Many problems appearing in scientific computing and other areas can be formulated as a graph partitioning problems. Examples include data distribution for parallel computers, decomposition of sparse matrices and VLSI-design. In this survey we present the graph partitioning problem, describe some applications and introduce many of the algorithms used to solve the problem.

L’objectif de ma thèse était d’étudier de façon plus approfondie l’influence des émotions sur la perception du temps, et surtout d’essayer d’identifier les mécanismes sous-jacents, encore méconnus aujourd’hui. Pour cela, j’ai mené 6 études organisées autour de 2 axes. L’objectif de l’Axe 1 était d’identifier les mécanismes impliqués dans l’effet des émotions sur la perception de la durée de stimuli émotionnels dans une tâche de bissection temporelle. L’objectif de l’Axe 2 était d’élaborer de nouvelles techniques d’induction permettant d’observer des effets plus robustes. Les résultats révèlent un allongement subjectif du temps, avec différents types de stimuli émotionnels fortement activateurs sur le plan physiologique. Cet effet apparait avec plusieurs types de stimuli menaçants et un large éventail de durées, allant de 0.3 à 8 s. Il apparait également de manière similaire chez l’adulte et l’enfant, du moins dès l’âge de 5 ans. De plus, cette dilatation du temps n’est pas accompagnée d’une amélioration de la sensibilité au temps, même quand la discrimination temporelle s’avère difficile. L’ensemble de ces données suggère un effet robuste des émotions fortement activatrices sur le jugement du temps, qui serait expliqué principalement par des mécanismes automatiques, et provoquerait l’accélération de l’horloge interne. Dans les situations de menace, cette accélération de l'horloge permettrait d’agir le plus rapidement possible. Quand le temps passe plus vite, l’individu est plus promptement prêt à agir. Toutefois, d’autres études sont encore nécessaires pour trancher la question de l’automaticité des processus et l’absence d’effet attentionnel. / The objective of this thesis was to study more precisely the influence of emotions on time perception, and above all to try to identify the underlying mechanisms, still unknown to this day. For this, I conducted 6 studies organized around 2 axes. The goal of Axis 1 was to identify the mechanisms involved in the effect of emotions on the duration judgment of emotional stimuli in a temporal bisection task. The goal of Axis 2 was to develop new induction methods in order to observe more robust effects.The results reveal a subjective lengthening of time, with different types of highly arousing emotional stimuli. This effect appears with several types of threatening stimuli and a wide range of durations, ranging from 0.3 to 8 s. It also appears similarly in adults and children, at least from the age of 5. Moreover, this time dilatation is not accompanied by an improvement in time sensitivity, even when temporal discrimination was difficult. All of these data suggest a robust effect of highly arousing emotions on time perception, which would be explained mainly by automatic mechanisms, and would cause the acceleration of the internal clock. In a situation of danger, this clock acceleration would allow the individual to react as quickly as possible. When time goes faster, the individual is prepared to act quicker. However, further studies are still needed to resolve the issue of automatic processes and the absence of attentional effects.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2015-10-22T11:43:16Z No. of bitstreams: 2 Dissertação - Maria Isabel Pereira Bezerra Almeida - 2014.pdf: 582154 bytes, checksum: 45944bf34da8b157f3aec7864ea90d4c (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2015-10-22T11:44:55Z (GMT) No. of bitstreams: 2 Dissertação - Maria Isabel Pereira Bezerra Almeida - 2014.pdf: 582154 bytes, checksum: 45944bf34da8b157f3aec7864ea90d4c (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2015-10-22T11:44:55Z (GMT). No. of bitstreams: 2 Dissertação - Maria Isabel Pereira Bezerra Almeida - 2014.pdf: 582154 bytes, checksum: 45944bf34da8b157f3aec7864ea90d4c (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2014-03-17 / This work aims to provide high school teachers and Higher Education workshop optimization with the aid of WinPlotr program to encourage the study of mathematics and its applications . Optimization in mathematics is to study problems for maximization or minimization of functions with the choice of methods that lead to the great demand . Given this context , this work has the theme of a workshop with the help of optimization software WinPlotr and aims to encourage the student ’s third year of high school or the early years of Higher Education to want to know new methods of learning. Presents troubleshooters of Applied Mathematics with maximization or minimization . With the use of derivatives and the Bisection method , and the second method is often used in calculations of roots and was adapted for this work . It also offers all the methodological planning a workshop to be worked into eight classes dynamic , creative and with the proposal to encourage students to have a more robust and accurate knowledge fashion . / Este trabalho tem o objetivo de oferecer aos professores do Ensino Medio e do Ensino Superior uma oficina de otimizacao com o auxılio do programa WinPlotr destinado a incentivar o estudo da Matem´atica e suas aplicacoes. Otimizacao em matematica consiste em estudar problemas para maximiza¸c˜ao ou minimizacao de funcoes com a escolha de metodos que conduzam `a procura do otimo. Diante deste contexto, este trabalho tem como tema uma oficina de otimiza¸c˜ao com o auxılio do software WinPlotr e tem como objetivo incentivar o aluno do terceiro ano do Ensino Medio ou dos anos iniciais do Ensino Superior a querer conhecer novos metodos de aprendizagem. Apresenta resolucoes de problemas de Matem´atica Aplicada com maximiza¸c˜ao ou minimiza¸c˜ao. Com o uso de derivadas e pelo m´etodo da Bissec¸c˜ao, sendo que o segundo m´etodo ´e muito utilizado em calculos de raızes e foi adaptado para esse trabalho. Contempla ainda todo o planejamento metodologico de uma oficina para ser trabalhada em oito aulas de forma dinamica, criativa e com a proposta de estimular os alunos a terem um conhecimento mais solido e rigoroso.

A Topological Interlocking Configuration (TIC) is an assembly where the shape and alignment of the blocks define the kinematic constraints. Conventional TICs are single-layered structures made of convex blocks. The interface between the blocks in an assembly is face-to-face contact. The traditional convention disregards the use of joinery, adhesive, or other mechanisms that keep two pieces next to each other. However, TICs require a support structure that prevents the lateral strain of the blocks.<br><br>The generation process of a TIC starts with a surface tessellation that describes a geometric domain. Each tile in the tessellation represents a traversal section of a block. For regular tessellations and uniform generation parameters, such sections lie in the middle of their respective blocks. Additionally, such conditions guarantee the blocks align adequately with each other. If one of such conditions does not hold, then the resultant blocks may not be aligned. Furthermore, there could be overlapping between the blocks, which makes a TIC invalid.<br><br>Traditionally, the generation parameters are angle values set at the edges of the tiles. The angles must match between tiles such that each block in the assembly has a geometry that imposes kinematic constraints to its neighboring blocks. Using the same angle values on regular and semi-regular tessellation produces feasible blocks. That is not the case for non-regular tessellations, curvilinear surfaces, and free-form 3D meshes. In such cases, the generation method must find specific angle values to design the blocks and reduce overlapping.<br><br>In this thesis, we propose a TIC generation framework focused on the generation of valid interlocking assemblies based on multiple types of surface tessellations. We start with the Height-Bisection method, a TIC generation approach that uses the distances from a tile to its respective evolution sections as the generation parameters. The method considers the bisector vectors between two tiles to define the parameters that generate aligned blocks to each other. We expand the generation model to a complete pipeline process that finds feasible generation parameters. The pipeline includes clipping parameters and methods in case that overlapping between blocks cannot be avoided.<br><br>Additionally, we describe a generalization of the mid-section evolution concept to include multiple evolution steps during the generation process. Our approach considers the angles and distances required to generate infinitely many TICs, including shapes that are not possible using the traditional generation method and the Height-Bisection method. Finally, we consider the interlocking assemblies that cannot maintain static equilibrium due to the shape of the surface tessellation. We consider the Structure Feasibility Analysis method to find the location and magnitude of the minimum tension forces that guarantee a TIC will reach a static equilibrium state. We describe how to update the generation parameters according to the results of the feasibility analysis. Our results show that the proposed pipeline generates valid TICs based on different surface tessellations, including closed and free-form shapes.

Many problems appearing in scientific computing and other areas can be formulated as a graph partitioning problems. Examples include data distribution for parallel computers, decomposition of sparse matrices and VLSI-design. In this survey we present the graph partitioning problem, describe some applications and introduce many of the algorithms used to solve the problem.

Submitted by Cláudia Bueno (claudiamoura18@gmail.com) on 2016-04-06T17:13:53Z No. of bitstreams: 2 Dissertação - Pâmella Almeida Quintino Jordan - 2015.pdf: 2688001 bytes, checksum: 877af9d2fad9e949f2e145fa570a729b (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2016-04-07T10:48:02Z (GMT) No. of bitstreams: 2 Dissertação - Pâmella Almeida Quintino Jordan - 2015.pdf: 2688001 bytes, checksum: 877af9d2fad9e949f2e145fa570a729b (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2016-04-07T10:48:02Z (GMT). No. of bitstreams: 2 Dissertação - Pâmella Almeida Quintino Jordan - 2015.pdf: 2688001 bytes, checksum: 877af9d2fad9e949f2e145fa570a729b (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2015-07-27 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / In the field of applied mathematics, the study of numerical methods has its relief importance and it shows a vast variety. In this word will be study four these methods: Bisection, Newton, Secant and Monte Carlo. The first three are methods studied in Numerical Calculus discipline, normally given in graduation from Exactn Science; their goal is the calculation of real functions roots; they work through recurrent sequence converge the exact roots. One of the applications of the fourth method is the calculation of the area under a curve, using, for this, random numbers. In each method we will study the motivation followed of their eff ectiveness. As we will see, there are many areas that need methods, as studied here, for solving problems; the engineering, Chemistry, Physics and Biology are just some of them. In order to demonstrate the importance of the four methods we will present situations that can be applied. Demonstrated the need and eff ectiveness of all methods, we can understand them as an excellent alternative to solving problems in mathematics applied area, especially when conventional resolution methods cannot play a satisfactory role. We must emphasize, also, the importance of using software and computer programs to better performance of the methods. / No campo da matemática aplicada o estudo de métodos numéricos tem sua relevada importância e apresenta uma vasta variedade. Neste trabalho serão estudados quatro destes métodos: Bissecção, Newton, Secante e Monte Carlo. Os três primeiros são métodos estudados na disciplina de Cálculo Numérico, normalmente ministrado nas graduações das Ciências Exatas; o objetivo deles é o cálculo de raízes de funções reais; eles trabalham através de sequências recorrentes que convergem para as raízes exatas. Uma das aplicações do quarto método é o cálculo da área sob uma curva, utilizando, para isso, números aleatórios. Em cada método estudaremos a motivação seguida de sua comprovada eficácia. Como veremos, são muitas as áreas que necessitam de métodos, como os aqui estudados, para resolução de problemas; as engenharias, a Química, a Física e a Biologia são apenas algumas delas. A fi m de demonstrar a importância dos quatro métodos apresentaremos situações em que podem ser aplicados. Demonstrada a necessidade e e ficiência de todos os métodos, podemos compreendê-los como uma excelente alternativa para resolução de problemas na área da matemática aplicada, principalmente quando os métodos convencionais de resolução não conseguem desempenhar papel satisfatório. Devemos enfatizar, ainda, a importância do uso de softwares e programas computacionais para melhor desempenho dos métodos.

5章（本文31～40ページ）と元となった論文の著作権はIEEEに属するため、規約に従い、本文79ページにおいて出典を示すともに、コピーライト表記を付している。本文39、40ページの全ての図の著作権は、IEEEに属する。このため、これら全ての図においてコピーライト表記を付している。 / 京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第19858号 / 情博第609号 / 新制||情||106(附属図書館) / 32894 / 京都大学大学院情報学研究科数理工学専攻 / (主査)教授 中村 佳正, 教授 梅野 健, 教授 中島 浩 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Nous sommes fonctionnellement et structurellement asymétriques. La symétrie parfaite que nous expérimentons subjectivement en observant l'espace qui nous entoure est, dans une certaine mesure, une illusion. La cognition visuospatiale, comme indiqué par les tâches de bissection de lignes, est généralement biaisée à gauche chez les sujets sains et à droite suite à des lésions de l'hémisphère droit causant la Négligence Spatiale Unilatérale (NSU). Ces biais peuvent être modulés et l'adaptation prismatique (AP) a démontré sa capacité à réduire les symptômes de la NSU et à induire des comportements similaires à la NSU chez les individus sains. La question de savoir comment ce type d'adaptation sensorimotrice module la cognition spatiale est encore débattue. L'objectif de cette thèse était d'utiliser des approches comportementales et physiologiques, pour examiner les mécanismes sous-jacents des effets de l'AP sur la cognition visuospatiale d'individus sains. Dans une 1ère étude comportementale, nous avons observé la présence d'une dynamique temporelle des effets survenant après l'AP. Suite à ce premier résultat, nous avons testé sur une période de temps plus longue les effets faisant suite à l'AP déviant la vision vers la droite ou la gauche, et nous avons dévoilé, dans une 2ème étude, des dynamiques temporelles différentes en fonction de la direction de l'AP. Dans une 3ème étude, nous avons utilisé la stimulation magnétique transcrânienne pour étudier la physiologie sous-tendant la modulation visuospatiale efficacement induite par l'AP. Les résultats de cette thèse appellent à un raffinement des modèles actuels de l'action de l'AP sur la cognition visuospatiale / We are functionally and structurally asymmetric. The perfect symmetry we subjectively experience through vision of the space around us is, to some extent, an illusion. Visuospatial cognition, as indexed by performance on line bisection tasks, is generally biased leftward in healthy individuals and pathologically rightward after right brain damage causing unilateral spatial neglect (USN). These biases can be modulated and prism adaptation (PA) is capable of both alleviating USN symptoms and inducing a rightward shift (the so-called “neglect-like behavior”) in healthy individuals. How this type of sensorimotor adaptation modulates spatial cognition is still debated. The goal of this thesis was to use both behavioral and physiological approaches to investigate the underlying mechanisms of PA’s effects on visuospatial cognition in healthy individuals. In a first behavioral study we found the presence of a temporal dynamic in PA after-effects. Based on this first finding we tested, over a longer period of time the PA after-effects following both right and leftward PA and unveiled, with the second study, different temporal dynamics depending on PA direction. In a third study we used transcranial magnetic stimulation to investigate the physiology underlying the effective visuospatial modulation induced by PA. The results of this thesis call for a refinement of the current models of PA action on visuospatial cognition

[pt] Uma reta bissetora divide uma região convexa do plano em duas partes com áreas iguais. É natural estudar as envoltórias destas linhas bissetoras, que em geral apresentam singularidades. O caso de polígonos é particularmente interessante, pois existem diversas noções distintas de envoltórias discretas. Nesta dissertação, nós estudamos três tipos diferentes de envoltórias discretas de retas bissetoras e as conexões entre elas. / [en] A bisection line divide a convex planar region into two parts with equal areas. It is natural to study the envelope of these lines, which in general present singularities. The polygonal case is particularly interesting, since there are several different notions of a discrete envelope. In this dissertation, we study three different notions of discrete envelopes of bisection lines and the connections between them.

Path integral Monte Carlo (PIMC) is a method suitable for quantum liquid simulations at finite temperature. We present in this thesis a study of PIMC dealing with the theory and algorithms related to it, and then two applications of PIMC to current research problems of quantum fluids in the Bolzmann regime. The first part encompasses a study of the different ingredients of a PIMC code: action, sampling and physical property estimators. Particular attention has been paid to Li-Broughton's higher order approximation to the action. Regarding sampling, several collective movement methods have been derived, including the bisection algorithm, that has been thoroughly tested. We also include a study of estimators for different physical properties, such as, the energy (through the thermodynamic and virial estimators), the pair distribution function, the structure factor, and the momentum distribution. In relation to the momentum distribution, we have developed a novel algorithm for its estimation, the trail method. It surmounts some of the problems exposed by previous approaches, such as the open chain method or McMillan's algorithm.The Richardson extrapolation used within PIMC simulations, is another contribution of this thesis. Up until now, this extrapolation has not been used in this context. We present studies of the energy dependence on the number of "beads", along with the betterment provide by the Richardson extrapolation. Inasmuch as our goal is to perform research of quantum liquids at finite temperature, we have produced a library of codes, written from scratch, that implement most of the features theoretically developed. The most elaborated parts of these codes are included in some of the appendixes.The second part shows two different applications of the algorithms coded. We present results of a PIMC calculation of the momentum distribution of Ne and normal 4He at low temperatures. In the range of temperatures analysed, exchanges can be disregarded and both systems are considered Boltzmann quantum liquids. Their quantum character is well reflected in their momentum distributions witch show clear departures from the classical limit. The PIMC momentum distributions which show clear departures from the classical limit. The PIMC momentum distributions are sampled using the trail method. Kinetic energies of both systems, as a function of temperature and at a fixed density, are also reported. Finally, the solid-liquid neon phase transition along the 35 K isotherm has been characterized.While thermodynamic properties of the solid phase are well known the behaviour of some properties, such as the energy or the dessity, during the trasition presen6 some uncertainties For example, experimental data for the place diagram, which determines solid and liquid boundaries, present sizeable differences. The temperature chosen is high enough so that Bose or Fermi statistics corrections are small, although the system is strongly quantum mechanical. The results obtained show a discontinuity in the kinetic energy during the transition.