Technical analysis has been thwarted in academic circles, due to the Efficient Market Hypothesis, which had significant empirical support early on. However recently, there is accumulating evidence that the markets are not as efficient and a new theory of price discovery, Heterogenous Market Hypothesis, is being proposed. As such, there is renewed interest and possibility in technical analysis, which identifies trends in price and volume based on aggregate repeatable human behavioural patterns. In this thesis we propose a new approach for modeling and working with technical analysis in high-frequency markets: dynamic Bayesian networks (DBNs). DBNs are a statistical modeling and learning framework that have had successful applications in other domains such as speech recognition, bio-sequencing, visual interpretation. It provides a coherent probabilistic framework (in a Bayesian sense), that can be used for both learning technical rules and inferring the hidden state of the system. We design a DBN to learn price and volume patterns in TSE60 stock market and find that our model is able to successfully identify runs and reversal out-of-sample in a statistically significant way.

Technical analysis has been thwarted in academic circles, due to the Efficient Market Hypothesis, which had significant empirical support early on. However recently, there is accumulating evidence that the markets are not as efficient and a new theory of price discovery, Heterogenous Market Hypothesis, is being proposed. As such, there is renewed interest and possibility in technical analysis, which identifies trends in price and volume based on aggregate repeatable human behavioural patterns. In this thesis we propose a new approach for modeling and working with technical analysis in high-frequency markets: dynamic Bayesian networks (DBNs). DBNs are a statistical modeling and learning framework that have had successful applications in other domains such as speech recognition, bio-sequencing, visual interpretation. It provides a coherent probabilistic framework (in a Bayesian sense), that can be used for both learning technical rules and inferring the hidden state of the system. We design a DBN to learn price and volume patterns in TSE60 stock market and find that our model is able to successfully identify runs and reversal out-of-sample in a statistically significant way.

One of the foremost characteristics of the mammalian visual system is the retinotopic mapping observed in the low-level visual processing centres; the spatial pattern of activation in the lateral geniculate nucleus and primary visual cortex corresponds topologically to the pattern of light falling on the retina. Various vision systems have been developed that take advantage of structured input such as retinotopy, however these systems are often not biologically plausible. Using a parsimonious approach for implementing retinotopy, one that is based on the biology of our visual pathway, we run simulations of visual learning using a deep belief network (DBN). Experiments show that we can successfully produce receptive fields and activation maps typical of the LGN and visual cortex respectively. These results may indicate a possible avenue of exploration into discovering the workings of the early visual system (and possibly more) on a neuronal level.

One of the foremost characteristics of the mammalian visual system is the retinotopic mapping observed in the low-level visual processing centres; the spatial pattern of activation in the lateral geniculate nucleus and primary visual cortex corresponds topologically to the pattern of light falling on the retina. Various vision systems have been developed that take advantage of structured input such as retinotopy, however these systems are often not biologically plausible. Using a parsimonious approach for implementing retinotopy, one that is based on the biology of our visual pathway, we run simulations of visual learning using a deep belief network (DBN). Experiments show that we can successfully produce receptive fields and activation maps typical of the LGN and visual cortex respectively. These results may indicate a possible avenue of exploration into discovering the workings of the early visual system (and possibly more) on a neuronal level.

Cette thèse a été réalisée dans le cadre d’une convention entre le Cerema (Centre d’Etude et d’Expertise sur les Risques, l’Environnement, la Mobilité et l’Aménagement) et l’ENTPE (Ecole Nationale des Travaux Publics de l’Etat). Ce travail concerne l’étude et la modélisation du comportement des enrobés bitumineux, dans les domaines des petites et grandes déformations, soit respectivement les comportements viscoélastique et viscoplastique.Après une étude bibliographique portant sur les propriétés mécaniques des enrobés bitumineux en petites et grandes déformations et la présentation des modèles de comportement viscoélastique linéaire et viscoplastique, ce mémoire s’organise en cinq chapitres. Les deux premières parties concernent le développement du modèle DBN, et la description de la version introduite dans cette thèse : la version EPPI (Elastique Parfaitement Plastique Isotrope). Cette formulation tridimensionnelle vise à décrire les comportements viscoélastique et viscoplastique des enrobés bitumineux avec un formalisme unique, suffisamment simple pour être implémenter dans un code de calcul aux éléments finis. Le modèle est ensuite développé et implémenté dans un code de calcul homogène, permettant ainsi de réaliser des simulations d’essais expérimentaux classiques.Les deux chapitres suivants traitent de la campagne expérimentale réalisée dans cette thèse. Il s’agit d’abord de décrire l’ensemble des procédures et conditions d’essais. Deux types d’essais sont réalisés : module complexe pour la caractérisation du comportement viscoélastique et traction simple pour la caractérisation du comportement viscoplastique. Ensuite, l’ensemble des résultats expérimentaux et observations issus de cette campagne sont présentés et permettent de dessiner quelques conclusions.Enfin, le dernier chapitre traite des simulations réalisées à partir du modèle DBNEPPI. Ces résultats permettent dans un premier temps de tester la validité du modèle (essais de module complexe, essais cycliques de traction-compression). Enfin une étude des effets transitoires lors des essais de module complexe est présentée. / The thesis has been realized within the framework of a partnership between the Cerema (Center for Studies and Expertise on Risks, Environment, Mobility, and Urban and Country Planning) and the ENTPE (National School for Public State Works). This work deals with the study and the simulation of the behavior of bituminous mixtures, in the domains of small and large amplitudes of solicitations. These domains correspond to the viscoelastic and viscoplastic behavior.First a bibliographical review on mechanical properties of the bituminous mixtures under small and large amplitudes of solicitations and a review of the several models describing the viscoelastic and viscoplastic behavior of bituminous mixtures is presented. This thesis is then composed of 5 main chapters.The two first deals with the development of the DBN model and mainly with the introduction of its new version EPPI (Elastic Perfectly Plastic and Isotropic). This formulation aims at describing the viscoelastic and viscoplastic behaviors together in a unique formalism, but simple enough to be implemented in a finite elements calculation program. This version of the DBN model is then implemented in a homogeneous computation code so as to be able to simulate classical experimental test.The two following chapter are devoted to the experimental campaign made during this research work. First a global description of the experimental procedures and test conditions is presented. Two kinds of tests are performed: complex modulus tests so as to characterize the viscoelastic behavior and simple traction tests for viscoplastic behavior. Finally results and observations issued from this campaign are detailed and some conclusions can be drawn.Finally the last chapter deals with the simulations performed thanks to the DBNEPPI model. These results aim at verifying the validity of the model introduced. Then a study about the transient effects during complex modulus tests on bituminous mixtures is detailed.

Machine learning has achieved great success in recent years, especially the deep learning algorithms based on Artificial Neural Network. However, high performance and large memories are needed for these models , which makes them not suitable for IoT device, as IoT devices have limited performance and should be low cost and less energy-consuming. Therefore, it is necessary to optimize the deep learning models to accommodate the resource-constrained IoT devices. This thesis is to seek for a possible solution of optimizing the ANN models to fit into the IoT devices and provide a hardware implementation of the ANN accelerator on FPGA. The contribution of this thesis mainly lies in two aspects: 1). analyze the sparsity in the two mainstream deep learning models – DBN and CNN. The DBN model consists of two hidden layers with Restricted Boltzmann Machines while the CNN model consists of 2 convolutional layers and 2 sub-sampling layer. Experiments have been done on the MNIST data set with the sparsity of 75%. The ratio of the multiplications resulting in near-zero values has been tested. 2). FPGA implementation of an ANN accelerator. This thesis designed a hardware accelerator for the inference process in ANN models on FPGA (Stratix IV: EP4SGX530KH40C2). The main part of hardware design is the processing array consists of 256 Multiply-Accumulators array, which can conduct multiply-accumulate operations of 256 synaptic connections simultaneously. 16-bit fixed point computation is used to reduce the hardware complexity, thus saving power and area. Based on the evaluation results, it is found that the ratio of the multiplications under the threshold of 2-5 is 75% for CNN with ReLU activation function, and is 83% for DBN with sigmoid activation function, respectively. Therefore, there still exists large space for complex ANN models to be optimized if the sparsity of data is fully utilized. Meanwhile, the implemented hardware accelerator is verified to provide correct results through 16-bit fixed point computation, which can be used as a hardware testing platform for evaluating the ANN models.

Relating statistical machine learning approaches to the automatic analysis of multiparty communicative events, such as meetings, is an ambitious research area. We have investigated automatic meeting segmentation both in terms of “Meeting Actions” and “Dialogue Acts”. Dialogue acts model the discourse structure at a fine grained level highlighting individual speaker intentions. Group meeting actions describe the same process at a coarse level, highlighting interactions between different meeting participants and showing overall group intentions. A framework based on probabilistic graphical models such as dynamic Bayesian networks (DBNs) has been investigated for both tasks. Our first set of experiments is concerned with the segmentation and structuring of meetings (recorded using multiple cameras and microphones) into sequences of group meeting actions such as monologue, discussion and presentation. We outline four families of multimodal features based on speaker turns, lexical transcription, prosody, and visual motion that are extracted from the raw audio and video recordings. We relate these lowlevel multimodal features to complex group behaviours proposing a multistreammodelling framework based on dynamic Bayesian networks. Later experiments are concerned with the automatic recognition of Dialogue Acts (DAs) in multiparty conversational speech. We present a joint generative approach based on a switching DBN for DA recognition in which segmentation and classification of DAs are carried out in parallel. This approach models a set of features, related to lexical content and prosody, and incorporates a weighted interpolated factored language model. In conjunction with this joint generative model, we have also investigated the use of a discriminative approach, based on conditional random fields, to perform a reclassification of the segmented DAs. The DBN based approach yielded significant improvements when applied both to the meeting action and the dialogue act recognition task. On both tasks, the DBN framework provided an effective factorisation of the state-space and a flexible infrastructure able to integrate a heterogeneous set of resources such as continuous and discrete multimodal features, and statistical language models. Although our experiments have been principally targeted on multiparty meetings; features, models, and methodologies developed in this thesis can be employed for a wide range of applications. Moreover both group meeting actions and DAs offer valuable insights about the current conversational context providing valuable cues and features for several related research areas such as speaker addressing and focus of attention modelling, automatic speech recognition and understanding, topic and decision detection.

In the area of image recognition, the so-called deep neural networks belong to the most promising models these days. They often achieve considerably better results than traditional techniques even without the necessity of any excessive task-oriented preprocessing. This thesis is devoted to the study and analysis of three basic variants of deep neural networks-namely the neocognitron, convolutional neural networks, and deep belief networks. Based on extensive testing of the described models on the standard task of handwritten digit recognition, the convolutional neural networks seem to be most suitable for the recognition of general image data. Therefore, we have used them also to classify images from two very large data sets-CIFAR-10 and ImageNet. In order to optimize the architecture of the applied networks, we have proposed a new pruning algorithm based on the Principal Component Analysis. Powered by TCPDF (www.tcpdf.org)

This thesis is dedicated to the development of statistical and computational methods for the analysis of DNA sequences and gene expression time series.<br /><br />First we study a parsimonious Markov model called Mixture Transition Distribution (MTD) model which is a mixture of Markovian transitions. The overly high number of constraints on the parameters of this model hampers the formulation of an analytical expression of the Maximum Likelihood Estimate (MLE). We propose to approach the MLE thanks to an EM algorithm. After comparing the performance of this algorithm to results from the litterature, we use it to evaluate the relevance of MTD modeling for bacteria DNA coding sequences in comparison with standard Markovian modeling.<br /><br />Then we propose two different approaches for genetic regulation network recovering. We model those genetic networks with Dynamic Bayesian Networks (DBNs) whose edges describe the dependency relationships between time-delayed genes expression. The aim is to estimate the topology of this graph despite the overly low number of repeated measurements compared with the number of observed genes. <br /><br />To face this problem of dimension, we first assume that the dependency relationships are homogeneous, that is the graph topology is constant across time. Then we propose to approximate this graph by considering partial order dependencies. The concept of partial order dependence graphs, already introduced for static and non directed graphs, is adapted and characterized for DBNs using the theory of graphical models. From these results, we develop a deterministic procedure for DBNs inference. <br /><br />Finally, we relax the homogeneity assumption by considering the succession of several homogeneous phases. We consider a multiple changepoint<br />regression model. Each changepoint indicates a change in the regression model parameters, which corresponds to the way an expression level depends on the others. Using reversible jump MCMC methods, we develop a stochastic algorithm which allows to simultaneously infer the changepoints location and the structure of the network within the phases delimited by the changepoints. <br /><br />Validation of those two approaches is carried out on both simulated and real data analysis.

Recent developments in social media and web technologies offer new opportunities to access, analyze and process ever-increasing amounts of fashion-related data. In the appealing context of design and fashion, our main goal is to automatically suggest fashionable outfits based on the preferences extracted from real-world data provided either by individual users or gathered from the internet. In our case, the clothing items have the form of 2D-images. Especially for visual data processing tasks, recent models of deep neural networks are known to surpass human performance. This fact inspired us to apply the idea of transfer learning to understand the actual variability in clothing items. The principle of transfer learning consists in extracting the internal representa- tions formed in large convolutional networks pre-trained on general datasets, e.g., ImageNet, and visualizing its (similarity) structure. Together with transfer learn- ing, clustering algorithms and the image color schemes can be, namely, utilized when searching for related outfit items. Viable means applicable to generating new out- fits include deep belief networks and genetic algorithms enhanced by a convolutional network that models the outfit fitness. Although fashion-related recommendations remain highly subjective, the results we have achieved...