No description available.

One main goal of this thesis is to show that for every K it is possible to find K consecutive natural numbers that cannot be written as sums of three nonnegative cubes. Since it is believed that approximately 10% of all natural numbers can be written in this way, this result indicates that the sums of three cubes distribute unevenly on the real line. These sums have been studied for almost a century, in relation with Waring's problem, but the existence of ``arbitrarily long gaps'' between them was not known. We will provide two proofs for this theorem. The first is relatively elementary and is based on the observation that the sums of three cubes have a positive bias towards being cubic residues modulo primes of the form p=1+3k. Thus, our first method to find consecutive non-sums of three cubes consists in searching them among the natural numbers that are non-cubic residues modulo ``many'' primes congruent to 1 modulo 3. Our second proof is more technical: it involves the computation of the Sato-Tate distribution of the underlying cubic Fermat variety {x^3+y^3+z^3=0}, via Jacobi sums of cubic characters and equidistribution theorems for Hecke L-functions of the Eisenstein quadratic number field Q(\sqrt{-3}). The advantage of the second approach is that it provides a nearly optimal quantitative estimate for the size of gaps: if N is large, there are >>\sqrt{log N}/(log log N)^4 consecutive non-sums of three cubes that are less than N. According to probabilistic models, an optimal estimate would be of the order of log N / log log N. In this thesis we also study other gap problems, e.g. between sums of four fourth powers, and we give an application to the arithmetic of cubic and biquadratic theta series. We also provide the following additional contributions to Number Theory and Combinatorics: a derivation of cubic identities from a parameterization of the pseudo-automorphisms of binary quadratic forms; a multiplicity estimate for multiprojective Chow forms, with applications to Transcendental Number Theory; a complete solution of a problem on planar graphs with everywhere positive combinatorial curvature.

Cette thèse de doctorat aborde spécifiquement le problème de la métamorphose entre différents maillages 3D, qui peut assurer un niveau élevé de qualité pour la séquence de transition, qui devrait être aussi lisse et progressive que possible, cohérente par rapport à la géométrie et la topologie, et visuellement agréable. Les différentes étapes impliquées dans le processus de transformation sont développées dans cette thèse. Nos premières contributions concernent deux approches différentes des paramétrisations: un algorithme de mappage barycentrique basé sur la préservation des rapports de longueur et une technique de paramétrisation sphérique, exploitant la courbure Gaussien. L'évaluation expérimentale, effectuées sur des modèles 3D de formes variées, démontré une amélioration considérable en termes de distorsion maillage pour les deux méthodes. Afin d’aligner les caractéristiques des deux modèles d'entrée, nous avons considéré une technique de déformation basée sur la fonction radial CTPS C2a approprié pour déformer le mappage dans le domaine paramétrique et maintenir un mappage valide a travers le processus de mouvement. La dernière contribution consiste d’une une nouvelle méthode qui construit un pseudo metamaillage qui évite l'exécution et le suivi des intersections d’arêtes comme rencontrées dans l'état-of-the-art. En outre, notre méthode permet de réduire de manière drastique le nombre de sommets normalement nécessaires dans une structure supermesh. Le cadre générale de métamorphose a été intégré dans une application prototype de morphing qui permet à l'utilisateur d'opérer de façon interactive avec des modèles 3D et de contrôler chaque étape du processus / This Ph.D. thesis specifically deals with the issue of metamorphosis of 3D objects represented as 3D triangular meshes. The objective is to elaborate a complete 3D mesh morphing methodology which ensures high quality transition sequences, smooth and gradual, consistent with respect to both geometry and topology, and visually pleasant. Our first contributions concern the two different approaches of parameterization: a new barycentric mapping algorithm based on the preservation of the mesh length ratios, and a spherical parameterization technique, exploiting a Gaussian curvature criterion. The experimental evaluation, carried out on 3D models of various shapes, demonstrated a considerably improvement in terms of mesh distortion for both methods. In order to align the features of the two input models, we have considered a warping technique based on the CTPS C2a radial basis function suitable to deform the models embeddings in the parametric domain maintaining a valid mapping through the entire movement process. We show how this technique has to be adapted in order to warp meshes specified in the parametric domains. A final contribution consists of a novel algorithm for constructing a pseudo-metamesh that avoids the complex process of edge intersections encountered in the state-of-the-art. The obtained mesh structure is characterized by a small number of vertices and it is able to approximate both the source and target shapes. The entire mesh morphing framework has been integrated in an interactive application that allows the user to control and visualize all the stages of the morphing process

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第21737号 / 工博第4554号 / 新制||工||1710(附属図書館) / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 木村 亮, 准教授 肥後 陽介, 准教授 木元 小百合 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

Betongpelare är en vanlig del i konstruktioner inom bygg-och anläggningsbranschen och kräver normalt en stor mängd armeringsjärn för att säkerställa dess styrka och stabilitet. Vid dimensionering av slanka betongpelare ska hänsyn inte bara tas till första ordningens moment och deformationer utan även andra ordningens teori ska beaktas. För att dimensionera förandra ordningens moment beskriver Eurokoderna tre olika metoder, en generell metod samt två förenklade metoder: nominell styvhet och nominell krökning. Dimensionering kan ske förhand eller med datorprogram. FEM-Design, som är ett avancerat analysprogram, baseras på finita elementmetoden som är en numerisk analysmetod och ett av de vanligaste sätten att beräkna fysikaliska fenomen. FEM-Design kan ofta ge ett bättre och säkrare resultat då handberäkningar approximerar för att de ska vara hanterbara.I arbetet jämförs beräknad armeringsmängd mellan handberäkningar med nominell styvhet,nominell krökning samt analysprogrammet FEM-Design. Arbetet har gjorts för att undersöka skillnaden i armering och därmed kunna avgöra vilken metod som ger minst respektive mest mängd armering. Betongpelarna som undersöks är slanka och har tre olika upplagsförhållanden, varje upplag belastas med tre olika belastningsfall. Beräkningarna är utförda enligt Eurokod 2 och resultatet blev att FEM-Design gav i sju av nio fall lägst andra ordningens moment. I åtta av nio fall gav FEM-Design lägst mängd armering medan nominell styvhet gav störst andra ordningens moment och störst mängd armering i samtliga fall. Nominell krökning gav ett andra ordningens moment som var nära FEM-Designs resultat medstörsta skillnaden på 30%. Beräknad armering för nominell krökning växlade mellan attstämma överens mest med nominell styvhet och FEM-Design. Utifrån resultatet har även skillnaden i pris på armering beräknats fram där nominell styvhet är det dyraste alternativet.FEM-Design är 61% billigare än nominell styvhet medan nominell krökning är 51% billigareän nominell krökning.Jämförelsen visar att nominell krökning kan i dessa belastningsfall som har undersökts ansesvara den bästa metoden av handberäkningarna men FEM-Design anses vara den bästa av samtliga metoder i detta arbete. Slutsatsen som kunde dras var att båda de förenklade handberäkningsmetoderna överdimensionerar armeringsmängden i pelarna och därmed anses FEM-Design som det bästa alternativet. FEM-Design gav inte bara minst armering och därmed lägsta armeringskostnaden utan dimensioneringen tog också kortast tid. / When designing slender concrete columns where the second-order theories need to be considered, the Eurocodes describe three different methods; a general method and two simplified methods - nominal stiffness and nominal curvature. Designing can be done manually or with programs. FEM-design, an advanced analysis program, is based on the finiteelement method, which is a numerical analysis method and is one of the most common ways to calculate big and complex problems. FEM-Design often provides more reliable results compared to calculations done by hand, which involve approximations to make them manageable.In this study the calculated reinforcement quantities are compared with hand calculationsusing the nominal stiffness, nominal curvature and FEM-Design. The purpose is to investigatethe difference in reinforcement and determine which method requires the least amount of reinforcement. The investigated columns are slender and have three different boundary conditions, each subject to three different load cases. The calculations are performed according to the Eurocode 2. The results show that in seven out of nine cases the FEM-Design method produced the lowest second-order moments. In eight out of nine cases, FEM-Design resulted in least amount of reinforcement, while nominal stiffness resulted in the highest second-order moments and greatest amount of reinforcement in all cases. Nominal curvature generally produced second-order moments that were close to FEM-Design, the largest difference being 30%. Regarding the calculated reinforcement , nominal curvature varied in agreement with nominal stiffness and FEM-Design. The cost of reinforcement was also analyzed, with nominal stiffness being 51% more expensive than nominal curvature and 61% more expensive than FEM-Design. Nominal curvature was the preferred manual method, but FEM-Design emerged as the best overall method, offering both minimal reinforcement and shorter design time.

Lifting surfaces of unmanned aerial vehicles (UAV) are often operated in low Reynolds number (Re) ranges, wherein the transition of boundary layer from laminar-to-turbulent plays a more significant role than in high-Re aerodynamics applications. This poses a challenge for traditional computational fluid dynamics (CFD) simulations, since typical modeling approaches assume either fully laminar or fully turbulent flow. In particular, the boundary layer state must be accurately predicted to successfully determine the separation behavior which significantly influences the aerodynamic characteristics of the airfoil. Reynolds-averaged Navier-Stokes (RANS) based CFD simulations of an elliptic airfoil are performed for time-varying angles of attack, and results are used to elucidate relevant flow physics and aerodynamic data for an elliptic airfoil under realistic operating conditions. Results are also used to evaluate the performance of several different RANS-based turbulence modeling approaches for this class of flowfield.

The far-reaching successes of deep neural networks in a wide variety of learning tasks have prompted research on how model properties account for high network performance. For a specific class of models whose activation functions are piecewise linear, one such property of interest is the number of linear regions that the network generates. Such models themselves define piecewise linear functions by partitioning input space into disjoint regions and fitting a different linear function on each such piece. It would be expected that the number or configuration of such regions would describe the model’s ability to fit complicated functions. However, previous works have shown difficulty in identifying linear regions as satisfactory predictors of model success. In this thesis, the question of whether the generation of linear regions due to training encode the properties of the learning problem is explored. More specifically, it is investigated whether change in linear region density due to model fitting is related to the geometric properties of the training data. In this work, data geometry is characterized in terms of the curvature of the underlying manifold. Models with ReLU activation functions are trained on a variety of regression problems defined on artificial manifolds and the change in linear region density is recorded along trajectories in input space. Learning is performed on problems defined on curves, surfaces and for image data. Experiments are repeated as the data geometry is varied and the change in density is compared with the manifold curvature measure used. In no experimental setting, was the observed change in density found to be clearly linked with curvature. However, density was observed to increase at points of discontinuity. This suggests that linear regions can in some instances model data complexities, however, the findings presented here do not support that data curvature is encoded by the formation of linear regions. Thus, the role that linear regions play in controlling the capacity of piecewise linear networks remains open. Future research is needed to gain further insights into how data geometry and linear regions are connected. / De breda framgångar som djupa neurala nätverk har uppvisat i en mängd olika inlärningsproblem har inspirerat ny forskning med syfte att förklara vilka modellegenskaper som resulterar i högpresterande nätverk. För neurala nätverk som använder styckvis linjära aktiveringsfunktioner är en intressant egenskap att studera de linjära regioner som nätverket genererar i det vektorrum som utgör träningsdatans definitionsmängd. Nätverk med styckvis linjära aktiveringsfunktioner delar upp definitionsmängden i distinkta regioner på vilka olika linjära funktioner avbildas. Dessa nätverk avbildar själva styckvis linjära funktioner. Genom att anpassa flera skilda linjära avbildningar går det att approximera funktioner som är icke-linjära. Därför skulle man kunna förvänta sig att antalet linjära regioner som en modell genererar och hur de är fördelade i rummet kunde fungera som mått på modellens förmåga att lära sig komplicerade funktioner. Tidigare efterforskingar inom detta område har dock inte kunnat demonstrera ett samband mellan antalet eller fördelningen av linjära regioner och modellens prestanda. I den här avhandlingen undersöks det vilken roll linjära regioner spelar i att förklara en modells kapacitet och vad den lär sig. Fångar de linjära regioner som ett nätverk lär sig de underliggande egenskaperna hos träningsdatan? Mer specifikt så studeras huruvida den lokala förändringen i antalet linjära regioner efter modellträning korrelerar med träningsdatans geometri. Träningsdata genereras från syntetiska mångfalder och datageometrin beskrivs i termer av mångfaldens krökning. På dessa mångfalder definieras regressionsproblem och träning upprepas för topologier av olika form och med olika krökning. Skillnaden i antalet linjära regioner efter träning mäts längs banor i definitionsdomänen och jämförs med datans krökning. Ingen av de experiment som utfördes lyckades påvisa något tydligt samband mellan förändring i antal regioner och datans krökning. Det observerades dock att antalet linjära regioner ökar i närheten av punkter som utgör diskontinuiteter. Detta antyder att linjära regioner under vissa omständigheter kan modellera komplexitet. Således förblir rollen som linjära regioner har i att förklara modellförmåga diffus.

No description available.

No description available.

This thesis investigates some problems related to the form and shape of statistical distributions with the main focus on goodness of fit and bump hunting. A bump is a distinctive characteristic of distributional shape. A search for bumps, or bump hunting, in a probability density function (PDF) has long been an important topic in statistical research. We introduce a new definition of a bump which relies on the notion of the curvature of a planar curve. We then propose a new method for bump hunting which is based on a kernel density estimator of the unknown PDF. The method gives not only the number of bumps but also the location of their centers and base points. In quantitative risk applications, the selection of distributions that properly capture upper tail behavior is essential for accurate modeling. We study tests of distributional form, or goodness-of-fit (GoF) tests, that assess simple hypotheses, i.e., when the parameters of the hypothesized distribution are completely specified. From theoretical and practical perspectives, we analyze the limiting properties of a family of weighted Cramér-von Mises GoF statistics W2 with weight function psi(t)=1/(1-t)^beta (for beta<=2) which focus on the upper tail. We demonstrate that W2 has no limiting distribution. For this reason, we provide a normalization of W2 that leads to a non-degenerate limiting distribution. Further, we study W2 for composite hypotheses, i.e., when distributional parameters must be estimated from a sample at hand. When the hypothesized distribution is heavy-tailed, we examine the finite sample properties of W2 under the Chen-Balakrishnan transformation that reduces the original GoF test (the direct test) to a test for normality (the indirect test). In particular, we compare the statistical level and power of the pairs of direct and indirect tests. We observe that decisions made by the direct and indirect tests agree well, and in many cases they become independent as sample size grows. / Thesis / Doctor of Philosophy (PhD)