characterizing the full posterior distribution of the problem. By learning the high dimensional prior from cosmological simulations, we demonstrate that we are able to reconstruct high-resolution 2D mass-maps alongside uncertainty quantification. Additionally, we present a new method for cosmic shear estimation based on forward modeling of the observation at the pixel level. This represents a new paradigm for weak lensing measurement as it does not rely on galaxy ellipticities anymore. In particular, we propose to build a hybrid generative and physical hierarchical Bayesian model and demonstrate that we can remove the source of model bias in the estimation of the cosmic shear
Cumin, Julien. "Reconnaissance et prédiction d'activités dans la maison connectée." Thesis, Université Grenoble Alpes (ComUE), 2018. http://www.theses.fr/2018GREAM071/document.
Full textUnderstanding the context of a home is essential in order to provide services to occupants that fit their situations and thus fulfil their needs. One example of service that such a context-aware smart home could provide is that of a communication assistant, which can for example advise correspondents outside the home on the availability for communication of occupants. In order to implement such a service, it is indeed required that the home understands the situations of occupants, in order to derive their availability.In this thesis, we first propose a definition of context in homes. We argue that one of the primary context dimensions necessary for a system to be context-aware is the activity of occupants. As such, we then study the problem of recognizing activities, from ambient smart home sensors. We propose a new supervised place-based approach which both improves activity recognition accuracy as well as computing times compared to standard approaches.Smart home services, such as our communication assistance example, may often need to anticipate future situations. In particular, they need to anticipate future activities of occupants. Therefore, we design a new supervised activity prediction model, based on previous state-of-the-art work. We propose a number of extensions to improve prediction accuracy based on the specificities of smart home environments.Finally, we study the problem of inferring the availability of occupants for communication, in order to illustrate the feasibility of our communication assistant example. We argue that availability can be inferred from primary context dimensions such as place and activity (which can be recognized or predicted using our previous contributions), and by taking into consideration the correspondent initiating the communication as well as the modality of communication used. We discuss the impact of the activity recognition step on availability inference.We evaluate those contributions on various state-of-the-art datasets, as well as on a new dataset of activities and availabilities in homes which we constructed specifically for the purposes of this thesis: Orange4Home. Through our contributions to these 3 problems, we demonstrate the way in which an example context-aware communication assistance service can be implemented, which can advise on future availability for communication of occupants. More generally, we show how secondary context dimensions such as availability can be inferred from other context dimensions, in particular from activity. Highly accurate activity recognition and prediction are thus mandatory for a smart home to achieve context awareness
Ndour, Cheikh. "Modélisation statistique de la mortalité maternelle et néonatale pour l'aide à la planification et à la gestion des services de santé en Afrique Sub-Saharienne." Phd thesis, Université de Pau et des Pays de l'Adour, 2014. http://tel.archives-ouvertes.fr/tel-00996996.
Full textKiendrébéogo, Ramodgwendé Weizmann. "Développements pour l'observation et la caractérisation des sources multi-messagers d'ondes gravitationnelles lors des campagnes d'observation LIGO-Virgo-KAGRA." Electronic Thesis or Diss., Université Côte d'Azur, 2024. http://www.theses.fr/2024COAZ5034.
Full textThe Advanced LIGO/Virgo observation campaigns have revealed the rich and diverse physics of binary neutron star (BNS) and binary black hole mergers. In 2017, the simultaneous discovery of GWs and electromagnetic (EM) counterparts from a BNS merger provided an exceptionally detailed view of this extreme phenomenon, yielding numerous results in both astrophysics and physics, particularly on the behavior of ultra-dense matter. However, despite enormous efforts, no new multi-messenger detections have been made since. This is due to the formidable observational challenge posed by the rapid and precise alerts of GWs, the immediate reactivity of a network of telescopes, and the online data processing required for the identification of EM counterparts.The identification of EM counterparts enables numerous high-priority scientific studies, such as constraints on the equation of state of neutron stars, the measurement of the universe's expansion rate, and the r-process nucleosynthesis of heavy elements produced during a kilonova. For a rapid follow-up of possible counterparts to these events, we must reduce the sky-localization area where the event occurs. However, the significantly different sensitivities of the detectors demonstrate how challenging gravitational-wave (GW) follow-up can be. This is the case for the fourth (ongoing) and fifth LIGO/Virgo/KAGRA (LVK) observation campaigns. Many GW signals from compact binary mergers are hidden by detector noise and can be detected if the noise is sufficiently reduced. To maximize the scientific outcome of the LVK GW detectors, such as the detectability of pre-merger signals, noise must be significantly reduced. Several factors contribute to this noise, undermining the detector's sensitivity, including environmental noise, instrumental artifacts, and some more fundamental and irreducible noises. The identification of additional sub-threshold events is therefore linked to our ability to reduce noise in the instruments. Noise and sensitivity directly influence our capacity to extract information from GW signals.To mitigate these effects, I initially developed new tools and techniques while also making several improvements to existing ones. These analysis tools include, among others, i) enhancing the capabilities of the Nuclear Multi-messenger Astronomy (NMMA), a Python library for probing nuclear physics and cosmology with multi-messenger analysis; ii) updating and configuring telescopes such as the Zwicky Transient Facility (ZTF), the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST), and the Ultraviolet Transient Astronomy Satellite (ULTRASAT) within Gravitational-wave Electromagnetic Optimization (gwemopt), a tool for simulating detections using a telescope and event sky map information; iii) injecting a new distribution, PBD/GWTC-3, into Ligo.Skymap for "observing scenarios". This new distribution can define all populations of compact binary coalescences with a single law; iv) developing NMMA-Skyportal, a pipeline that integrates ZTF alerts, the Skyportal tool, a collaborative platform for time-domain astronomy, and NMMA to discriminate the nature of light curves in real-time.Moreover, this work provides projections for astronomers interested in data produced by GW detectors, as well as expected constraints on the universe's expansion rate based on forthcoming data. These results are useful to those analyzing GW data and those seeking EM counterparts to neutron star mergers. Finally, to address the problem of "astrophysical signals bathing" below the noise threshold, I applied the DeepClean algorithm, a one-dimensional convolutional neural network, to estimate, analyze and subtract stationary and non-stationary noises in the Virgo detector. A first for the Virgo detector. In addition to preserving the integrity of the astrophysical signal, the algorithm improves the detector's signal-to-noise ratio
Deregnaucourt, Thomas. "Prédiction spatio-temporelle de surfaces issues de l'imagerie en utilisant des processus stochastiques." Thesis, Université Clermont Auvergne (2017-2020), 2019. http://www.theses.fr/2019CLFAC088.
Full textThe prediction of a surface is now an important problem due to its use in multiple domains, such as computer vision, the simulation of avatars for cinematography or video games, etc. Since a surface can be static or dynamic, i.e. evolving with time, this problem can be separated in two classes: a spatial prediction problem and a spatio-temporal one. In order to propose a new approach for each of these problems, this thesis works have been separated in two parts.First of all, we have searched to predict a static surface, which is supposed cylindrical, knowing it partially from curves. The proposed approach consisted in deforming a cylinder on the known curves in order to reconstruct the surface of interest. First, a correspondence between known curves and the cylinder is generated with the help of shape analysis tools. Once this step done, an interpolation of the deformation field, which is supposed Gaussian, have been estimated using maximum likelihood and Bayesian inference. This methodology has then been applied to real data from two domains of imaging: medical imaging and infography. The obtained results show that the proposed approach exceeds the existing methods in the literature, with better results using Bayesian inference.In a second hand, we have been interested in the spatio-temporal prediction of dynamic surfaces. The objective was to predict a dynamic surface based on its initial surface. Since the prediction needs to learn on known observations, we first have developed a spatio-temporal surface analysis tool. This analysis is based on shape analysis tools, and allows a better learning. Once this preliminary step done, we have estimated the temporal deformation of the dynamic surface of interest. More precisely, an adaptation, with is usable on the space of surfaces, of usual statistical estimators has been used. Using this estimated deformation on the initial surface, an estimation of the dynamic surface has been created. This process has then been applied for predicting 4D expressions of faces, which allow us to generate visually convincing expressions
Gerchinovitz, Sébastien. "Prédiction de suites individuelles et cadre statistique classique : étude de quelques liens autour de la régression parcimonieuse et des techniques d'agrégation." Phd thesis, Université Paris Sud - Paris XI, 2011. http://tel.archives-ouvertes.fr/tel-00653550.
Full textFrichot, Eric. "Modèles à facteurs latents pour les études d'association écologique en génétique des populations." Thesis, Grenoble, 2014. http://www.theses.fr/2014GRENS018/document.
Full textWe introduce a set of latent factor models dedicated to landscape genomics and ecological association tests. It includes statistical methods for correcting principal component maps for effects of spatial autocorrelation (spFA); methods for estimating ancestry coefficients from large genotypic matrices and evaluating the number of ancestral populations (sNMF); and methods for identifying genetic polymorphisms that exhibit high correlation with some environmental gradient or with the variables used as proxies for ecological pressures (LFMM). We also developed a set of open source softwares associated with the methods, based on optimized C programs that can scale with the dimension of very large data sets, to run analyses of population structure and genome scans for local adaptation
Willot, Hénoïk. "Certified explanations of robust models." Electronic Thesis or Diss., Compiègne, 2024. http://www.theses.fr/2024COMP2812.
Full textWith the advent of automated or semi-automated decision systems in artificial intelligence comes the need of making them more reliable and transparent for an end-user. While the role of explainable methods is in general to increase transparency, reliability can be achieved by providing certified explanations, in the sense that those are guaranteed to be true, and by considering robust models that can abstain when having insufficient information, rather than enforcing precision for the mere sake of avoiding indecision. This last aspect is commonly referred to as skeptical inference. This work participates to this effort, by considering two cases: - The first one considers classical decision rules used to enforce fairness, which are the Ordered Weighted Averaging (OWA) with decreasing weights. Our main contribution is to fully characterise from an axiomatic perspective convex sets of such rules, and to provide together with this sound and complete explanation schemes that can be efficiently obtained through heuristics. Doing so, we also provide a unifying framework between the restricted and generalized Lorenz dominance, two qualitative criteria, and precise decreasing OWA. - The second one considers that our decision rule is a classification model resulting from a learning procedure, where the resulting model is a set of probabilities. We study and discuss the problem of providing prime implicant as explanations in such a case, where in addition to explaining clear preferences of one class over the other, we also have to treat the problem of declaring two classes as being incomparable. We describe the corresponding problems in general ways, before studying in more details the robust counter-part of the Naive Bayes Classifier
Khomh, Foutse. "Patterns and quality of object-oriented software systems." Thèse, 2010. http://hdl.handle.net/1866/4601.
Full textMaintenance costs during the past decades have reached more than 70% of the overall costs of object-oriented systems, because of many factors, such as changing software environments, changing users' requirements, and the overall quality of systems. One factor on which we have a control is the quality of systems. Many object-oriented software quality models have been introduced in the literature to help assess and control quality. However, these models usually use metrics of classes (such as number of methods) or of relationships between classes (for example coupling) to measure internal attributes of systems. Yet, the quality of object-oriented systems does not depend on classes' metrics solely: it also depends on the organisation of classes, i.e. the system design that concretely manifests itself through design styles, such as design patterns and antipatterns. In this dissertation, we propose the method DEQUALITE to systematically build quality models that take into account the internal attributes of the systems (through metrics) but also their design (through design patterns and antipatterns). This method uses a machine learning approach based on Bayesian Belief Networks and builds on the results of a series of experiments aimed at evaluating the impact of design patterns and antipatterns on the quality of systems. These experiments, performed on 9 large object-oriented open source systems enable us to draw the following conclusions: • Counter-intuitively, design patterns do not always improve the quality of systems; tangled implementations of design patterns for example significantly affect the structure of classes and negatively impact their change- and fault-proneness. • Classes participating in antipatterns are significantly more likely to be subject to changes and to be involved in fault-fixing changes than other classes. • A non negligible percentage of classes participate in co-occurrences of antipatterns and design patterns in systems. On these classes, design patterns have a positive effect in mitigating antipatterns. We apply and validate our method on three open-source object-oriented systems to demonstrate the contribution of the design of system in quality assessment.