This thesis is devoted to the study of holomorphic Poisson structures and Lie algebroids, and their relationship with differential equations, singularity theory and noncommutative algebra. After reviewing and developing the basic theory of Lie algebroids in the framework of complex analytic and algebraic geometry, we focus on Lie algebroids over complex curves and their application to the study of meromorphic connections. We give concrete constructions of the corresponding Lie groupoids, using blowups and the uniformization theorem. These groupoids are complex surfaces that serve as the natural domains of definition for the fundamental solutions of ordinary differential equations with singularities. We explore the relationship between the convergent Taylor expansions of these fundamental solutions and the divergent asymptotic series that arise when one attempts to solve an ordinary differential equation at an irregular singular point. We then turn our attention to Poisson geometry. After discussing the basic structure of Poisson brackets and Poisson modules on analytic spaces, we study the geometry of the degeneracy loci---where the dimension of the symplectic leaves drops. We explain that Poisson structures have natural residues along their degeneracy loci, analogous to the Poincar\'e residue of a meromorphic volume form. We discuss the local structure of degeneracy loci that have small codimensions, and place strong constraints on the singularities of the degeneracy hypersurfaces of log symplectic manifolds. We use these results to give new evidence for a conjecture of Bondal. Finally, we discuss the problem of quantization in noncommutative projective geometry. Using Cerveau and Lins Neto's classification of degree-two foliations of projective space, we give normal forms for unimodular quadratic Poisson structures in four dimensions, and describe the quantizations of these Poisson structures to noncommutative graded algebras. As a result, we obtain a (conjecturally complete) list of families of quantum deformations of projective three-space. Among these algebras is an ``exceptional'' one, associated with a twisted cubic curve. This algebra has a number of remarkable properties: for example, it supports a family of bimodules that serve as quantum analogues of the classical Schwarzenberger bundles.

abstract: Transfer learning refers to statistical machine learning methods that integrate the knowledge of one domain (source domain) and the data of another domain (target domain) in an appropriate way, in order to develop a model for the target domain that is better than a model using the data of the target domain alone. Transfer learning emerged because classic machine learning, when used to model different domains, has to take on one of two mechanical approaches. That is, it will either assume the data distributions of the different domains to be the same and thereby developing one model that fits all, or develop one model for each domain independently. Transfer learning, on the other hand, aims to mitigate the limitations of the two approaches by accounting for both the similarity and specificity of related domains. The objective of my dissertation research is to develop new transfer learning methods and demonstrate the utility of the methods in real-world applications. Specifically, in my methodological development, I focus on two different transfer learning scenarios: spatial transfer learning across different domains and temporal transfer learning along time in the same domain. Furthermore, I apply the proposed spatial transfer learning approach to modeling of degenerate biological systems.Degeneracy is a well-known characteristic, widely-existing in many biological systems, and contributes to the heterogeneity, complexity, and robustness of biological systems. In particular, I study the application of one degenerate biological system which is to use transcription factor (TF) binding sites to predict gene expression across multiple cell lines. Also, I apply the proposed temporal transfer learning approach to change detection of dynamic network data. Change detection is a classic research area in Statistical Process Control (SPC), but change detection in network data has been limited studied. I integrate the temporal transfer learning method called the Network State Space Model (NSSM) and SPC and formulate the problem of change detection from dynamic networks into a covariance monitoring problem. I demonstrate the performance of the NSSM in change detection of dynamic social networks. / Dissertation/Thesis / Doctoral Dissertation Industrial Engineering 2015

The problem of epipolar geometry estimation together with correspondence establishment in case of wide baseline and large scale changes and rotation has been addressed in this work. This work deals with cases that are heavily noised by outliers. The jump diffusion MCMC method has been employed to search for the non-degenerate epipolar geometry with the highest probabilistic support of putative correspondences. At the same time, inliers in the putative set are also identified. The jump steps involve large movements guided by a distribution of similarity based priors while diffusion steps are small movements guided by a distribution of likelihoods given by the Joint Feature Distribution (JFD). The 'best so far' samples are accepted in accordance to Metropolis-Hastings method. The diffusion steps are carried out by sampling conditioned on the 'best so far', making it local to the 'best so far' sample, while jump steps remain unconditioned and span across the correspondence and motion space according to a similarity based proposal distribution making large movements. We advance the theory in three novel ways. First, a similarity based prior proposal distribution which guide jump steps. Second, JFD based likelihoods which guide diffusion steps allowing more focused correspondence establishment while searching for epipolar geometry. Third, a measure of degeneracy that allows to rule out degenerate configurations. The jump diffusion framework thus defined allows handling over 90% outliers even in cases where the number of inliers is very few. Practically, the advancement lies in higher precision and accuracy that has been detailed in this work by comparisons. In this work, BLOGS is compared with LO-RANSAC, NAPSAC, MAPSAC and BEEM algorithm, which are the current state of the art competing methods, on a dataset that has significantly more change in baseline, rotation, and scale than those used in the state of the art. Performance of these algorithms and BLOGS are quantitatively benchmark for a comparison by estimating the error in the epipolar geometry given by root mean Sampson's distance from manually specified corresponding point pairs which serve as a ground truth. Not just is BLOGS able to tolerate very high outlier rates, but also gives result of similar quality in 10 times lesser number of iterations than the most competitive among the compared algorithms.

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第17973号 / 理博第3917号 / 新制||理||1565(附属図書館) / 80817 / 京都大学大学院理学研究科物理学・宇宙物理学専攻 / (主査)教授 高橋 義朗, 教授 田中 耕一郎, 教授 石田 憲二 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DGAM

L’étude des propriétés fractales des séries biologiques fait l’objet d’un intérêt croissant. Néanmoins la littérature met en évidence une ambiguïté quand à l’explication causale de la présence de ces séries temporelles ne permettant pas de distinguer entre l’adaptation effective réalisée par un sujet ou ses capacités d’adaptabilité globales. La présente thèse a pour objectif de décorréler ces deux notions, notamment en liant le niveau comportemental au niveau cérébral. Notre première étude a permise de mettre en évidence que les propriétés mono-fractales pourraient refléter l’adaptabilité des sujets tandis que les propriétés multifractales seraient liées à l’adaptation effective réalisée au cours de la tâche. La seconde étude à mise en évidence une corrélation entre les propriétés multifractales et le nombre de réseaux cérébraux mis en oeuvre au cours de la tâche, reflétant l’adaptation effective aux contraintes expérimentales imposées. Les résultats de ces travaux de thèse nous ont permis de mieux comprendre la signification fonctionnelle des analyses fractales en terme d’adaptation effective et d’adaptabilité. / The study of fractal properties in biological time series is of increasing interest. Nevertheless, the literature highlights an ambiguity on the causal explanation of the presence of these time series which does not make it possible to distinguish between the effective adaptation made by a subject or his overall adaptability capacities. The aim of this dissertation is to decorrelate these two notions, notably by linking the behavioral level to the cerebral level. Our first study allowed to highlight that the mono-fractal properties could reflect the adaptability of the subjects whereas the multifractal properties would be related to the effective adaptation carried out during the task. The second study showed a correlation between the multifractal properties and the number of brain networks implemented during the task, reflecting the effective adaptation to the experimental constraints imposed. The results of this work have allowed us to better understand the functional meaning of fractal analyzes in terms of effective adaptation and adaptability.

O caso finito de teorias topológicas de 1-gauge, quando nenhuma simetria global está presente, é bastante bem compreendido e classificado. Nos últimos anos, as tentativas de generalizar as teorias de 1-gauge através das chamadas teorias de 2-gauge abriram a porta para novos modelos interessantes e novas fases topológicas, as quais não são descritas pelos esquemas de classificação anteriores. Nesta tese, vamos além da construção de 2-gauge, e consideramos uma classe de modelos que vivem em maiores dimensões. Esses modelos estão inseridos em uma estrutura de complexos de cadeia de grupos abelianos, forçando a generalizar o conceito usual de configurações de gauge. A vantagem de tal abordagem é que, a ordem topológica fica manifestamente explcita. Isto é feito em ter- mos de uma cohomologia com coeficientes em um complexo de cadeia finita. Além disso, mostramos que a degenerescência do estado fundamental suporta um conjunto conve- niente de números quânticos que indexam os estados e que, além, foram completamente caracterizados. Consequentemente, nós também mostramos que muitos dos exemplos abelianos de teorias de 1 -gauge 2-gauge são recuperados como casos especiais desta construção. / The finite case of 1-gauge topological theories, when no global symmetries are present, is fairly well understood and classified. In recent years, attempts to generalize the latter situation through the so called 2-gauge theories have opened the door to interesting new models and new topological phases, not described by the previous schemes of classifica- tion. In this paper we go even beyond the 2-gauge construction by considering a class of models that live in arbitrary higher dimensions. These models are embedded in a structure of chain complexes of abelian groups, forcing to generalize the usual notion of gauge configurations. The advantage of such an approach is that, the topological order is explicitly manifest when the ground state space of these models is described. This is done in terms of a cohomology with coefficients in a finite chain complex. Furthermore, we show that the ground state degeneracy underpins a convenient set of quantum num- bers that label the states and that have been completely characterized. We also show that abelian examples of 1-gauge 2-gauge theories are recovered as special cases of this construction.

O caso finito de teorias topológicas de 1-gauge, quando nenhuma simetria global está presente, é bastante bem compreendido e classificado. Nos últimos anos, as tentativas de generalizar as teorias de 1-gauge através das chamadas teorias de 2-gauge abriram a porta para novos modelos interessantes e novas fases topológicas, as quais não são descritas pelos esquemas de classificação anteriores. Nesta tese, vamos além da construção de 2-gauge, e consideramos uma classe de modelos que vivem em maiores dimensões. Esses modelos estão inseridos em uma estrutura de complexos de cadeia de grupos abelianos, forçando a generalizar o conceito usual de configurações de gauge. A vantagem de tal abordagem é que, a ordem topológica fica manifestamente explcita. Isto é feito em ter- mos de uma cohomologia com coeficientes em um complexo de cadeia finita. Além disso, mostramos que a degenerescência do estado fundamental suporta um conjunto conve- niente de números quânticos que indexam os estados e que, além, foram completamente caracterizados. Consequentemente, nós também mostramos que muitos dos exemplos abelianos de teorias de 1 -gauge 2-gauge são recuperados como casos especiais desta construção. / The finite case of 1-gauge topological theories, when no global symmetries are present, is fairly well understood and classified. In recent years, attempts to generalize the latter situation through the so called 2-gauge theories have opened the door to interesting new models and new topological phases, not described by the previous schemes of classifica- tion. In this paper we go even beyond the 2-gauge construction by considering a class of models that live in arbitrary higher dimensions. These models are embedded in a structure of chain complexes of abelian groups, forcing to generalize the usual notion of gauge configurations. The advantage of such an approach is that, the topological order is explicitly manifest when the ground state space of these models is described. This is done in terms of a cohomology with coefficients in a finite chain complex. Furthermore, we show that the ground state degeneracy underpins a convenient set of quantum num- bers that label the states and that have been completely characterized. We also show that abelian examples of 1-gauge 2-gauge theories are recovered as special cases of this construction.

Linear degeneracy of a PDE is a concept that is related to a number of interesting geometric constructions. We first take a quadratic line complex, which is a three parameter family of lines in projective space P3 specified by a single quadratic relation in the Plucker coordinates. This complex supplies us with a conformal structure in P3. With this conformal structure, we associate a three-dimensional second order quasilinear wave equation. We show that any PDE arising in this way is linearly degenerate, furthermore, any linearly degenerate PDE can be obtained by this construction. We classify Segre types of quadratic complexes for which the structure is conformally flat, as well as Segre types for which the corresponding PDE is integrable. These results were published in [1]. We then introduce the notion of characteristic integrals, discuss characteristic integrals in 3D and show that, for certain classes of second-order linearly degenerate dispersionless integrable PDEs, the corresponding characteristic integrals are parameterised by points on the Veronese variety. These results were published in [2].

Cette thèse porte sur la construction de nouveaux exemples de surfaces minimales dérivées de la famille de surfaces de Costa-Hoffman-Meeks. Il s'agit d'une famille de surfaces minimales complètes plongées avec trois bouts et genre k > 0. Soit M_k la surface de Costa_Hoffman_Meeks de genre k. Dans le chapitre 1, j'ai démontré que M_k est non dégénérée pour k > 37. J'ai donc étendu les résultats de S. Nayatani qui assuraient que la surface M_k est non dégénérée seulement pour k=1,...,37. Ce résultat permet de montrer dans les chapitres 2 et 3 l'existence de nouveaux exemples de surfaces minimales de genre g arbitraire à l'aide d'une procédure de collage d'autres surfaces déjà connues (parmi lesquelles y figure la surface M_k). Sans ceci, ces résultats ne seraient valables que pour k < 38. En particulier dans le chapitre 2, j'ai démontré l'existence, dans H^2 x R, (H^2 étant le plan hyperbolique) d'une famille de surfaces minimales plongées inspirées de M_k, pour tout k > 0. Ce résultat peut être censé un cas particulier d'un théorème générale de désingularisation de l'intersection de deux surfaces minimales annoncé par N. Kapouleas et jamais publié. Le chapitre 3 est consacré à la construction de trois familles de surfaces minimales simplement périodiques plongées dans R^3 dont le quotient a genre arbitraire. Les résultats présentés dans ce chapitre (obtenus en collaborations avec L. Hauswirth et M. Rodríguez) généralisent plusieurs anciennes constructions / This thesis is devoted to the construction of new examples of minimal surfaces derived from the family of surfaces if Costa-Hoffman-Meeks. Surfaces in this family are complete embedded with 3 ends and genus k > 0. Let M_k denote the surface of Costa-Hoffman-Meeks of genus k. In chapter 1 I showed M_k is non degenerate for k > 37. So I extended the results of S. Nayatani which insured M_k is non degenerate only for k=1,...,37. That allows to prove in chapters 2 and 3 the existence of new examples of minimal surfaces by a gluing procedure involving already known surfaces (among which figures M_k). Without it theses results would hold only for k < 38. In particular in chapter 2 I showed the existence in H^2 x R (where H^2 denotes the hyperbolic plane) of a family of surfaces inspired to M_k, for all k > 0, which are complete and embedded. This result can be considered as a particular case of a general theorem of desingularization of the intersection of two minimal surfaces announced by N. Kapouleas and never published. Chapter 3 is devoted to the construction of 3 families of singly periodic minimal surfaces, embedded in R^3, whose quotient has an arbitrary value of the genus. The results showed in this chapter (obtained in collaboration with L. Hauswirth and M. Rodríguez) generalize many previous constructions

<p>The question about how to determine the rank of a tensor has been widely studied in the literature. However the analytical methods to compute the decomposition of tensors have not been so much developed even for low-rank tensors.</p><p>In this report we present analytical methods for finding real and complex PARAFAC decompositions of 2 x 2 x 2 tensors before computing the actual rank of the tensor.</p><p>These methods are also implemented in MATLAB.</p><p>We also consider the question of how best lower-rank approximation gives rise to problems of degeneracy, and give some analytical explanations for these issues.</p>