Many concrete problems, ranging from Portfolio selection to Water resource management, may be cast into a multiobjective programming framework. The simplistic way of superseding blindly conflictual goals by one objective function let no chance to the model but to churn out meaningless outcomes. Hence interest of discussing ways for tackling Multiobjective Programming Problems. More than this, in many real-life situations, uncertainty and imprecision are in the state of affairs. In this dissertation we discuss ways for solving Multiobjective Programming Problems with fixed and fuzzy coefficients. No preference, a priori, a posteriori, interactive and metaheuristic methods are discussed for the deterministic case. As far as the fuzzy case is concerned, two approaches based respectively on possibility measures and on Embedding Theorem for fuzzy numbers are described. A case study is also carried out for the sake of illustration. We end up with some concluding remarks along with lines for further development, in this field. / Operations Research / M. Sc. (Operations Research)

No description available.

Evaluating separability is fundamental to pattern recognition. A plethora of embedding methods, such as dimension reduction and network embedding algorithms, have been developed to reveal the emergence of geometrical patterns in a low-dimensional space, where high-dimensional sample and node similarities are approximated by geometrical distances. However, statistical measures to evaluate the separability attained by the embedded representations are missing. Traditional cluster validity indices (CVIs) might be applied in this context, but they present multiple limitations because they are not specifically tailored for evaluating the separability of embedded results. This work introduces a new rationale called projection separability (PS), which provides a methodology expressly designed to assess the separability of data samples in a reduced (i.e., low-dimensional) geometrical space. In a first case study, using this rationale, a new class of indices named projection separability indices (PSIs) is implemented based on four statistical measures: Mann-Whitney U-test p-value, Area Under the ROC-Curve, Area Under the Precision-Recall Curve, and Matthews Correlation Coefficient. The PSIs are compared to six representative cluster validity indices and one geometrical separability index using seven nonlinear datasets and six different dimension reduction algorithms. In a second case study, the PS rationale is extended to define and measure the geometric separability (linear and nonlinear) of mesoscale patterns in complex data visualization by solving the traveling salesman problem, offering experimental evidence on the evaluation of community separability of network embedding results using eight real network datasets and three network embedding algorithms. The results of both studies provide evidence that the implemented statistical-based measures designed on the basis of the PS rationale are more accurate than the other indices and can be adopted not only for evaluating and comparing the separability of embedded results in the low-dimensional space but also for fine-tuning embedding algorithms’ hyperparameters. Besides these advantages, the PS rationale can be used to design new statistical-based separability measures other than the ones presented in this work, providing the community with a novel and flexible framework for assessing separability.

We explore in this study the different Natural Language Processing state-of-the art technologies that allow transforming textual data into numerical representation. We go through the theory of the existing traditional methods as well as the most recent ones. This thesis focuses on the recent advances in Natural Language processing being developed upon the Transfer model. One of the most relevant innovations was the release of a deep bidirectional encoder called BERT that broke several state of the art results. BERT utilises Transfer Learning to improve modelling language dependencies in text. BERT is used for several different languages, other specialized model were released like the french BERT: Camembert. This thesis compares the language models of these different pre-trained models and compares their capability to insure a domain adaptation. Using the multilingual and the french pre-trained version of BERT and a dataset from AXA France’s emails, clients’ messages, legal documents, insurance documents containing over 60 million words. We fine-tuned the language models in order to adapt them on the Axa insurance’s french context to create a French AXAInsurance BERT model. We evaluate the performance of this model on the capability of the language model of predicting a masked token based on the context. BERT proves to perform better : modelling better the french AXA’s insurance text without finetuning than Camembert. However, with this small amount of data, Camembert is more capable of adaptation to this specific domain of insurance. / I denna studie undersöker vi de senaste teknologierna för Natural Language Processing, som gör det möjligt att omvandla textdata till numerisk representation. Vi går igenom teorin om befintliga traditionella metoder såväl som de senaste. Denna avhandling fokuserar på de senaste framstegen inom bearbetning av naturliga språk som utvecklats med hjälp av överföringsmodellen. En av de mest relevanta innovationerna var lanseringen av en djup dubbelriktad kodare som heter BERT som bröt flera toppmoderna resultat. BERT använder Transfer Learning för att förbättra modelleringsspråkberoenden i text. BERT används för flera olika språk, andra specialmodeller släpptes som den franska BERT: Camembert. Denna avhandling jämför språkmodellerna för dessa olika förutbildade modeller och jämför deras förmåga att säkerställa en domänanpassning. Med den flerspråkiga och franska förutbildade versionen av BERT och en dataset från AXA Frankrikes epostmeddelanden, kundmeddelanden, juridiska dokument, försäkringsdokument som innehåller över 60 miljoner ord. Vi finjusterade språkmodellerna för att anpassa dem till Axas försäkrings franska sammanhang för att skapa en fransk AXAInsurance BERT-modell. Vi utvärderar prestandan för denna modell på förmågan hos språkmodellen att förutsäga en maskerad token baserat på sammanhanget. BERTpresterar bättre: modellerar bättre den franska AXA-försäkringstexten utan finjustering än Camembert. Men med denna lilla mängd data är Camembert mer kapabel att anpassa sig till denna specifika försäkringsdomän.

No description available.

Cette thèse traite du développement et de l’implémentation de nouvelles méthodes visant à décrire la corrélation électronique forte dans les molécules et les solides. Après avoir introduit l’état de l’art des méthodes utilisées en chimie quantique et en physique de la matière condensée, une nouvelle méthode hybride combinant théorie de la fonction d’onde et théorie de la fonctionnelle de la densité (DFT) est présentée et s’intitule “site-occupation embedding theory” (SOET). Celle-ci est appliquée au modèle de Hubbard à une dimension. Ensuite, le problème du gap fondamental est revisité en DFT pour les ensembles, où la dérivée discontinue est réécrite comme une fonctionnelle de la densité de l'état fondamental. Enfin, une extension à la chimie quantique est proposée, basée sur une fonction d’onde de séniorité zéro complémentée par une fonctionnelle de la matrice densité, et exprimée dans la base des orbitales naturelles. / The thesis deals with the development and implementation of new methods for the description of strong electron correlation effects in molecules and solids. After introducing the state of the art in quantum chemistry and in condensed matter physics, a new hybrid method so-called ``site-occupation embedding theory'' (SOET) is presented and is based on the merging of wavefunction theory and density functional theory (DFT). Different formulations of this theory are described and applied to the one-dimensional Hubbard model. In addition, a novel ensemble density functional theory approach has been derived to extract the fundamental gap exactly. In the latter approach, the infamous derivative discontinuity is reformulated as a derivative of a weight-dependent exchange-correlation functional. Finally, a quantum chemical extension of SOET is proposed and based on a seniority-zero wavefunction, completed by a functional of the density matrix and expressed in the natural orbital basis.

Cette thèse porte sur l’étude des méthodes de mesure pour l’analyse vectorielle des circuits microélectroniques en technologie intégrée aux fréquences millimétriques. Pour réussir à extraire les paramètres intrinsèques de circuits réalisés aux longueurs d'ondes millimétriques, les méthodes actuelles de calibrage et de de-embedding sont d'autant moins précises que les fréquences de fonctionnement visées augmentent au-delà de 100 GHz notamment. Cela est d’autant plus vrai pour la caractérisation des dispositifs passifs tels que des lignes de propagation. La motivation initiale de ces travaux de thèse venait du fait qu'il était difficile d'expliquer l’origine exacte des pertes mesurées pour des lignes coplanaires à ondes lentes (lignes S-CPW) aux fréquences millimétriques. Etait-ce un problème de mesure brute, un problème de méthode de-embedding qui sous-estime les pertes, une modélisation insuffisante des effets des cellules adjacentes, ou encore la création d'un mode de propagation perturbatif ?Le travail a principalement consisté à évaluer une dizaine de méthodes de de-embedding au-delà de 65 GHz et à classifier ces méthodes en 3 groupes pour pouvoir les comparer de manière pertinente. Cette étude s’est déroulée en 3 phases.Dans la première phase, il s’agissait de comparer les méthodes de de-embedding tout en maitrisant les modèles électriques des plots et des lignes d’accès. Cette phase a permis de dégager les conditions optimales d’utilisation pour pouvoir appliquer ces différentes méthodes de de-embedding.Dans la deuxième phase, la modélisation des structures de test a été réalisée à l’aide d’un simulateur électromagnétique 3D basé sur la méthode des éléments finis. Cette phase a permis de tester la robustesse des méthodes et d’envisager une méthode de-embedding originale nommée Half-Thru Method. Cette méthode donne des résultats comparables à la méthode TRL, méthode qui reste la plus performante actuellement. Cependant il reste difficile d'expliquer l'origine des pertes supplémentaires obtenues notamment dans la mesure des lignes à ondes lentes S-CPW.Une troisième phase de modélisation a alors consisté à prendre en compte les pointes de mesure et les cellules adjacentes à notre dispositif sous test. Plus de 80 structures de test ont été conçues en technologie AMS 0,35μm afin de comparer les différentes méthodes de de-embedding et d’en analyser les couplages avec les structures adjacentes, les pointes de mesure et les modes de propagation perturbatifs.Finalement, ce travail a permis de dégager un certain nombre de précautions à considérer à l’attention des concepteurs de circuits microélectroniques désirant caractériser leur circuit avec précision au-delà de 110 GHz. Il a également permis de mettre en place la méthode de de-embedding Half-Thru Method qui n'est basée sur aucun modèle électrique, au contraire des autres méthodes. / This thesis focuses on the study of vectorial measurement methods for analysing microelectronic circuits in integrated technology at millimeter wave frequencies. Current calibration and de-embedding methods are less precise for successfully extracting the intrinsic parameters of devices and circuits at millimeter wave frequencies, while the targeted operating frequencies are above 100 GHz. This is especially true for the characterization of passive devices such as propagation lines. The initial motivation of this thesis work was to explain the exact origin of the additional loss measured in Slow-Wave Coplanar Waveguides (S-CPW) lines at millimeter wave frequencies. Was it a problem of raw measurement or a problem of de-embedding method, which underestimates the losses? Or was it a problem of insufficient modeling of the effects of adjacent cells, or even the creation of a perturbation mode of propagation?This work consists of estimating many de-embedding methods beyond 65 GHz and classifies these methods into three groups to be able to compare them in a meaningful way. This study was conducted in three phases.In the first phase, we compared all the de-embedding methods with known electrical model parasitics of pad/accessline. This phase identifies the optimal conditions to use and apply these de-embedding methods.In the second phase, the modeling of test structures is performed using a 3D electromagnetic simulator based on finite element method. This phase tested the robustness of the methods and considered an original de-embedding method called Half-Thru de-embedding method. This method gives comparable results to the TRL method, which remains the most effective method. However, it remains difficult to explain the origin of additional losses obtained in measured S-CPW line.A third modeling phase was analysed to take into account the measurement of probes and the adjacent cells near our device under test. More than 80 test structures were designed in AMS 0.35 μm CMOS technology to compare the different de-embedding methods and analyse the link with adjacent cells, measuring probes and perturbation mode of propagation.Finally, this work has identified a number of precautions to consider for the attention of microelectronic circuit designers wishing to characterize their circuit with precision beyond 110 GHz. It also helped to establish Half-Thru Method de-embedding method, which is not based on electrical model, unlike other methods.

This thesis consists of four research papers. In the first paper we construct nonmetrizable compact convex sets with pathological sets of simpliciality, show- ing that the properties of the set of simpliciality known in the metrizable case do not hold without the assumption of metrizability. In the second paper we construct an example concerning remotal sets, answering thus a question of Martín and Rao, and present a new proof of the fact that in every infinite- dimensional Banach space there exists a closed convex bounded set which is not remotal. The third paper is a study of the relations between polynomials on Banach spaces and linear identities. We investigate under which conditions a linear identity is satisfied only by polynomials, and describe the space of poly- nomials satisfying such linear identity. In the last paper we study the coarse and uniform embeddability between Orlicz sequence spaces. We show that the embeddability between two Orlicz sequence spaces is in most cases determined only by the values of their upper Matuszewska-Orlicz indices. 1

Le travail de doctorat réalisé s'attache à étudier les nouveaux types d'interconnexions comme les TSV (Through Silicon Via), les lignes de redistribution (RDL) et les piliers de cuivre (Cu-Pillar) présentes dans le domaine de l'intégration 3D en microélectronique avancée, par exemple pour des applications de type « imager » où une puce « capteur optique » est empilée sur une puce « processeur ». Afin de comprendre et quantifier le comportement électrique de ces nouveaux composants d'interconnexion, une première problématique de la thèse s'articulait autour de la caractérisation électrique, sur une très large bande de fréquence (10 MHz - 60 GHz) de ces éléments, enfouis dans leurs environnements complexes d'intégration, en particulier avec l'analyse de l'impact des pertes dans les substrats de silicium dans une gamme de conductivités allant de très faible (0 S/m) à très forte (10 000 S/m). Par la suite, une nouvelle problématique prend alors naissance sur la nécessité de développer des modèles mathématiques permettant de prédire le comportement électrique des interconnexions 3D. Les modèles électriques développés doivent tenir compte des pertes, des couplages ainsi que de certains phénomènes liés à la montée en fréquence (courants de Foucault) en fonction des caractéristiques matériaux, des dimensions et des architectures (haute à faible densité d'intégration). Enfin, à partir des modèles développés, une dernière partie propose une étude sur les stratégies de routage dans les empilements 3D de puces à partir d'une analyse sur l'intégrité de signaux. En opposant différents environnements, débit de signaux binaires ou dimensions des TSV et des RDL des conclusions émergent sur les stratégies à adopter pour améliorer les performances des circuits conçus en intégration 3D. / The aim of this doctoral work is to study the new kind of interconnections like TSV (Through Silicon Via), redistribution lines (RDL) and copper pillars used in 3D integration context in advanced microelectronic components. An example of 3D integration application could be an imager designed by staking an optical sensor chip upon a processor chip. In order to understand and quantify the electrical behaviour of these new interconnection components, the first issue was about electrical characterization in a very wide frequency band (10 MHz - 60 GHz) of these elements, buried in their complex environment, in particular with the analysis of the silicon substrate loss impact which can be found in a wide band of conductivities from very low (0 S/m) to very high (10 000 S/m). Subsequently, a second issue appears from the need to develop mathematical models to predict the electrical behavior of 3D interconnects. The developed models have to take into account losses, coupling effects and some phenomena appearing with the rise of frequency (eddy currents) according to material characteristics, dimensions and architecture (from high to low density of integration). Finally, based on developed models, the last part presents a study on routing strategies in the 3D stacking chip from the analysis of signal integrity. By contrasting various environments, binary signals flow or dimensions of TSV and RDL, conclusions emerge on the best strategies to use to improve performances of circuits designed in 3D integration.

In this thesis we study embeddings of spaces of functions defined on Carnot- Carathéodory spaces. Main results of this work consist of conditions for Sobolev- type embeddings of higher order between rearrangement-invariant spaces. In a special case when the underlying measure space is the so-called X-PS domain in the Heisenberg group we obtain full characterization of a Sobolev embedding. The next set of main results concerns compactness of the above-mentioned em- beddings. In these cases we obtain sufficient conditions. We apply the general results to important particular examples of function spaces. In the final part of the thesis we present a new algorithm for approximation of the least concave majorant of a function defined on an interval complemented with the estimate of the error of such approximation. 1