La multiplication des contenus stéréoscopique augmente les risques de piratage numérique. La solution technologique par tatouage relève ce défi. En pratique, le défi d’une approche de tatouage est d'atteindre l’équilibre fonctionnel entre la transparence, la robustesse, la quantité d’information insérée et le coût de calcul. Tandis que la capture et l'affichage du contenu 3D ne sont fondées que sur les deux vues gauche/droite, des représentations alternatives, comme les cartes de disparité devrait également être envisagée lors de la transmission/stockage. Une étude spécifique sur le domaine d’insertion optimale devient alors nécessaire. Cette thèse aborde les défis mentionnés ci-dessus. Tout d'abord, une nouvelle carte de disparité (3D video-New Three Step Search- 3DV-SNSL) est développée. Les performances des 3DV-NTSS ont été évaluées en termes de qualité visuelle de l'image reconstruite et coût de calcul. En comparaison avec l'état de l'art (NTSS et FS-MPEG) des gains moyens de 2dB en PSNR et 0,1 en SSIM sont obtenus. Le coût de calcul est réduit par un facteur moyen entre 1,3 et 13. Deuxièmement, une étude comparative sur les principales classes héritées des méthodes de tatouage 2D et de leurs domaines d'insertion optimales connexes est effectuée. Quatre méthodes d'insertion appartenant aux familles SS, SI et hybride (Fast-IProtect) sont considérées. Les expériences ont mis en évidence que Fast-IProtect effectué dans la nouvelle carte de disparité (3DV-NTSS) serait suffisamment générique afin de servir une grande variété d'applications. La pertinence statistique des résultats est donnée par les limites de confiance de 95% et leurs erreurs relatives inférieurs er <0.1 / The explosion in stereoscopic video distribution increases the concerns over its copyright protection. Watermarking can be considered as the most flexible property right protection technology. The watermarking applicative issue is to reach the trade-off between the properties of transparency, robustness, data payload and computational cost. While the capturing and displaying of the 3D content are solely based on the two left/right views, some alternative representations, like the disparity maps should also be considered during transmission/storage. A specific study on the optimal (with respect to the above-mentioned properties) insertion domain is also required. The present thesis tackles the above-mentioned challenges. First, a new disparity map (3D video-New Three Step Search - 3DV-NTSS) is designed. The performances of the 3DV-NTSS were evaluated in terms of visual quality of the reconstructed image and computational cost. When compared with state of the art methods (NTSS and FS-MPEG) average gains of 2dB in PSNR and 0.1 in SSIM are obtained. The computational cost is reduced by average factors between 1.3 and 13. Second, a comparative study on the main classes of 2D inherited watermarking methods and on their related optimal insertion domains is carried out. Four insertion methods are considered; they belong to the SS, SI and hybrid (Fast-IProtect) families. The experiments brought to light that the Fast-IProtect performed in the new disparity map domain (3DV-NTSS) would be generic enough so as to serve a large variety of applications. The statistical relevance of the results is given by the 95% confidence limits and their underlying relative errors lower than er<0.1

Cette thèse présente trois principaux sujets de recherche, les deux premiers étant indépendants et le dernier indiquant la relation des deux premières problématiques dans un cas concret.Dans la première partie nous nous intéressons au problème de transport optimal martingale dans l’espace de Skorokhod, dont le premier but est d’étudier systématiquement la tension des plans de transport martingale. On s’intéresse tout d’abord à la semicontinuité supérieure du problème primal par rapport aux distributions marginales. En utilisant la S-topologie introduite par Jakubowski, on dérive la semicontinuité supérieure et on montre la première dualité. Nous donnons en outre deux problèmes duaux concernant la surcouverture robuste d’une option exotique, et nous établissons les dualités correspondantes, en adaptant le principe de la programmation dynamique et l’argument de discrétisation initie par Dolinsky et Soner.La deuxième partie de cette thèse traite le problème du plongement de Skorokhod optimal. On formule tout d’abord ce problème d’optimisation en termes de mesures de probabilité sur un espace élargi et ses problèmes duaux. En utilisant l’approche classique de la dualité; convexe et la théorie d’arrêt optimal, nous obtenons les résultats de dualité. Nous rapportons aussi ces résultats au transport optimal martingale dans l’espace des fonctions continues, d’où les dualités correspondantes sont dérivées pour une classe particulière de fonctions de paiement. Ensuite, on fournit une preuve alternative du principe de monotonie établi par Beiglbock, Cox et Huesmann, qui permet de caractériser les optimiseurs par leur support géométrique. Nous montrons à la fin un résultat de stabilité qui contient deux parties: la stabilité du problème d’optimisation par rapport aux marginales cibles et le lien avec un autre problème du plongement optimal.La dernière partie concerne l’application de contrôle stochastique au transport optimal martingale avec la fonction de paiement dépendant du temps local, et au plongement de Skorokhod. Pour le cas d’une marginale, nous retrouvons les optimiseurs pour les problèmes primaux et duaux via les solutions de Vallois, et montrons en conséquence l’optimalité des solutions de Vallois, ce qui regroupe le transport optimal martingale et le plongement de Skorokhod optimal. Quand au cas de deux marginales, on obtient une généralisation de la solution de Vallois. Enfin, un cas spécial de plusieurs marginales est étudié, où les temps d’arrêt donnés par Vallois sont bien ordonnés. / This PhD dissertation presents three research topics, the first two being independent and the last one relating the first two issues in a concrete case.In the first part we focus on the martingale optimal transport problem on the Skorokhod space, which aims at studying systematically the tightness of martingale transport plans. Using the S-topology introduced by Jakubowski, we obtain the desired tightness which yields the upper semicontinuity of the primal problem with respect to the marginal distributions, and further the first duality. Then, we provide also two dual formulations that are related to the robust superhedging in financial mathematics, and we establish the corresponding dualities by adapting the dynamic programming principle and the discretization argument initiated by Dolinsky and Soner.The second part of this dissertation addresses the optimal Skorokhod embedding problem under finitely-many marginal constraints. We formulate first this optimization problem by means of probability measures on an enlarged space as well as its dual problems. Using the classical convex duality approach together with the optimal stopping theory, we obtain the duality results. We also relate these results to the martingale optimal transport on the space of continuous functions, where the corresponding dualities are derived for a special class of reward functions. Next, We provide an alternative proof of the monotonicity principle established in Beiglbock, Cox and Huesmann, which characterizes the optimizers by their geometric support. Finally, we show a stability result that is twofold: the stability of the optimization problem with respect to target marginals and the relation with another optimal embedding problem.The last part concerns the application of stochastic control to the martingale optimal transport with a payoff depending on the local time, and the Skorokhod embedding problem. For the one-marginal case, we recover the optimizers for both primal and dual problems through Vallois' solutions, and show further the optimality of Vallois' solutions, which relates the martingale optimal transport and the optimal Skorokhod embedding. As for the two-marginal case, we obtain a generalization of Vallois' solution. Finally, a special multi-marginal case is studied, where the stopping times given by Vallois are well ordered.

No description available.

No description available.

No description available.

We give a strongly polynomial time combinatorial algorithm to minimise the largest eigenvalue of the weighted Laplacian of a bipartite graph. This is accomplished by solving the dual graph embedding problem which arises from a semidefinite programming formulation. In particular, the problem for trees can be solved in time cubic in the number of vertices.

Digital innovation (DI) has enabled businesses to enhance their existing market offerings by integrating digital features. Despite advanced technologies, substantial marketing efforts, and global recognition, businesses can still struggle to convince customers to adopt their digital market offering. This process of spreading novel innovation is known as diffusion. In the fast-growing digital world, due to the unique characteristics of DI, traditional diffusion theories and models shows limited explanatory power, creating challenges for researchers and practitioners alike. With the aim to explore these challenges, we position our research within the IS literature with the following research question: "How and why do diffusion enablers and barriers emerge during digital innovation?". We conducted an interpretive case study of Company X, one of the world's largest consulting firms and an active DI practitioner. Our findings suggest that digital innovation diffusion can be enabled or hindered by several understudied interdependencies in its technological architecture. Furthermore, for successful diffusion, how DI distributed division of labor between layers must be effectively embedded and aligned for value in the clients' context. This study provides novel insights, exciting research avenues, and a diffusion strategy for DI practitioners.

<p>Efficient topic modeling is needed to support applications that aim at identifying main themes from a collection of documents. In this thesis, a reduced vector embedding representation and particle swarm optimization (PSO) are combined to develop a topic modeling strategy that is able to identify representative themes from a large collection of documents. Documents are encoded using a reduced, contextual vector embedding from a general-purpose pre-trained language model (sBERT). A modified PSO algorithm (pPSO) that tracks particle fitness on a dimension-by-dimension basis is then applied to these embeddings to create clusters of related documents. The proposed methodology is demonstrated on three datasets across different domains. The first dataset consists of posts from the online health forum r/Cancer. The second dataset is a collection of NY Times abstracts and is used to compare</p> <p>the proposed model to LDA. The third is a standard benchmark dataset for topic modeling which consists of a collection of messages posted to 20 different news groups. It is used to compare state-of-the-art generative document models (i.e., ETM and NVDM) to pPSO. The results show that pPSO is able to produce interpretable clusters. Moreover, pPSO is able to capture both common topics as well as emergent topics. The topic coherence of pPSO is comparable to that of ETM and its topic diversity is comparable to NVDM. The assignment parity of pPSO on a document completion task exceeded 90% for the 20News-Groups dataset. This rate drops to approximately 30% when pPSO is applied to the same Skip-Gram embedding derived from a limited, corpus specific vocabulary which is used by ETM and NVDM.</p>

[pt] A tarefa de resolução de correferência consiste em identificar e agrupar trechos de um texto de acordo com as entidades do mundo real a que se referem. Apesar de já ter sido abordada em outras conferências, a CoNLL de 2012 é um marco pela qualidade das bases de dados, das métricas e das soluções apresentadas. Naquela edição, o modelo vencedor utilizou um perceptron estruturado para otimizar uma árvore latente de antecedentes, atingindo a pontuação de 63.4 na métrica oficial para o dataset de teste em inglês. Nos anos seguintes, as bases e métricas apresentadas na conferência se tornaram o benchmark para a tarefa de correferência. Com novas técnicas de aprendizado de máquina desenvolvidas, soluções mais elaboradas foram apresentadas. A utilização de redes neurais rasas atingiu a pontuação de 68.8; a adição de representação contextual elevou o estado da arte para 73.0; redes neurais profundas melhoraram o baseline para 76.9 e o estado da arte atual, que é uma combinação de várias dessas técnicas, está em 79.6. Neste trabalho é apresentado uma análise de como as técnicas de representação de palavras Bag of Words, GloVe, BERT e SpanBERT utilizadas com árvores latentes de antecedentes se comparam com o modelo original de 2012. O melhor modelo encontrado foi o que utiliza SpanBERT com uma margem muito larga, o qual atingiu pontuação de 61.3 na métrica da CoNLL 2012, utilizando o dataset de teste. Com estes resultados, mostramos que é possível utilizar técnicas avançadas em estruturas mais simples e ainda obter resultados competitivos na tarefa de correferência. Além disso, melhoramos a performance de um framework de código aberto para correferência, a fim de contemplar soluções com maior demanda de memória e processamento. / [en] The coreference resolution task consists of to identify and group spans of text related to the same real-world entity. Although it has been approached in other conferences, the 2012 CoNLL is a milestone due to the improvement in the quality of its dataset, metrics, and the presented solutions. In that edition, the winning model used a structured perceptron to optimize an antecedent latent tree, achieving 63.4 on the official metric for the English test dataset. During the following years, the metrics and dataset presented in that conference became the benchmark for the coreference task. With new machine learning techniques, more elaborated solutions were presented. The use of shallow neural networks achieved 68.8; adding contextual representation raised the state-of-the-art to 73.0; deep neural networks improved the baseline to 76.9 and the current state-of-the-art, which is a combination of many of these techniques, is at 79.6. This work presents an analysis of how the word embedding mechanisms Bag of Words, GloVe, BERT and SpanBERT, used with antecedent latent trees, are compared to the original model of 2012. The best model found used SpanBERT with a very large margin, achieving 61.3 in the CoNLL 2012 metric using the test dataset. With these results, we show that it is possible to use advanced techniques in simpler structures and still achieve competitive results in the coreference task. Besides that, we improved the performance of an open source framework for coreference, so it can manage solution that demand more memory and processing.

Vid stor mängd data i form av kundomdömen kan det vara ett relativt tidskrävande arbeteatt bedöma varje omdömes sentiment manuellt, om det är positivt eller negativt laddat. Denna avhandling har utförts för att automatiskt kunna klassificera kundomdömen efter positiva eller negativa omdömen vilket hanterades med hjälp av maskininlärning. Tre olika djupa neurala nätverk testades och jämfördes med hjälp av två olika ramverk, TensorFlow och Keras, på både större och mindre datamängder. Även olika inbäddningsmetoder testades med de neurala nätverken. Den bästa kombination av neuralt nätverk, ramverk och inbäddningsmetod var ett Convolutional Neural Network (CNN) som använde ordinbäddningsmetoden Word2Vec, var skriven i ramverket Keras och gav en träffsäkerhetpå ca 88.87% med en avvikelse på ca 0.4%. CNN gav bäst resultat i alla olika tester framför de andra två neurala nätverken, Recurrent Neural Network (RNN) och Convolutional Recurrent Neural Network (CRNN) / With large amount of data in the form of customer reviews, it could be time consuming to manually go through each review and decide if its sentiment is positive or negative. This thesis have been done to automatically classify client reviews to determine if a review is positive or negative. This was dealt with by machine learning. Three different deep neural network was tested on greater and lesser datasets, and compared with the help of two different frameworks, TensorFlow and Keras. Different embedding methods were tested on the neural networks. The best combination of a neural network, a framework and anembedding was the Convolutional Neural Network (CNN) which used the word embedding method Word2Vec, was written in Keras framework and gave an accuracy of approximately 88.87% with a deviation of approximately 0.4%. CNN scored a better result in all of the tests in comparison with the two other neural networks, Recurrent NeuralNetwork (RNN) and Convolutional Recurrent Neural Network (CRNN).