Dissertations / Theses on the topic 'Apprentissage profond – Réseaux neuronaux (informatique)'
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 profond – Réseaux neuronaux (informatique).'
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.
Haykal, Vanessa. "Modélisation des séries temporelles par apprentissage profond." Thesis, Tours, 2019. http://www.theses.fr/2019TOUR4019.
Full textTime series prediction is a problem that has been addressed for many years. In this thesis, we have been interested in methods resulting from deep learning. It is well known that if the relationships between the data are temporal, it is difficult to analyze and predict accurately due to non-linear trends and the existence of noise specifically in the financial and electrical series. From this context, we propose a new hybrid noise reduction architecture that models the recursive error series to improve predictions. The learning process fusessimultaneouslyaconvolutionalneuralnetwork(CNN)andarecurrentlongshort-term memory network (LSTM). This model is distinguished by its ability to capture globally a variety of hybrid properties, where it is able to extract local signal features, to learn long-term and non-linear dependencies, and to have a high noise resistance. The second contribution concerns the limitations of the global approaches because of the dynamic switching regimes in the signal. We present a local unsupervised modification with our previous architecture in order to adjust the results by adapting the Hidden Markov Model (HMM). Finally, we were also interested in multi-resolution techniques to improve the performance of the convolutional layers, notably by using the variational mode decomposition method (VMD)
Caron, Stéphane. "Détection d'anomalies basée sur les représentations latentes d'un autoencodeur variationnel." Master's thesis, Université Laval, 2021. http://hdl.handle.net/20.500.11794/69185.
Full textIn this master's thesis, we propose a methodology that aims to detect anomalies among complex data, such as images. In order to do that, we use a specific type of neural network called the varitionnal autoencoder (VAE). This non-supervised deep learning approach allows us to obtain a simple representation of our data on which we then use the Kullback-Leibler distance to discriminate between anomalies and "normal" observations. To determine if an image should be considered "abnormal", our approach is based on a proportion of observations to be filtered, which is easier and more intuitive to establish than applying a threshold based on the value of a distance metric. By using our methodology on real complex images, we can obtain superior anomaly detection performances in terms of area under the ROC curve (AUC),precision and recall compared to other non-supervised methods. Moreover, we demonstrate that the simplicity of our filtration level allows us to easily adapt the method to datasets having different levels of anomaly contamination.
Ostertag, Cécilia. "Analyse des pathologies neuro-dégénératives par apprentissage profond." Thesis, La Rochelle, 2022. http://www.theses.fr/2022LAROS003.
Full textMonitoring and predicting the cognitive state of a subject affected by a neuro-degenerative disorder is crucial to provide appropriate treatment as soon as possible. Thus, these patients are followed for several years, as part of longitudinal medical studies. During each visit, a large quantity of data is acquired : risk factors linked to the pathology, medical imagery (MRI or PET scans for example), cognitive tests results, sampling of molecules that have been identified as bio-markers, etc. These various modalities give information about the disease's progression, some of them are complementary and others can be redundant. Several deep learning models have been applied to bio-medical data, notably for organ segmentation or pathology diagnosis. This PhD is focused on the conception of a deep neural network model for cognitive decline prediction, using multimodal data, here both structural brain MRI images and clinical data. In this thesis we propose an architecture made of sub-modules tailored to each modality : 3D convolutional network for the brain MRI, and fully connected layers for the quantitative and qualitative clinical data. To predict the patient's evolution, this model takes as input data from two medical visits for each patient. These visits are compared using a siamese architecture. After training and validating this model with Alzheimer's disease as our use case, we look into knowledge transfer to other neuro-degenerative pathologies, and we use transfer learning to adapt our model to Parkinson's disease. Finally, we discuss the choices we made to take into account the temporal aspect of our problem, both during the ground truth creation using the long-term evolution of a cognitive score, and for the choice of using pairs of visits as input instead of longer sequences
Mercier, Jean-Philippe. "Deep learning for object detection in robotic grasping contexts." Doctoral thesis, Université Laval, 2021. http://hdl.handle.net/20.500.11794/69801.
Full textIn the last decade, deep convolutional neural networks became a standard for computer vision applications. As opposed to classical methods which are based on rules and hand-designed features, neural networks are optimized and learned directly from a set of labeled training data specific for a given task. In practice, both obtaining sufficient labeled training data and interpreting network outputs can be problematic. Additionnally, a neural network has to be retrained for new tasks or new sets of objects. Overall, while they perform really well, deployment of deep neural network approaches can be challenging. In this thesis, we propose strategies aiming at solving or getting around these limitations for object detection. First, we propose a cascade approach in which a neural network is used as a prefilter to a template matching approach, allowing an increased performance while keeping the interpretability of the matching method. Secondly, we propose another cascade approach in which a weakly-supervised network generates object-specific heatmaps that can be used to infer their position in an image. This approach simplifies the training process and decreases the number of required training images to get state-of-the-art performances. Finally, we propose a neural network architecture and a training procedure allowing detection of objects that were not seen during training, thus removing the need to retrain networks for new objects.
Boussaha, Basma El Amel. "Response selection for end-to-end retrieval-based dialogue systems." Thesis, Nantes, 2019. http://www.theses.fr/2019NANT4080.
Full textThe increasing need of human assistance pushed researchers to develop automatic, smart and tireless dialogue systems that can converse with humans in natural language to be either their virtual assistant or their chat companion. The industry of dialogue systems has been very popular in the last decade and many systems from industry and academia have been developed. In this thesis, we study retrieval-based dialogue systems which aim to find the most appropriate response to the conversation among a set of predefined responses. The main challenge of these systems is to understand the conversation and identify the elements that describe the problem and the solution which are usually implicit. Most of the recent approaches are based on deep learning techniques which can automatically capture implicit information. However these approaches are either complex or domain dependent. We propose a simple, end-to-end and efficient retrieval-based dialogue system that first matches the response with the history of the conversation on the sequence-level and then we extend the system to multiple levels while keeping the architecture simple and domain independent. We perform several analyzes to determine possible improvements
Katranji, Mehdi. "Apprentissage profond de la mobilité des personnes." Thesis, Bourgogne Franche-Comté, 2019. http://www.theses.fr/2019UBFCA024.
Full textKnowledge of mobility is a major challenge for authorities mobility organisers and urban planning. Due to the lack of formal definition of human mobility, the term "people's mobility" will be used in this book. This topic will be introduced by a description of the ecosystem by considering these actors and applications.The creation of a learning model has prerequisites: an understanding of the typologies of the available data sets, their strengths and weaknesses. This state of the art in mobility knowledge is based on the four-step model that has existed and been used since 1970, ending with the renewal of the methodologies of recent years.Our models of people's mobility are then presented. Their common point is the emphasis on the individual, unlike traditional approaches that take the locality as a reference. The models we propose are based on the fact that the intake of individuals' decisions is based on their perception of the environment.This finished book on the study of the deep learning methods of Boltzmann machines restricted. After a state of the art of this family of models, we are looking for strategies to make these models viable in the application world. This last chapter is our contribution main theoretical, by improving robustness and performance of these models
Sablayrolles, Alexandre. "Mémorisation et apprentissage de structures d'indexation avec les réseaux de neurones." Thesis, Université Grenoble Alpes, 2020. https://thares.univ-grenoble-alpes.fr/2020GRALM044.pdf.
Full textMachine learning systems, and in particular deep neural networks, aretrained on large quantities of data. In computer vision for instance, convolutionalneural networks used for image classification, scene recognition,and object detection, are trained on datasets which size ranges from tensof thousands to billions of samples. Deep parametric models have a largecapacity, often in the order of magnitude of the number of datapoints.In this thesis, we are interested in the memorization aspect of neuralnetworks, under two complementary angles: explicit memorization,i.e. memorization of all samples of a set, and implicit memorization,that happens inadvertently while training models. Considering explicitmemorization, we build a neural network to perform approximate setmembership, and show that the capacity of such a neural network scaleslinearly with the number of data points. Given such a linear scaling, weresort to another construction for set membership, in which we build aneural network to produce compact codes, and perform nearest neighborsearch among the compact codes, thereby separating “distribution learning”(the neural network) from storing samples (the compact codes), theformer being independent of the number of samples and the latter scalinglinearly with a small constant. This nearest neighbor system performs amore generic task, and can be plugged in to perform set membership.In the second part of this thesis, we analyze the “unintended” memorizationthat happens during training, and assess if a particular data pointwas used to train a model (membership inference). We perform empiricalmembership inference on large networks, on both individual and groupsof samples. We derive the Bayes-optimal membership inference, andconstruct several approximations that lead to state-of-the-art results inmembership attacks. Finally, we design a new technique, radioactive data,that slightly modifies datasets such that any model trained on them bearsan identifiable mark
Groueix, Thibault. "Learning 3D Generation and Matching." Thesis, Paris Est, 2020. http://www.theses.fr/2020PESC1024.
Full textThe goal of this thesis is to develop deep learning approaches to model and analyse 3D shapes. Progress in this field could democratize artistic creation of 3D assets which currently requires time and expert skills with technical software.We focus on the design of deep learning solutions for two particular tasks, key to many 3D modeling applications: single-view reconstruction and shape matching.A single-view reconstruction (SVR) method takes as input a single image and predicts the physical world which produced that image. SVR dates back to the early days of computer vision. In particular, in the 1960s, Lawrence G. Roberts proposed to align simple 3D primitives to the input image under the assumption that the physical world is made of cuboids. Another approach proposed by Berthold Horn in the 1970s is to decompose the input image in intrinsic images and use those to predict the depth of every input pixel.Since several configurations of shapes, texture and illumination can explain the same image, both approaches need to form assumptions on the distribution of images and 3D shapes to resolve the ambiguity. In this thesis, we learn these assumptions from large-scale datasets instead of manually designing them. Learning allows us to perform complete object reconstruction, including parts which are not visible in the input image.Shape matching aims at finding correspondences between 3D objects. Solving this task requires both a local and global understanding of 3D shapes which is hard to achieve explicitly. Instead we train neural networks on large-scale datasets to solve this task and capture this knowledge implicitly through their internal parameters.Shape matching supports many 3D modeling applications such as attribute transfer, automatic rigging for animation, or mesh editing.The first technical contribution of this thesis is a new parametric representation of 3D surfaces modeled by neural networks.The choice of data representation is a critical aspect of any 3D reconstruction algorithm. Until recently, most of the approaches in deep 3D model generation were predicting volumetric voxel grids or point clouds, which are discrete representations. Instead, we present an alternative approach that predicts a parametric surface deformation ie a mapping from a template to a target geometry. To demonstrate the benefits of such a representation, we train a deep encoder-decoder for single-view reconstruction using our new representation. Our approach, dubbed AtlasNet, is the first deep single-view reconstruction approach able to reconstruct meshes from images without relying on an independent post-processing, and can do it at arbitrary resolution without memory issues. A more detailed analysis of AtlasNet reveals it also generalizes better to categories it has not been trained on than other deep 3D generation approaches.Our second main contribution is a novel shape matching approach purely based on reconstruction via deformations. We show that the quality of the shape reconstructions is critical to obtain good correspondences, and therefore introduce a test-time optimization scheme to refine the learned deformations. For humans and other deformable shape categories deviating by a near-isometry, our approach can leverage a shape template and isometric regularization of the surface deformations. As category exhibiting non-isometric variations, such as chairs, do not have a clear template, we learn how to deform any shape into any other and leverage cycle-consistency constraints to learn meaningful correspondences. Our reconstruction-for-matching strategy operates directly on point clouds, is robust to many types of perturbations, and outperforms the state of the art by 15% on dense matching of real human scans
Asselin, Louis-Philippe. "Une approche d'apprentissage profond pour l’estimation de l'apparence des matériaux à partir d’images." Master's thesis, Université Laval, 2021. http://hdl.handle.net/20.500.11794/69186.
Full textCohen-Hadria, Alice. "Estimation de descriptions musicales et sonores par apprentissage profond." Electronic Thesis or Diss., Sorbonne université, 2019. http://www.theses.fr/2019SORUS607.
Full textIn Music Information Retrieval (MIR) and voice processing, the use of machine learning tools has become in the last few years more and more standard. Especially, many state-of-the-art systems now rely on the use of Neural Networks.In this thesis, we propose a wide overview of four different MIR and voice processing tasks, using systems built with neural networks. More precisely, we will use convolutional neural networks, an image designed class neural networks. The first task presented is music structure estimation. For this task, we will show how the choice of input representation can be critical, when using convolutional neural networks. The second task is singing voice detection. We will present how to use a voice detection system to automatically align lyrics and audio tracks.With this alignment mechanism, we have created the largest synchronized audio and speech data set, called DALI. Singing voice separation is the third task. For this task, we will present a data augmentation strategy, a way to significantly increase the size of a training set. Finally, we tackle voice anonymization. We will present an anonymization method that both obfuscate content and mask the speaker identity, while preserving the acoustic scene
Vialatte, Jean-Charles. "Convolution et apprentissage profond sur graphes." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2018. http://www.theses.fr/2018IMTA0118/document.
Full textConvolutional neural networks have proven to be the deep learning model that performs best on regularly structured datasets like images or sounds. However, they cannot be applied on datasets with an irregular structure (e.g. sensor networks, citation networks, MRIs). In this thesis, we develop an algebraic theory of convolutions on irregular domains. We construct a family of convolutions that are based on group actions (or, more generally, groupoid actions) that acts on the vertex domain and that have properties that depend on the edges. With the help of these convolutions, we propose extensions of convolutional neural netowrks to graph domains. Our researches lead us to propose a generic formulation of the propagation between layers, that we call the neural contraction. From this formulation, we derive many novel neural network models that can be applied on irregular domains. Through benchmarks and experiments, we show that they attain state-of-the-art performances, and beat them in some cases
Langlois, Julien. "Vision industrielle et réseaux de neurones profonds : application au dévracage de pièces plastiques industrielles." Thesis, Nantes, 2019. http://www.theses.fr/2019NANT4010/document.
Full textThis work presents a pose estimation method from a RGB image of industrial parts placed in a bin. In a first time, neural networks are used to segment a certain number of parts in the scene. After applying an object mask to the original image, a second network is inferring the local depth of the part. Both the local pixel coordinates of the part and the local depth are used in two networks estimating the orientation of the object as a quaternion and its translation on the Z axis. Finally, a registration module working on the back-projected local depth and the 3D model of the part is refining the pose inferred from the previous networks. To deal with the lack of annotated real images in an industrial context, an data generation process is proposed. By using various light parameters, the dataset versatility allows to anticipate multiple challenging exploitation scenarios within an industrial environment
Hardy, Corentin. "Contribution au développement de l’apprentissage profond dans les systèmes distribués." Thesis, Rennes 1, 2019. http://www.theses.fr/2019REN1S020/document.
Full textDeep learning enables the development of a growing number of services. However, it requires large training databases and a lot of computing power. In order to reduce the costs of this deep learning, we propose a distributed computing setup to enable collaborative learning. Future users can participate with their devices and their data without moving private data in datacenters. We propose methods to train deep neural network in this distibuted system context
Voerman, Joris. "Classification automatique à partir d’un flux de documents." Electronic Thesis or Diss., La Rochelle, 2022. http://www.theses.fr/2022LAROS025.
Full textAdministrative documents can be found everywhere today. They are numerous, diverse and can be of two types: physical and numerical. The need to switch between these two forms required the development of new solutions. After document digitization (mainly with a scanner), one of the first problems is to determine the type of the document, which will simplify all future processes. Automatic classification is a complex process that has multiple solutions in the state of the art. Therefore, the document classification, the imbalanced context and industrial constraints will heavily challenge these solutions. This thesis focuses on the automatic classification of document streams with research of solutions to the three major problems previously introduced. To this end, we first propose an evaluation of existing methods adaptation to document streams context. In addition, this work proposes an evaluation of state-of-the-art solutions to contextual constraints and possible combinations between them. Finally, we propose a new combination method that uses a cascade of systems to offer a gradual solution. The most effective solutions are, at first, a multimodal neural network reinforced by an attention model that is able to classify a great variety of documents. In second, a cascade of three complementary networks with : a one network for text classification, one for image classification and one for low represented classes. These two options provide good results as well in ideal context than in imbalanced context. In the first case, it challenges the state of the art. In the second case, it shows an improvement of +6% F0.5-Measure in comparison to the state of the art
Wolinski, Pierre. "Structural Learning of Neural Networks." Thesis, université Paris-Saclay, 2020. http://www.theses.fr/2020UPASS026.
Full textThe structure of a neural network determines to a large extent its cost of training and use, as well as its ability to learn. These two aspects are usually in competition: the larger a neural network is, the better it will perform the task assigned to it, but the more it will require memory and computing time resources for training. Automating the search of efficient network structures -of reasonable size and performing well- is then a very studied question in this area. Within this context, neural networks with various structures are trained, which requires a new set of training hyperparameters for each new structure tested. The aim of the thesis is to address different aspects of this problem. The first contribution is a training method that operates within a large perimeter of network structures and tasks, without needing to adjust the learning rate. The second contribution is a network training and pruning technique, designed to be insensitive to the initial width of the network. The last contribution is mainly a theorem that makes possible to translate an empirical training penalty into a Bayesian prior, theoretically well founded. This work results from a search for properties that theoretically must be verified by training and pruning algorithms to be valid over a wide range of neural networks and objectives
Etienne, Caroline. "Apprentissage profond appliqué à la reconnaissance des émotions dans la voix." Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLS517.
Full textThis thesis deals with the application of artificial intelligence to the automatic classification of audio sequences according to the emotional state of the customer during a commercial phone call. The goal is to improve on existing data preprocessing and machine learning models, and to suggest a model that is as efficient as possible on the reference IEMOCAP audio dataset. We draw from previous work on deep neural networks for automatic speech recognition, and extend it to the speech emotion recognition task. We are therefore interested in End-to-End neural architectures to perform the classification task including an autonomous extraction of acoustic features from the audio signal. Traditionally, the audio signal is preprocessed using paralinguistic features, as part of an expert approach. We choose a naive approach for data preprocessing that does not rely on specialized paralinguistic knowledge, and compare it with the expert approach. In this approach, the raw audio signal is transformed into a time-frequency spectrogram by using a short-term Fourier transform. In order to apply a neural network to a prediction task, a number of aspects need to be considered. On the one hand, the best possible hyperparameters must be identified. On the other hand, biases present in the database should be minimized (non-discrimination), for example by adding data and taking into account the characteristics of the chosen dataset. We study these aspects in order to develop an End-to-End neural architecture that combines convolutional layers specialized in the modeling of visual information with recurrent layers specialized in the modeling of temporal information. We propose a deep supervised learning model, competitive with the current state-of-the-art when trained on the IEMOCAP dataset, justifying its use for the rest of the experiments. This classification model consists of a four-layer convolutional neural networks and a bidirectional long short-term memory recurrent neural network (BLSTM). Our model is evaluated on two English audio databases proposed by the scientific community: IEMOCAP and MSP-IMPROV. A first contribution is to show that, with a deep neural network, we obtain high performances on IEMOCAP, and that the results are promising on MSP-IMPROV. Another contribution of this thesis is a comparative study of the output values of the layers of the convolutional module and the recurrent module according to the data preprocessing method used: spectrograms (naive approach) or paralinguistic indices (expert approach). We analyze the data according to their emotion class using the Euclidean distance, a deterministic proximity measure. We try to understand the characteristics of the emotional information extracted autonomously by the network. The idea is to contribute to research focused on the understanding of deep neural networks used in speech emotion recognition and to bring more transparency and explainability to these systems, whose decision-making mechanism is still largely misunderstood
Carbajal, Guillaume. "Apprentissage profond bout-en-bout pour le rehaussement de la parole." Electronic Thesis or Diss., Université de Lorraine, 2020. http://www.theses.fr/2020LORR0017.
Full textThis PhD falls within the development of hands-free telecommunication systems, more specifically smart speakers in domestic environments. The user interacts with another speaker at a far-end point and can be typically a few meters away from this kind of system. The microphones are likely to capture sounds of the environment which are added to the user's voice, such background noise, acoustic echo and reverberation. These types of distortion degrade speech quality, intelligibility and listening comfort for the far-end speaker, and must be reduced. Filtering methods can reduce individually each of these types of distortion. Reducing all of them implies combining the corresponding filtering methods. As these methods interact with each other which can deteriorate the user's speech, they must be jointly optimized. First of all, we introduce an acoustic echo reduction approach which combines an echo cancellation filter with a residual echo postfilter designed to adapt to the echo cancellation filter. To do so, we propose to estimate the postfilter coefficients using the short term spectra of multiple known signals, including the output of the echo cancellation filter, as inputs to a neural network. We show that this approach improves the performance and the robustness of the postfilter in terms of echo reduction, while limiting speech degradation, on several scenarios in real conditions. Secondly, we describe a joint approach for multichannel reduction of echo, reverberation and noise. We propose to simultaneously model the target speech and undesired residual signals after echo cancellation and dereveberation in a probabilistic framework, and to jointly represent their short-term spectra by means of a recurrent neural network. We develop a block-coordinate ascent algorithm to update the echo cancellation and dereverberation filters, as well as the postfilter that reduces the undesired residual signals. We evaluate our approach on real recordings in different conditions. We show that it improves speech quality and reduction of echo, reverberation and noise compared to a cascade of individual filtering methods and another joint reduction approach. Finally, we present an online version of our approach which is suitable for time-varying acoustic conditions. We evaluate the perceptual quality achieved on real examples where the user moves during the conversation
Douillard, Arthur. "Continual Learning for Computer Vision." Electronic Thesis or Diss., Sorbonne université, 2022. http://www.theses.fr/2022SORUS165.
Full textI first review the existing methods based on regularization for continual learning. While regularizing a model's probabilities is very efficient to reduce forgetting in large-scale datasets, there are few works considering constraints on intermediate features. I cover in this chapter two contributions aiming to regularize directly the latent space of ConvNet. The first one, PODNet, aims to reduce the drift of spatial statistics between the old and new model, which in effect reduces drastically forgetting of old classes while enabling efficient learning of new classes. I show in a second part a complementary method where we avoid pre-emptively forgetting by allocating locations in the latent space for yet unseen future class. Then, I describe a recent application of CIL to semantic segmentation. I show that the very nature of CSS offer new specific challenges, namely forgetting on large images and a background shift. We tackle the first problem by extending our distillation loss introduced in the previous chapter to multi-scales. The second problem is solved by an efficient pseudo-labeling strategy. Finally, we consider the common rehearsal learning, but applied this time to CSS. I show that it cannot be used naively because of memory complexity and design a light-weight rehearsal that is even more efficient. Finally, I consider a completely different approach to continual learning: dynamic networks where the parameters are extended during training to adapt to new tasks. Previous works on this domain are hard to train and often suffer from parameter count explosion. For the first time in continual computer vision, we propose to use the Transformer architecture: the model dimension mostly fixed and shared across tasks, except for an expansion of learned task tokens. With an encoder/decoder strategy where the decoder forward is specialized by a task token, we show state-of-the-art robustness to forgetting while our memory and computational complexities barely grow
Biasutto-Lervat, Théo. "Modélisation de la coarticulation multimodale : vers l'animation d'une tête parlante intelligible." Electronic Thesis or Diss., Université de Lorraine, 2021. http://www.theses.fr/2021LORR0019.
Full textThis thesis deals with neural network based coarticulation modeling, and aims to synchronize facial animation of a 3D talking head with speech. Predicting articulatory movements is not a trivial task, as it is well known that production of a phoneme is greatly affected by its phonetic context, a phoneme called coarticulation. We propose in this work a coarticulation model, i.e. a model able to predict spatial trajectories of articulators from speech. We rely on a sequential model, the recurrent neural networks, and more specifically the Gated Recurrent Units, which are able to consider the articulation dynamic as a central component of its modeling. Unfortunately, the typical amount of data in articulatory and audiovisual databases seems to be quite low for a deep learning approach. To overcome this difficulty, we propose to integrate articulatory knowledge into the networks during its initialization. The RNNs robustness allow uw to apply our coarticulation model to predict both face and tongue movements, in french and german for the face, and in english and german for the tongue. Evaluation has been conducted through objective measures of the trajectories, and through experiments to ensure a complete reach of critical articulatory targets. We also conducted a subjective evaluation to attest the perceptual quality of the predicted articulation once applied to our facial animation system. Finally, we analyzed the model after training to explore phonetic knowledges learned
Bertrand, Hadrien. "Optimisation d'hyper-paramètres en apprentissage profond et apprentissage par transfert : applications en imagerie médicale." Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLT001/document.
Full textIn the last few years, deep learning has changed irrevocably the field of computer vision. Faster, giving better results, and requiring a lower degree of expertise to use than traditional computer vision methods, deep learning has become ubiquitous in every imaging application. This includes medical imaging applications. At the beginning of this thesis, there was still a strong lack of tools and understanding of how to build efficient neural networks for specific tasks. Thus this thesis first focused on the topic of hyper-parameter optimization for deep neural networks, i.e. methods for automatically finding efficient neural networks on specific tasks. The thesis includes a comparison of different methods, a performance improvement of one of these methods, Bayesian optimization, and the proposal of a new method of hyper-parameter optimization by combining two existing methods: Bayesian optimization and Hyperband.From there, we used these methods for medical imaging applications such as the classification of field-of-view in MRI, and the segmentation of the kidney in 3D ultrasound images across two populations of patients. This last task required the development of a new transfer learning method based on the modification of the source network by adding new geometric and intensity transformation layers.Finally this thesis loops back to older computer vision methods, and we propose a new segmentation algorithm combining template deformation and deep learning. We show how to use a neural network to predict global and local transformations without requiring the ground-truth of these transformations. The method is validated on the task of kidney segmentation in 3D US images
Tran, Khanh-Hung. "Semi-supervised dictionary learning and Semi-supervised deep neural network." Thesis, université Paris-Saclay, 2021. http://www.theses.fr/2021UPASP014.
Full textSince the 2010's, machine learning (ML) has been one of the topics that attract a lot of attention from scientific researchers. Many ML models have been demonstrated their ability to produce excellent results in various fields such as Computer Vision, Natural Language Processing, Robotics... However, most of these models use supervised learning, which requires a massive annotation. Therefore, the objective of this thesis is to study and to propose semi-supervised learning approaches that have many advantages over supervised learning. Instead of directly applying a semi-supervised classifier on the original representation of data, we rather use models that integrate a representation learning stage before the classification stage, to better adapt to the non-linearity of the data. In the first step, we revisit tools that allow us to build our semi-supervised models. First, we present two types of model that possess representation learning in their architecture: dictionary learning and neural network, as well as the optimization methods for each type of model. Moreover, in the case of neural network, we specify the problem with adversarial examples. Then, we present the techniques that often accompany with semi-supervised learning such as variety learning and pseudo-labeling. In the second part, we work on dictionary learning. We synthesize generally three steps to build a semi-supervised model from a supervised model. Then, we propose our semi-supervised model to deal with the classification problem typically in the case of a low number of training samples (including both labelled and non-labelled samples). On the one hand, we apply the preservation of the data structure from the original space to the sparse code space (manifold learning), which is considered as regularization for sparse codes. On the other hand, we integrate a semi-supervised classifier in the sparse code space. In addition, we perform sparse coding for test samples by taking into account also the preservation of the data structure. This method provides an improvement on the accuracy rate compared to other existing methods. In the third step, we work on neural network models. We propose an approach called "manifold attack" which allows reinforcing manifold learning. This approach is inspired from adversarial learning : finding virtual points that disrupt the cost function on manifold learning (by maximizing it) while fixing the model parameters; then the model parameters are updated by minimizing this cost function while fixing these virtual points. We also provide criteria for limiting the space to which the virtual points belong and the method for initializing them. This approach provides not only an improvement on the accuracy rate but also a significant robustness to adversarial examples. Finally, we analyze the similarities and differences, as well as the advantages and disadvantages between dictionary learning and neural network models. We propose some perspectives on both two types of models. In the case of semi-supervised dictionary learning, we propose some techniques inspired by the neural network models. As for the neural network, we propose to integrate manifold attack on generative models
Elbayad, Maha. "Une alternative aux modèles neuronaux séquence-à-séquence pour la traduction automatique." Thesis, Université Grenoble Alpes, 2020. http://www.theses.fr/2020GRALM012.
Full textIn recent years, deep learning has enabled impressive achievements in Machine Translation.Neural Machine Translation (NMT) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another.One crucial factor to the success of NMT is the design of new powerful and efficient architectures. State-of-the-art systems are encoder-decoder models that first encode a source sequence into a set of feature vectors and then decode the target sequence conditioning on the source features.In this thesis we question the encoder-decoder paradigm and advocate for an intertwined encoding of the source and target so that the two sequences interact at increasing levels of abstraction. For this purpose, we introduce Pervasive Attention, a model based on two-dimensional convolutions that jointly encode the source and target sequences with interactions that are pervasive throughout the network.To improve the efficiency of NMT systems, we explore online machine translation where the source is read incrementally and the decoder is fed partial contexts so that the model can alternate between reading and writing. We investigate deterministic agents that guide the read/write alternation through a rigid decoding path, and introduce new dynamic agents to estimate a decoding path for each sample.We also address the resource-efficiency of encoder-decoder models and posit that going deeper in a neural network is not required for all instances.We design depth-adaptive Transformer decoders that allow for anytime prediction and sample-adaptive halting mechanisms to favor low cost predictions for low complexity instances and save deeper predictions for complex scenarios
Alqasir, Hiba. "Apprentissage profond pour l'analyse de scènes de remontées mécaniques : amélioration de la généralisation dans un contexte multi-domaines." Thesis, Lyon, 2020. http://www.theses.fr/2020LYSES045.
Full textThis thesis presents our work on chairlift safety using deep learning techniques as part of the Mivao project, which aims to develop a computer vision system that acquires images of the chairlift boarding station, analyzes the crucial elements, and detects dangerous situations. In this scenario, we have different chairlifts spread over different ski resorts, with a high diversity of acquisition conditions and geometries; thus, each chairlift is considered a domain. When the system is installed for a new chairlift, the objective is to perform an accurate and reliable scene analysis, given the lack of labeled data on this new domain (chairlift).In this context, we mainly concentrate on the chairlift safety bar and propose to classify each image into two categories, depending on whether the safety bar is closed (safe) or open (unsafe). Thus, it is an image classification problem with three specific features: (i) the image category depends on a small detail (the safety bar) in a cluttered background, (ii) manual annotations are not easy to obtain, (iii) a classifier trained on some chairlifts should provide good results on a new one (generalization). To guide the classifier towards the important regions of the images, we have proposed two solutions: object detection and Siamese networks. Furthermore, we analyzed the generalization property of these two approaches. Our solutions are motivated by the need to minimize human annotation efforts while improving the accuracy of the chairlift safety problem. However, these contributions are not necessarily limited to this specific application context, and they may be applied to other problems in a multi-domain context
Mlynarski, Pawel. "Apprentissage profond pour la segmentation des tumeurs cérébrales et des organes à risque en radiothérapie." Thesis, Université Côte d'Azur (ComUE), 2019. http://www.theses.fr/2019AZUR4084.
Full textMedical images play an important role in cancer diagnosis and treatment. Oncologists analyze images to determine the different characteristics of the cancer, to plan the therapy and to observe the evolution of the disease. The objective of this thesis is to propose efficient methods for automatic segmentation of brain tumors and organs at risk in the context of radiotherapy planning, using Magnetic Resonance (MR) images. First, we focus on segmentation of brain tumors using Convolutional Neural Networks (CNN) trained on MRIs manually segmented by experts. We propose a segmentation model having a large 3D receptive field while being efficient in terms of computational complexity, based on combination of 2D and 3D CNNs. We also address problems related to the joint use of several MRI sequences (T1, T2, FLAIR). Second, we introduce a segmentation model which is trained using weakly-annotated images in addition to fully-annotated images (with voxelwise labels), which are usually available in very limited quantities due to their cost. We show that this mixed level of supervision considerably improves the segmentation accuracy when the number of fully-annotated images is limited.\\ Finally, we propose a methodology for an anatomy-consistent segmentation of organs at risk in the context of radiotherapy of brain tumors. The segmentations produced by our system on a set of MRIs acquired in the Centre Antoine Lacassagne (Nice, France) are evaluated by an experienced radiotherapist
Ferré, Paul. "Adéquation algorithme-architecture de réseaux de neurones à spikes pour les architectures matérielles massivement parallèles." Thesis, Toulouse 3, 2018. http://www.theses.fr/2018TOU30318/document.
Full textThe last decade has seen the re-emergence of machine learning methods based on formal neural networks under the name of deep learning. Although these methods have enabled a major breakthrough in machine learning, several obstacles to the possibility of industrializing these methods persist, notably the need to collect and label a very large amount of data as well as the computing power necessary to perform learning and inference with this type of neural network. In this thesis, we propose to study the adequacy between inference and learning algorithms derived from biological neural networks and massively parallel hardware architectures. We show with three contribution that such adequacy drastically accelerates computation times inherent to neural networks. In our first axis, we study the adequacy of the BCVision software engine developed by Brainchip SAS for GPU platforms. We also propose the introduction of a coarse-to-fine architecture based on complex cells. We show that GPU portage accelerates processing by a factor of seven, while the coarse-to-fine architecture reaches a factor of one thousand. The second contribution presents three algorithms for spike propagation adapted to parallel architectures. We study exhaustively the computational models of these algorithms, allowing the selection or design of the hardware system adapted to the parameters of the desired network. In our third axis we present a method to apply the Spike-Timing-Dependent-Plasticity rule to image data in order to learn visual representations in an unsupervised manner. We show that our approach allows the effective learning a hierarchy of representations relevant to image classification issues, while requiring ten times less data than other approaches in the literature
Feutry, Clément. "Two sides of relevant information : anonymized representation through deep learning and predictor monitoring." Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLS479.
Full textThe work presented here is for a first part at the cross section of deep learning and anonymization. A full framework was developed in order to identify and remove to a certain extant, in an automated manner, the features linked to an identity in the context of image data. Two different kinds of processing data were explored. They both share the same Y-shaped network architecture despite components of this network varying according to the final purpose. The first one was about building from the ground an anonymized representation that allowed a trade-off between keeping relevant features and tampering private features. This framework has led to a new loss. The second kind of data processing specified no relevant information about the data, only private information, meaning that everything that was not related to private features is assumed relevant. Therefore the anonymized representation shares the same nature as the initial data (e.g. an image is transformed into an anonymized image). This task led to another type of architecture (still in a Y-shape) and provided results strongly dependent on the type of data. The second part of the work is relative to another kind of relevant information: it focuses on the monitoring of predictor behavior. In the context of black box analysis, we only have access to the probabilities outputted by the predictor (without any knowledge of the type of structure/architecture producing these probabilities). This monitoring is done in order to detect abnormal behavior that is an indicator of a potential mismatch between the data statistics and the model statistics. Two methods are presented using different tools. The first one is based on comparing the empirical cumulative distribution of known data and to be tested data. The second one introduces two tools: one relying on the classifier uncertainty and the other relying on the confusion matrix. These methods produce concluding results
Firmo, Drumond Thalita. "Apports croisées de l'apprentissage hiérarchique et la modélisation du système visuel : catégorisation d'images sur des petits corpus de données." Thesis, Bordeaux, 2020. https://tel.archives-ouvertes.fr/tel-03129189.
Full textDeep convolutional neural networks (DCNN) have recently protagonized a revolution in large-scale object recognition. They have changed the usual computer vision practices of hand-engineered features, with their ability to hierarchically learn representative features from data with a pertinent classifier. Together with hardware advances, they have made it possible to effectively exploit the ever-growing amounts of image data gathered online. However, in specific domains like healthcare and industrial applications, data is much less abundant, and expert labeling costs higher than those of general purpose image datasets. This scarcity scenario leads to this thesis' core question: can these limited-data domains profit from the advantages of DCNNs for image classification? This question has been addressed throughout this work, based on an extensive study of literature, divided in two main parts, followed by proposal of original models and mechanisms.The first part reviews object recognition from an interdisciplinary double-viewpoint. First, it resorts to understanding the function of vision from a biological stance, comparing and contrasting to DCNN models in terms of structure, function and capabilities. Second, a state-of-the-art review is established aiming to identify the main architectural categories and innovations in modern day DCNNs. This interdisciplinary basis fosters the identification of potential mechanisms - inspired both from biological and artificial structures — that could improve image recognition under difficult situations. Recurrent processing is a clear example: while not completely absent from the "deep vision" literature, it has mostly been applied to videos — due to their inherently sequential nature. From biology however it is clear such processing plays a role in refining our perception of a still scene. This theme is further explored through a dedicated literature review focused on recurrent convolutional architectures used in image classification.The second part carries on in the spirit of improving DCNNs, this time focusing more specifically on our central question: deep learning over small datasets. First, the work proposes a more detailed and precise discussion of the small sample problem and its relation to learning hierarchical features with deep models. This discussion is followed up by a structured view of the field, organizing and discussing the different possible paths towards adapting deep models to limited data settings. Rather than a raw listing, this review work aims to make sense out of the myriad of approaches in the field, grouping methods with similar intent or mechanism of action, in order to guide the development of custom solutions for small-data applications. Second, this study is complemented by an experimental analysis, exploring small data learning with the proposition of original models and mechanisms (previously published as a journal paper).In conclusion, it is possible to apply deep learning to small datasets and obtain good results, if done in a thoughtful fashion. On the data path, one shall try gather more information from additional related data sources if available. On the complexity path, architecture and training methods can be calibrated in order to profit the most from any available domain-specific side-information. Proposals concerning both of these paths get discussed in detail throughout this document. Overall, while there are multiple ways of reducing the complexity of deep learning with small data samples, there is no universal solution. Each method has its own drawbacks and practical difficulties and needs to be tailored specifically to the target perceptual task at hand
Moukari, Michel. "Estimation de profondeur à partir d'images monoculaires par apprentissage profond." Thesis, Normandie, 2019. http://www.theses.fr/2019NORMC211/document.
Full textComputer vision is a branch of artificial intelligence whose purpose is to enable a machine to analyze, process and understand the content of digital images. Scene understanding in particular is a major issue in computer vision. It goes through a semantic and structural characterization of the image, on one hand to describe its content and, on the other hand, to understand its geometry. However, while the real space is three-dimensional, the image representing it is two-dimensional. Part of the 3D information is thus lost during the process of image formation and it is therefore non trivial to describe the geometry of a scene from 2D images of it.There are several ways to retrieve the depth information lost in the image. In this thesis we are interested in estimating a depth map given a single image of the scene. In this case, the depth information corresponds, for each pixel, to the distance between the camera and the object represented in this pixel. The automatic estimation of a distance map of the scene from an image is indeed a critical algorithmic brick in a very large number of domains, in particular that of autonomous vehicles (obstacle detection, navigation aids).Although the problem of estimating depth from a single image is a difficult and inherently ill-posed problem, we know that humans can appreciate distances with one eye. This capacity is not innate but acquired and made possible mostly thanks to the identification of indices reflecting the prior knowledge of the surrounding objects. Moreover, we know that learning algorithms can extract these clues directly from images. We are particularly interested in statistical learning methods based on deep neural networks that have recently led to major breakthroughs in many fields and we are studying the case of the monocular depth estimation
Medrouk, Indira Lisa. "Réseaux profonds pour la classification des opinions multilingue." Electronic Thesis or Diss., Paris 8, 2018. http://www.theses.fr/2018PA080081.
Full textIn the era of social networks where everyone can claim to be a contentproducer, the growing interest in research and industry is an indisputablefact for the opinion mining domain.This thesis is mainly addressing a Web inherent characteristic reflectingits globalized and multilingual character.To address the multilingual opinion mining issue, the proposed model isinspired by the process of acquiring simultaneous languages with equal intensityamong young children. The incorporate corpus-based input is raw, usedwithout any pre-processing, translation, annotation nor additional knowledgefeatures. For the machine learning approach, we use two different deep neuralnetworks. The evaluation of the proposed model was executed on corpusescomposed of four different languages, namely French, English, Greek and Arabic,to emphasize the ability of a deep learning model in order to establishthe sentiment polarity of reviews and topics classification in a multilingualenvironment. The various experiments combining corpus size variations forbi and quadrilingual grouping languages, presented to our models withoutadditional modules, have shown that, such as children bilingual competencedevelopment, which is linked to quality and quantity of their immersion in thelinguistic context, the network learns better in a rich and varied environment.As part of the problem of opinion classification, the second part of thethesis presents a comparative study of two models of deep networks : convolutionalnetworks and recurrent networks. Our contribution consists in demonstratingtheir complementarity according to their combinations in a multilingualcontext
Paillassa, Maxime. "Détection robuste de sources astronomiques par réseaux de neurones à convolutions." Thesis, Bordeaux, 2020. http://www.theses.fr/2020BORD0147.
Full textExtracting reliable source catalogs from images is crucial for a broad range of astronomical research topics.However, the efficiency of current source detection methods becomes severely limited in crowded fields, or when images are contaminated by optical, electronic and environmental defects.Performance in terms of reliability and completeness is now often insufficient with regard to the scientific requirements of large imaging surveys.In this thesis, we develop new methods to produce more robust and reliable source catalogs.We leverage recent advances in deep supervised learning to design generic and reliable models based on convolutional neural networks (CNNs).We present MaxiMask and MaxiTrack, two convolutional neural networks that we trained to automatically identify 13 different types of image defects in astronomical exposures.We also introduce a prototype of a multi-scale CNN-based source detector robust to image defects, which we show to significantly outperform existing algorithms.We discuss the current limitations and potential improvements of our approach in the scope of forthcoming large scale surveys such as Euclid
Wilson, Dennis G. "Évolution des principes de la conception des réseaux de neurones artificiels." Thesis, Toulouse 3, 2019. http://www.theses.fr/2019TOU30075.
Full textThe biological brain is an ensemble of individual components which have evolved over millions of years. Neurons and other cells interact in a complex network from which intelligence emerges. Many of the neural designs found in the biological brain have been used in computational models to power artificial intelligence, with modern deep neural networks spurring a revolution in computer vision, machine translation, natural language processing, and many more domains. However, artificial neural networks are based on only a small subset of biological functionality of the brain, and often focus on global, homogeneous changes to a system that is complex and locally heterogeneous. In this work, we examine the biological brain, from single neurons to networks capable of learning. We examine individually the neural cell, the formation of connections between cells, and how a network learns over time. For each component, we use artificial evolution to find the principles of neural design that are optimized for artificial neural networks. We then propose a functional model of the brain which can be used to further study select components of the brain, with all functions designed for automatic optimization such as evolution. Our goal, ultimately, is to improve the performance of artificial neural networks through inspiration from modern neuroscience. However, through evaluating the biological brain in the context of an artificial agent, we hope to also provide models of the brain which can serve biologists
Godet, Pierre. "Approches par apprentissage pour l’estimation de mouvement multiframe en vidéo." Thesis, université Paris-Saclay, 2021. http://www.theses.fr/2021UPASG005.
Full textThis work concerns the use of temporal information on a sequence of more than two images for optical flow estimation. Optical flow is defined as the dense field (in any pixel) of the apparent movements in the image plane. We study on the one hand the use of a basis of temporal models, learned by principal component analysis from the studied data, to model the temporal dependence of the movement. This first study focuses on the context of particle image velocimetry in fluid mechanics. On the other hand, the new state of the art of optical flow estimation having recently been established by methods based on deep learning, we train convolutional neural networks to estimate optical flow by taking advantage of temporal continuity, in the case of natural image sequences. We then propose STaRFlow, a convolutional neural network exploiting a memory of information from the past by using a temporal recurrence. By repeated application of the same recurrent cell, the same learned parameters are used for the different time steps and for the different levels of a multiscale process. This architecture is lighter than competing networks while giving STaRFlow state-of-the-art performance. In the course of our work, we highlight several cases where the use of temporal information improves the quality of the estimation, in particular in the presence of occlusions, when the image quality is degraded (blur, noise), or in the case of thin objects
Antipov, Grigory. "Apprentissage profond pour la description sémantique des traits visuels humains." Thesis, Paris, ENST, 2017. http://www.theses.fr/2017ENST0071/document.
Full textThe recent progress in artificial neural networks (rebranded as deep learning) has significantly boosted the state-of-the-art in numerous domains of computer vision. In this PhD study, we explore how deep learning techniques can help in the analysis of gender and age from a human face. In particular, two complementary problem settings are considered: (1) gender/age prediction from given face images, and (2) synthesis and editing of human faces with the required gender/age attributes.Firstly, we conduct a comprehensive study which results in an empirical formulation of a set of principles for optimal design and training of gender recognition and age estimation Convolutional Neural Networks (CNNs). As a result, we obtain the state-of-the-art CNNs for gender/age prediction according to the three most popular benchmarks, and win an international competition on apparent age estimation. On a very challenging internal dataset, our best models reach 98.7% of gender classification accuracy and an average age estimation error of 4.26 years.In order to address the problem of synthesis and editing of human faces, we design and train GA-cGAN, the first Generative Adversarial Network (GAN) which can generate synthetic faces of high visual fidelity within required gender and age categories. Moreover, we propose a novel method which allows employing GA-cGAN for gender swapping and aging/rejuvenation without losing the original identity in synthetic faces. Finally, in order to show the practical interest of the designed face editing method, we apply it to improve the accuracy of an off-the-shelf face verification software in a cross-age evaluation scenario
Zimmer, Matthieu. "Apprentissage par renforcement développemental." Thesis, Université de Lorraine, 2018. http://www.theses.fr/2018LORR0008/document.
Full textReinforcement learning allows an agent to learn a behavior that has never been previously defined by humans. The agent discovers the environment and the different consequences of its actions through its interaction: it learns from its own experience, without having pre-established knowledge of the goals or effects of its actions. This thesis tackles how deep learning can help reinforcement learning to handle continuous spaces and environments with many degrees of freedom in order to solve problems closer to reality. Indeed, neural networks have a good scalability and representativeness. They make possible to approximate functions on continuous spaces and allow a developmental approach, because they require little a priori knowledge on the domain. We seek to reduce the amount of necessary interaction of the agent to achieve acceptable behavior. To do so, we proposed the Neural Fitted Actor-Critic framework that defines several data efficient actor-critic algorithms. We examine how the agent can fully exploit the transitions generated by previous behaviors by integrating off-policy data into the proposed framework. Finally, we study how the agent can learn faster by taking advantage of the development of his body, in particular, by proceeding with a gradual increase in the dimensionality of its sensorimotor space
Jacques, Céline. "Méthodes d'apprentissage automatique pour la transcription automatique de la batterie." Electronic Thesis or Diss., Sorbonne université, 2019. http://www.theses.fr/2019SORUS150.
Full textThis thesis focuses on learning methods for automatic transcription of the battery. They are based on a transcription algorithm using a non-negative decomposition method, NMD. This thesis raises two main issues: the adaptation of methods to the analyzed signal and the use of deep learning. Taking into account the information of the signal analyzed in the model can be achieved by their introduction during the decomposition steps. A first approach is to reformulate the decomposition step in a probabilistic context to facilitate the introduction of a posteriori information with methods such as SI-PLCA and statistical NMD. A second approach is to implement an adaptation strategy directly in the NMD: the application of modelable filters to the patterns to model the recording conditions or the adaptation of the learned patterns directly to the signal by applying strong constraints to preserve their physical meaning. The second approach concerns the selection of the signal segments to be analyzed. It is best to analyze segments where at least one percussive event occurs. An onset detector based on a convolutional neural network (CNN) is adapted to detect only percussive onsets. The results obtained being very interesting, the detector is trained to detect only one instrument allowing the transcription of the three main drum instruments with three CNNs. Finally, the use of a CNN multi-output is studied to transcribe the part of battery with a single network
Mollaret, Sébastien. "Artificial intelligence algorithms in quantitative finance." Thesis, Paris Est, 2021. http://www.theses.fr/2021PESC2002.
Full textArtificial intelligence has become more and more popular in quantitative finance given the increase of computer capacities as well as the complexity of models and has led to many financial applications. In the thesis, we have explored three different applications to solve financial derivatives challenges, from model selection, to model calibration and pricing. In Part I, we focus on a regime-switching model to price equity derivatives. The model parameters are estimated using the Expectation-Maximization (EM) algorithm and a local volatility component is added to fit vanilla option prices using the particle method. In Part II, we then use deep neural networks to calibrate a stochastic volatility model, where the volatility is modelled as the exponential of an Ornstein-Uhlenbeck process, by approximating the mapping between model parameters and corresponding implied volatilities offline. Once the expensive approximation has been performed offline, the calibration reduces to a standard & fast optimization problem.In Part III, we finally use deep neural networks to price American option on large baskets to solve the curse of the dimensionality. Different methods are studied with a Longstaff-Schwartz approach, where we approximate the continuation values, and a stochastic control approach, where we solve the pricing partial differential equation by reformulating the problem as a stochastic control problem using the non-linear Feynman-Kac formula
Martinez, Coralie. "Classification précoce de séquences temporelles par de l'apprentissage par renforcement profond." Thesis, Université Grenoble Alpes (ComUE), 2019. http://www.theses.fr/2019GREAT123.
Full textEarly classification (EC) of time series is a recent research topic in the field of sequential data analysis. It consists in assigning a label to some data that is sequentially collected with new data points arriving over time, and the prediction of a label has to be made using as few data points as possible in the sequence. The EC problem is of paramount importance for supporting decision makers in many real-world applications, ranging from process control to fraud detection. It is particularly interesting for applications concerned with the costs induced by the acquisition of data points, or for applications which seek for rapid label prediction in order to take early actions. This is for example the case in the field of health, where it is necessary to provide a medical diagnosis as soon as possible from the sequence of medical observations collected over time. Another example is predictive maintenance with the objective to anticipate the breakdown of a machine from its sensor signals. In this doctoral work, we developed a new approach for this problem, based on the formulation of a sequential decision making problem, that is the EC model has to decide between classifying an incomplete sequence or delaying the prediction to collect additional data points. Specifically, we described this problem as a Partially Observable Markov Decision Process noted EC-POMDP. The approach consists in training an EC agent with Deep Reinforcement Learning (DRL) in an environment characterized by the EC-POMDP. The main motivation for this approach was to offer an end-to-end model for EC which is able to simultaneously learn optimal patterns in the sequences for classification and optimal strategic decisions for the time of prediction. Also, the method allows to set the importance of time against accuracy of the classification in the definition of rewards, according to the application and its willingness to make this compromise. In order to solve the EC-POMDP and model the policy of the EC agent, we applied an existing DRL algorithm, the Double Deep-Q-Network algorithm, whose general principle is to update the policy of the agent during training episodes, using a replay memory of past experiences. We showed that the application of the original algorithm to the EC problem lead to imbalanced memory issues which can weaken the training of the agent. Consequently, to cope with those issues and offer a more robust training of the agent, we adapted the algorithm to the EC-POMDP specificities and we introduced strategies of memory management and episode management. In experiments, we showed that these contributions improved the performance of the agent over the original algorithm, and that we were able to train an EC agent which compromised between speed and accuracy, on each sequence individually. We were also able to train EC agents on public datasets for which we have no expertise, showing that the method is applicable to various domains. Finally, we proposed some strategies to interpret the decisions of the agent, validate or reject them. In experiments, we showed how these solutions can help gain insight in the choice of action made by the agent
Arnold, Ludovic. "Learning Deep Representations : Toward a better new understanding of the deep learning paradigm." Phd thesis, Université Paris Sud - Paris XI, 2013. http://tel.archives-ouvertes.fr/tel-00842447.
Full textChandra, Siddhartha. "Apprentissage Profond pour des Prédictions Structurées Efficaces appliqué à la Classification Dense en Vision par Ordinateur." Thesis, Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLC033/document.
Full textIn this thesis we propose a structured prediction technique that combines the virtues of Gaussian Conditional Random Fields (G-CRFs) with Convolutional Neural Networks (CNNs). The starting point of this thesis is the observation that while being of a limited form GCRFs allow us to perform exact Maximum-APosteriori (MAP) inference efficiently. We prefer exactness and simplicity over generality and advocate G-CRF based structured prediction in deep learning pipelines. Our proposed structured prediction methods accomodate (i) exact inference, (ii) both shortand long- term pairwise interactions, (iii) rich CNN-based expressions for the pairwise terms, and (iv) end-to-end training alongside CNNs. We devise novel implementation strategies which allow us to overcome memory and computational challenges
Carvalho, Micael. "Deep representation spaces." Electronic Thesis or Diss., Sorbonne université, 2018. http://www.theses.fr/2018SORUS292.
Full textIn recent years, Deep Learning techniques have swept the state-of-the-art of many applications of Machine Learning, becoming the new standard approach for them. The architectures issued from these techniques have been used for transfer learning, which extended the power of deep models to tasks that did not have enough data to fully train them from scratch. This thesis' subject of study is the representation spaces created by deep architectures. First, we study properties inherent to them, with particular interest in dimensionality redundancy and precision of their features. Our findings reveal a strong degree of robustness, pointing the path to simple and powerful compression schemes. Then, we focus on refining these representations. We choose to adopt a cross-modal multi-task problem, and design a loss function capable of taking advantage of data coming from multiple modalities, while also taking into account different tasks associated to the same dataset. In order to correctly balance these losses, we also we develop a new sampling scheme that only takes into account examples contributing to the learning phase, i.e. those having a positive loss. Finally, we test our approach in a large-scale dataset of cooking recipes and associated pictures. Our method achieves a 5-fold improvement over the state-of-the-art, and we show that the multi-task aspect of our approach promotes a semantically meaningful organization of the representation space, allowing it to perform subtasks never seen during training, like ingredient exclusion and selection. The results we present in this thesis open many possibilities, including feature compression for remote applications, robust multi-modal and multi-task learning, and feature space refinement. For the cooking application, in particular, many of our findings are directly applicable in a real-world context, especially for the detection of allergens, finding alternative recipes due to dietary restrictions, and menu planning
Ran, Peipei. "Imaging and diagnostic of sub-wavelength micro-structures, from closed-form algorithms to deep learning." Electronic Thesis or Diss., université Paris-Saclay, 2020. http://www.theses.fr/2020UPASG061.
Full textElectromagnetic probing of a gridlike, finite set of infinitely long circular cylindrical dielectric rods affected by missing ones is investigated from time-harmonic single and multiple frequency data. Sub-wavelength distances between adjacent rods and sub-wavelength rod diameters are assumed throughout the frequency band of operation and this leads to a severe challenge due to need of super-resolution within the present micro-structure, well beyond the Rayleigh criterion. A wealth of solution methods is investigated and comprehensive numerical simulations illustrate pros and cons, completed by processing laboratory-controlled experimental data acquired on a micro-structure prototype in a microwave anechoic chamber. These methods, which differ per a priori information accounted for and consequent versatility, include time-reversal, binary-specialized contrast-source and sparsity-constrained inversions, and convolutional neural networks possibly combined with recurrent ones
Diallo, Boubacar. "Mesure de l'intégrité d'une image : des modèles physiques aux modèles d'apprentissage profond." Thesis, Poitiers, 2020. http://www.theses.fr/2020POIT2293.
Full textDigital images have become a powerful and effective visual communication tool for delivering messages, diffusing ideas, and proving facts. The smartphone emergence with a wide variety of brands and models facilitates the creation of new visual content and its dissemination in social networks and image sharing platforms. Related to this phenomenon and helped by the availability and ease of use of image manipulation softwares, many issues have arisen ranging from the distribution of illegal content to copyright infringement. The reliability of digital images is questioned for common or expert users such as court or police investigators. A well known phenomenon and widespread examples are the "fake news" which oftenly include malicious use of digital images.Many researchers in the field of image forensic have taken up the scientific challenges associated with image manipulation. Many methods with interesting performances have been developed based on automatic image processing and more recently the adoption of deep learning. Despite the variety of techniques offered, performance are bound to specific conditions and remains vulnerable to relatively simple malicious attacks. Indeed, the images collected on the Internet impose many constraints on algorithms questioning many existing integrity verification techniques. There are two main peculiarities to be taken into account for the detection of a falsification: one is the lack of information on pristine image acquisition, the other is the high probability of automatic transformations linked to the image-sharing platforms such as lossy compression or resizing.In this thesis, we focus on several of these image forensic challenges including camera model identification and image tampering detection. After reviewing the state of the art in the field, we propose a first data-driven method for identifying camera models. We use deep learning techniques based on convolutional neural networks (CNNs) and develop a learning strategy considering the quality of the input data versus the applied transformation. A family of CNN networks has been designed to learn the characteristics of the camera model directly from a collection of images undergoing the same transformations as those commonly used on the Internet. Our interest focused on lossy compression for our experiments, because it is the most used type of post-processing on the Internet. The proposed approach, therefore, provides a robust solution to compression for camera model identification. The performance achieved by our camera model detection approach is also used and adapted for image tampering detection and localization. The performances obtained underline the robustness of our proposals for camera model identification and image forgery detection
Dvornik, Mikita. "Learning with Limited Annotated Data for Visual Understanding." Thesis, Université Grenoble Alpes (ComUE), 2019. http://www.theses.fr/2019GREAM050.
Full textThe ability of deep-learning methods to excel in computer vision highly depends on the amount of annotated data available for training. For some tasks, annotation may be too costly and labor intensive, thus becoming the main obstacle to better accuracy. Algorithms that learn from data automatically, without human supervision, perform substantially worse than their fully-supervised counterparts. Thus, there is a strong motivation to work on effective methods for learning with limited annotations. This thesis proposes to exploit prior knowledge about the task and develops more effective solutions for scene understanding and few-shot image classification.Main challenges of scene understanding include object detection, semantic and instance segmentation. Similarly, all these tasks aim at recognizing and localizing objects, at region- or more precise pixel-level, which makes the annotation process difficult. The first contribution of this manuscript is a Convolutional Neural Network (CNN) that performs both object detection and semantic segmentation. We design a specialized network architecture, that is trained to solve both problems in one forward pass, and operates in real-time. Thanks to the multi-task training procedure, both tasks benefit from each other in terms of accuracy, with no extra labeled data.The second contribution introduces a new technique for data augmentation, i.e., artificially increasing the amount of training data. It aims at creating new scenes by copy-pasting objects from one image to another, within a given dataset. Placing an object in a right context was found to be crucial in order to improve scene understanding performance. We propose to model visual context explicitly using a CNN that discovers correlations between object categories and their typical neighborhood, and then proposes realistic locations for augmentation. Overall, pasting objects in ``right'' locations allows to improve object detection and segmentation performance, with higher gains in limited annotation scenarios.For some problems, the data is extremely scarce, and an algorithm has to learn new concepts from a handful of examples. Few-shot classification consists of learning a predictive model that is able to effectively adapt to a new class, given only a few annotated samples. While most current methods concentrate on the adaptation mechanism, few works have tackled the problem of scarce training data explicitly. In our third contribution, we show that by addressing the fundamental high-variance issue of few-shot learning classifiers, it is possible to significantly outperform more sophisticated existing techniques. Our approach consists of designing an ensemble of deep networks to leverage the variance of the classifiers, and introducing new strategies to encourage the networks to cooperate, while encouraging prediction diversity. By matching different networks outputs on similar input images, we improve model accuracy and robustness, comparing to classical ensemble training. Moreover, a single network obtained by distillation shows similar to the full ensemble performance and yields state-of-the-art results with no computational overhead at test time
Millan, Mégane. "L'apprentissage profond pour l'évaluation et le retour d'information lors de l'apprentissage de gestes." Thesis, Sorbonne université, 2020. http://www.theses.fr/2020SORUS057.
Full textLearning a new sport or manual work is complex. Indeed, many gestures have to be assimilated in order to reach a good level of skill. However, learning these gestures cannot be done alone. Indeed, it is necessary to see the gesture execution with an expert eye in order to indicate corrections for improvement. However, experts, whether in sports or in manual works, are not always available to analyze and evaluate a novice’s gesture. In order to help experts in this task of analysis, it is possible to develop virtual coaches. Depending on the field, the virtual coach will have more or less skills, but an evaluation according to precise criteria is always mandatory. Providing feedback on mistakes is also essential for the learning of a novice. In this thesis, different solutions for developing the most effective virtual coaches are proposed. First of all, and as mentioned above, it is necessary to evaluate the gestures. From this point of view, a first part consisted in understanding the stakes of automatic gesture analysis, in order to develop an automatic evaluation algorithm that is as efficient as possible. Subsequently, two algorithms for automatic quality evaluation are proposed. These two algorithms, based on deep learning, were then tested on two different gestures databases in order to evaluate their genericity. Once the evaluation has been carried out, it is necessary to provide relevant feedback to the learner on his errors. In order to maintain continuity in the work carried out, this feedback is also based on neural networks and deep learning. A method has been developed based on neural network explanability methods. It allows to go back to the moments of the gestures when errors were made according to the evaluation model. Finally, coupled with semantic segmentation, this method makes it possible to indicate to learners which part of the gesture was badly performed, and to provide them with statistics and a learning curve
Sanabria, Rosas Laura Melissa. "Détection et caractérisation des moments saillants pour les résumés automatiques." Thesis, Université Côte d'Azur, 2021. http://www.theses.fr/2021COAZ4104.
Full textVideo content is present in an ever-increasing number of fields, both scientific and commercial. Sports, particularly soccer, is one of the industries that has invested the most in the field of video analytics, due to the massive popularity of the game. Although several state-of-the-art methods rely on handcrafted heuristics to generate summaries of soccer games, they have proven that multiple modalities help detect the best actions of the game. On the other hand, the field of general-purpose video summarization has advanced rapidly, offering several deep learning approaches. However, many of them are based on properties that are not feasible for sports videos. Video content has been for many years the main source for automatic tasks in soccer but the data that registers all the events happening on the field have become lately very important in sports analytics, since these event data provide richer information and requires less processing. Considering that in automatic sports summarization, the goal is not only to show the most important actions of the game, but also to evoke as much emotion as those evoked by human editors, we propose a method to generate the summary of a soccer match video exploiting the event metadata of the entire match and the content broadcast on TV. We have designed an architecture, introducing (1) a Multiple Instance Learning method that takes into account the sequential dependency among events, (2) a hierarchical multimodal attention layer that grasps the importance of each event in an action and (3) a method to automatically generate multiple summaries of a soccer match by sampling from a ranking distribution, providing multiple candidate summaries which are similar enough but with relevant variability to provide different options to the final user.We also introduced solutions to some additional challenges in the field of sports summarization. Based on the internal signals of an attention model that uses event data as input, we proposed a method to analyze the interpretability of our model through a graphical representation of actions where the x-axis of the graph represents the sequence of events, and the y-axis is the weight value learned by the attention layer. This new representation provides a new tool for the editor containing meaningful information to decide whether an action is important. We also proposed the use of keyword spotting and boosting techniques to detect every time a player is mentioned by the commentators as a solution for the missing event data
Le, Thien-Hoa. "Neural Methods for Sentiment Analysis and Text Summarization." Electronic Thesis or Diss., Université de Lorraine, 2020. http://www.theses.fr/2020LORR0037.
Full textThis thesis focuses on two Natural Language Processing tasks that require to extract semantic information from raw texts: Sentiment Analysis and Text Summarization. This dissertation discusses issues and seeks to improve neural models on both tasks, which have become the dominant paradigm in the past several years. Accordingly, this dissertation is composed of two parts: the first part (Neural Sentiment Analysis) deals with the computational study of people's opinions, sentiments, and the second part (Neural Text Summarization) tries to extract salient information from a complex sentence and rewrites it in a human-readable form. Neural Sentiment Analysis. Similar to computer vision, numerous deep convolutional neural networks have been adapted to sentiment analysis and text classification tasks. However, unlike the image domain, these studies are carried on different input data types and on different datasets, which makes it hard to know if a deep network is truly needed. In this thesis, we seek to find elements to address this question, i.e. whether neural networks must compute deep hierarchies of features for textual data in the same way as they do in vision. We thus propose a new adaptation of the deepest convolutional architecture (DenseNet) for text classification and study the importance of depth in convolutional models with different atom-levels (word or character) of input. We show that deep models indeed give better performances than shallow networks when the text input is represented as a sequence of characters. However, a simple shallow-and-wide network outperforms the deep DenseNet models with word inputs. Besides, to further improve sentiment classifiers and contextualize them, we propose to model them jointly with dialog acts, which are a factor of explanation and correlate with sentiments but are nevertheless often ignored. We have manually annotated both dialogues and sentiments on a Twitter-like social medium, and train a multi-task hierarchical recurrent network on joint sentiment and dialog act recognition. We show that transfer learning may be efficiently achieved between both tasks, and further analyze some specific correlations between sentiments and dialogues on social media. Neural Text Summarization. Detecting sentiments and opinions from large digital documents does not always enable users of such systems to take informed decisions, as other important semantic information is missing. People also need the main arguments and supporting reasons from the source documents to truly understand and interpret the document. To capture such information, we aim at making the neural text summarization models more explainable. We propose a model that has better explainability properties and is flexible enough to support various shallow syntactic parsing modules. More specifically, we linearize the syntactic tree into the form of overlapping text segments, which are then selected with reinforcement learning (RL) and regenerated into a compressed form. Hence, the proposed model is able to handle both extractive and abstractive summarization. Further, we observe that RL-based models are becoming increasingly ubiquitous for many text summarization tasks. We are interested in better understanding what types of information is taken into account by such models, and we propose to study this question from the syntactic perspective. We thus provide a detailed comparison of both RL-based and syntax-aware approaches and of their combination along several dimensions that relate to the perceived quality of the generated summaries such as number of repetitions, sentence length, distribution of part-of-speech tags, relevance and grammaticality. We show that when there is a resource constraint (computation and memory), it is wise to only train models with RL and without any syntactic information, as they provide nearly as good results as syntax-aware models with less parameters and faster training convergence
Belilovsky, Eugene. "Apprentissage de graphes structuré et parcimonieux dans des données de haute dimension avec applications à l’imagerie cérébrale." Thesis, Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLC027.
Full textThis dissertation presents novel structured sparse learning methods on graphs that address commonly found problems in the analysis of neuroimaging data as well as other high dimensional data with few samples. The first part of the thesis proposes convex relaxations of discrete and combinatorial penalties involving sparsity and bounded total variation on a graph as well as bounded `2 norm. These are developed with the aim of learning an interpretable predictive linear model and we demonstrate their effectiveness on neuroimaging data as well as a sparse image recovery problem.The subsequent parts of the thesis considers structure discovery of undirected graphical models from few observational data. In particular we focus on invoking sparsity and other structured assumptions in Gaussian Graphical Models (GGMs). To this end we make two contributions. We show an approach to identify differences in Gaussian Graphical Models (GGMs) known to have similar structure. We derive the distribution of parameter differences under a joint penalty when parameters are known to be sparse in the difference. We then show how this approach can be used to obtain confidence intervals on edge differences in GGMs. We then introduce a novel learning based approach to the problem structure discovery of undirected graphical models from observational data. We demonstrate how neural networks can be used to learn effective estimators for this problem. This is empirically shown to be flexible and efficient alternatives to existing techniques
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.
Full textThe 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
Pascal, Lucas. "Optimization of deep multi-task networks." Electronic Thesis or Diss., Sorbonne université, 2021. http://www.theses.fr/2021SORUS535.
Full textMulti-task learning (MTL) is a learning paradigm involving the joint optimization of parameters with respect to multiple tasks. By learning multiple related tasks, a learner receives more complete and complementary information on the input domain from which the tasks are issued. This allows to gain better understanding of the domain by building a more accurate set of assumptions of it. However, in practice, the broader use of MTL is hindered by the lack of consistent performance gains observed by deep multi-task networks. It is often the case that deep MTL networks suffer from performance degradation caused by task interference. This thesis addresses the problem of task interference in Multi-Task learning, in order to improve the generalization capabilities of deep neural networks
Chen, Mickaël. "Learning with weak supervision using deep generative networks." Electronic Thesis or Diss., Sorbonne université, 2020. http://www.theses.fr/2020SORUS024.
Full textMany successes of deep learning rely on the availability of massive annotated datasets that can be exploited by supervised algorithms. Obtaining those labels at a large scale, however, can be difficult, or even impossible in many situations. Designing methods that are less dependent on annotations is therefore a major research topic, and many semi-supervised and weakly supervised methods have been proposed. Meanwhile, the recent introduction of deep generative networks provided deep learning methods with the ability to manipulate complex distributions, allowing for breakthroughs in tasks such as image edition and domain adaptation. In this thesis, we explore how these new tools can be useful to further alleviate the need for annotations. Firstly, we tackle the task of performing stochastic predictions. It consists in designing systems for structured prediction that take into account the variability in possible outputs. We propose, in this context, two models. The first one performs predictions on multi-view data with missing views, and the second one predicts possible futures of a video sequence. Then, we study adversarial methods to learn a factorized latent space, in a setting with two explanatory factors but only one of them is annotated. We propose models that aim to uncover semantically consistent latent representations for those factors. One model is applied to the conditional generation of motion capture data, and another one to multi-view data. Finally, we focus on the task of image segmentation, which is of crucial importance in computer vision. Building on previously explored ideas, we propose a model for object segmentation that is entirely unsupervised