Anonymous Call Detail Records (CDR’s) from mobile phone towers provide a unique opportunity to aggregate individual location data to overall human mobility patterns. Flowminder uses this data to improve the welfare of low- and middle-income countries. The movement patterns are studied through key measurements of mobility. This thesis seeks to evaluate the estimates of key measurements obtained with mobile phone towers through simulation of continuous human movement on a linear network. Simulation is made with an agent based approach. Spatial point processes are used to distribute continuous start points of the agents on the linear network. The start point is then equipped with a mark, a path with an end point dependent on the start point. A path from the start point to the end point of an agent is modeled with a Markov Decision Process. The simulated human movement can then be captured with different types of mobile phone tower distributions realized from spatial point processes. The thesis will initially consider homogeneous Poisson and Simple Sequential Inhibition (SSI) processes on a plane and then introduce local clusters (heterogeneity) with Matérn Cluster and SSI processes. The goal of the thesis is to investigate the effects of change in mobile phone tower distribution and call frequency on the estimates of key measurements of mobility. The effects of call frequency are unclear and invite more detailed study. The results suggest that a decrease in the total number of towers generally worsens the estimates and that introducing local clusters also has a negative effect on the estimates. The presented methodology provides a flexible and new way to model continuous human movement along a linear network.

L’un des principaux objectifs de la physique des réacteurs nucléaires est de caractériser la répartition aléatoire de la population de neutrons au sein d’un réacteur. Les fluctuations de cette population sont liées à la nature stochastique des interactions des neutrons avec les noyaux fissiles du milieu : diffusion, capture stérile, ou encore émission de plusieurs neutrons lors de la fission d’un noyau. L’ensemble de ces mécanismes physiques confère une structure aléatoire branchante à la trajectoire des neutrons, alors modélisée par des marches aléatoires. Avec environs 10⁸ neutrons par centimètre cube dans un réacteur de type REP à pleine puissance en conditions stationnaires, les grandeurs physiques du système (flux, taux de réaction, énergie déposée) sont, en première approximation, bien représentées par leurs valeurs moyennes respectives. Ces observables physiques moyennes obéissent alors à l’équation de transport linéaire de Boltzmann. Au cours de ma thèse, je me suis penchée sur deux aspects du transport qui ne sont pas décrits par cette équation, et pour lesquels je me suis appuyée sur des outils issus de la théorie des marches aléatoires. Tout d’abord, grâce au formalisme de Feynman-Kac, j’ai étudié les fluctuations statistiques de la population de neutrons, et plus particulièrement le phénomène de « clustering neutronique », qui a été mis en évidence numériquement pour de faibles densités de neutrons (typiquement un réacteur au démarrage). Je me suis ensuite intéressée à différentes propriétés de la statistique d’occupation des neutrons effectuant un transport anormal (càd non-exponentiel). Ce type de transport permet de modéliser le transport dans des matériaux fortement hétérogènes et désordonnés, tel que les réacteurs à lit de boulets. L’un des aspects novateurs de ce travail est la prise en compte de la présence de bords. En effet, bien que les systèmes réels soient de taille finie, la plupart des résultats théoriques pré-existants sur ces thématiques ont été obtenus sur des systèmes de taille infinie. / One of the key goals of nuclear reactor physics is to determine the distribution of the neutron population within a reactor core. This population indeed fluctuates due to the stochastic nature of the interactions of the neutrons with the nuclei of the surrounding medium: scattering, emission of neutrons from fission events and capture by nuclear absorption. Due to these physical mechanisms, the stochastic process performed by neutrons is a branching random walk. For most applications, the neutron population considered is very large, and all physical observables related to its behaviour, such as the heat production due to fissions, are well characterised by their average values. Generally, these mean quantities are governed by the classical neutron transport equation, called linear Boltzmann equation. During my PhD, using tools from branching random walks and anomalous diffusion, I have tackled two aspects of neutron transport that cannot be approached by the linear Boltzmann equation. First, thanks to the Feynman-Kac backward formalism, I have characterised the phenomenon of “neutron clustering” that has been highlighted for low-density configuration of neutrons and results from strong fluctuations in space and time of the neutron population. Then, I focused on several properties of anomalous (non-exponential) transport, that can model neutron transport in strongly heterogeneous and disordered media, such as pebble-bed reactors. One of the novel aspects of this work is that problems are treated in the presence of boundaries. Indeed, even though real systems are finite (confined geometries), most of previously existing results were obtained for infinite systems.

Syftet med denna studie är att undersöka om barn i enlighet med FN:s konvention om barns rättigheter har möjlighet till att fritt delta i det kulturella och konstnärliga livet i förskolans utbildning samt om de utifrån läroplanen ges tid, rum och ro till eget skapande. Studien syftar till att undersöka hur barn åskådliggör sina möjligheter till eget/fritt skapande, vilket skapande som barnen anser vara viktigt och vad de väljer att framhäva samt i vilken form skapandet tar sig uttryck. Studien behandlar dessutom de handlingserbjudanden barn möter i förskolan och hur dessa påverkar skapandet. Vi har använt oss av vår nya metod digitala gåturer, där barnen själva och med stöd av en förskollärare har fått visa och berätta om sitt skapande för oss. Resultatet av studien visar att platsen är viktig för de möjligheter och handlingserbjudanden barn möter i förskolan. Vi ser även hur barns skapande tar sig många olika uttrycksformer samt att barn genom sitt aktörskap transformerar och erövrar platser och material. Slutsatsen av studien är att det är viktigt att vi noggrant ser över utformningen av förskolans miljöer samt tillgång till material för att eget/fritt skapande ska kunna uppnås. Då barnen använder många olika tillvägagångssätt i processen av sitt skapande blir det även viktigt att synliggöra och värdesätta resultatet av vad processen har lett till. Slutligen visar även studien att metoden digitala gåturer är en tillförlitlig och modern metod för att åskådliggöra barns perspektiv.

Robbins' problem is an optimal stopping problem where one seeks to minimize the expected rank of their observations among all observations. We examine random walk analogs to Robbins' problem in both discrete and continuous time. In discrete time, we consider full information and relative ranks versions of this problem. For three step walks, we give the optimal stopping rule and the expected rank for both versions. We also give asymptotic upper bounds for the expected rank in discrete time. Finally, we give upper and lower bounds for the expected rank in continuous time, and we show that the expected rank in the continuous time problem is at least as large as the normalized asymptotic expected rank in the full information discrete time version.

No description available.

Machine learning is evidenced as a research area with the main purpose of developing computational methods that are capable of learning with their previously acquired experiences. Although a large amount of machine learning techniques has been proposed and successfully applied in real systems, there are still many challenging issues, which need be addressed. In the last years, an increasing interest in techniques based on complex networks (large-scale graphs with nontrivial connection patterns) has been verified. This emergence is explained by the inherent advantages provided by the complex network representation, which is able to capture the spatial, topological and functional relations of the data. In this work, we investigate the new features and possible advantages offered by complex networks in the machine learning domain. In fact, we do show that the network-based approach really brings interesting features for supervised, semisupervised, and unsupervised learning. Specifically, we reformulate a previously proposed particle competition technique for both unsupervised and semisupervised learning using a stochastic nonlinear dynamical system. Moreover, an analytical analysis is supplied, which enables one to predict the behavior of the proposed technique. In addition to that, data reliability issues are explored in semisupervised learning. Such matter has practical importance and is found to be of little investigation in the literature. With the goal of validating these techniques for solving real problems, simulations on broadly accepted databases are conducted. Still in this work, we propose a hybrid supervised classification technique that combines both low and high orders of learning. The low level term can be implemented by any classification technique, while the high level term is realized by the extraction of features of the underlying network constructed from the input data. Thus, the former classifies the test instances by their physical features, while the latter measures the compliance of the test instances with the pattern formation of the data. Our study shows that the proposed technique not only can realize classification according to the semantic meaning of the data, but also is able to improve the performance of traditional classification techniques. Finally, it is expected that this study will contribute, in a relevant manner, to the machine learning area / Aprendizado de máquina figura-se como uma área de pesquisa que visa a desenvolver métodos computacionais capazes de aprender com a experiência. Embora uma grande quantidade de técnicas de aprendizado de máquina foi proposta e aplicada, com sucesso, em sistemas reais, existem ainda inúmeros problemas desafiantes que necessitam ser explorados. Nos últimos anos, um crescente interesse em técnicas baseadas em redes complexas (grafos de larga escala com padrões de conexão não triviais) foi verificado. Essa emergência é explicada pelas inerentes vantagens que a representação em redes complexas traz, sendo capazes de capturar as relações espaciais, topológicas e funcionais dos dados. Nesta tese, serão investigadas as possíveis vantagens oferecidas por redes complexas quando utilizadas no domínio de aprendizado de máquina. De fato, será mostrado que a abordagem por redes realmente proporciona melhorias nos aprendizados supervisionado, semissupervisionado e não supervisionado. Especificamente, será reformulada uma técnica de competição de partículas para o aprendizado não supervisionado e semissupervisionado por meio da utilização de um sistema dinâmico estocástico não linear. Em complemento, uma análise analítica de tal modelo será desenvolvida, permitindo o entendimento evolucional do modelo no tempo. Além disso, a questão de confiabilidade de dados será investigada no aprendizado semissupervisionado. Tal tópico tem importância prática e é pouco estudado na literatura. Com o objetivo de validar essas técnicas em problemas reais, simulações computacionais em bases de dados consagradas pela literatura serão conduzidas. Ainda nesse trabalho, será proposta uma técnica híbrica de classificação supervisionada que combina tanto o aprendizado de baixo como de alto nível. O termo de baixo nível pode ser implementado por qualquer técnica de classificação tradicional, enquanto que o termo de alto nível é realizado pela extração das características de uma rede construída a partir dos dados de entrada. Nesse contexto, aquele classifica as instâncias de teste segundo qualidades físicas, enquanto que esse estima a conformidade da instância de teste com a formação de padrões dos dados. Os estudos aqui desenvolvidos mostram que o método proposto pode melhorar o desempenho de técnicas tradicionais de classificação, além de permitir uma classificação de acordo com o significado semântico dos dados. Enfim, acredita-se que este estudo possa gerar contribuições relevantes para a área de aprendizado de máquina.

The work consists of two parts. In the first part which is concerned with random walks, we construct the conditional versions of a multidimensional random walk given that it does not leave the Weyl chambers of type C and of type D, respectively, in terms of a Doob h-transform. Furthermore, we prove functional limit theorems for the rescaled random walks. This is an extension of recent work by Eichelsbacher and Koenig who studied the analogous conditioning for the Weyl chamber of type A. Our proof follows recent work by Denisov and Wachtel who used martingale properties and a strong approximation of random walks by Brownian motion. Therefore, we are able to keep minimal moment assumptions. Finally, we present an alternate function that is amenable to an h-transform in the Weyl chamber of type C. In the second part which is concerned with Brownian motion, we examine the non-exit probability of a multidimensional Brownian motion from a growing truncated Weyl chamber. Different regimes are identified according to the growth speed, ranging from polynomial decay over stretched-exponential to exponential decay. Furthermore we derive associated large deviation principles for the empirical measure of the properly rescaled and transformed Brownian motion as the dimension grows to infinity. Our main tool is an explicit eigenvalue expansion for the transition probabilities before exiting the truncated Weyl chamber.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This paper has as its main purpose to highlight some mathematical concepts, which are behind the ranking given by a research made on the website mostly used in the world: Google. At the beginning, we briefly approached some High School’s concepts, such as: Matrices, Linear Systems and Probability. After that, we presented some basic notions related to Directed Graphs and Markov Chains of Discrete Time. From this last one, we gave more emphasis to the Steady State Vector because it ensures foreknowledge results from long-term. These concepts are extremely important to our paper, because they will be used to explain the involvement of Mathematic behind the web search “Google”. Then, we tried to detail the ranking operation of the search pages on Google, i.e., how the results of a research are classified, determining which results are presented in a sequential way in order of relevance. Finally we obtained “PageRank”, an algorithm which creates what we call Google’s Matrices and ranks the pages of a search. We finished making a brief comment about the historical arising of the web searches, from their founders to the rise and hegemony of Google. / O presente trabalho tem como objetivo destacar alguns conceitos matemáticos que estão por trás do ranqueamento dado por uma pesquisa feita no site de busca mais usados do mundo, o “Google”. Inicialmente abordamos de forma breve alguns conteúdos da matemática do ensino médio, a exemplo de: matrizes, sistemas lineares, probabilidades. Em seguida são introduzidas noções básicas de grafos dirigidos e cadeias de Markov de tempo discreto; essa última, é dada uma ênfase ao vetor estado estacionário, por ele garantir resultados de previsão de longo prazo. Esses conceitos são de grande importância em nosso trabalho, pois serão usados para explicar o envolvimento da matemática por trás do site de buscas “Google”. Na sequência, buscamos detalhar o funcionamento do ranqueamento das páginas de uma busca no “Google”, isto é, como são classificados os resultados de uma pesquisa, determinando quais resultados serão apresentados de modo sequencial em ordem de relevância. Finalmente, chegamos na obtenção do “PageRank”, algoritmo que gera a chamada Matriz do Google e ranqueia as páginas de uma busca. Encerramos com um breve histórico do surgimento dos sites de buscas, desde os seus fundadores até a ascensão e hegemonia do Google.

Mon travail de thèse est composé de deux parties bien distinctes, la première partie est consacrée à l’analysestochastique en temps discret des marches aléatoires obtuses quant à la deuxième partie, elle est liée aux probabili-tés libres. Dans la première partie, on donne une construction des intégrales stochastiques itérées par rapport à unefamille de martingales normales d-dimentionelles. Celle-ci permet d’étudier la propriété de représentation chaotiqueen temps discret et mène à une construction des opérateurs gradient et divergence sur les chaos de Wiener correspon-dant. [...] d’une EDP non linéaire alors que la deuxième est de nature combinatoire.Dans un second temps, on a revisité la description de la mesure spectrale de la partie radiale du mouvement Browniensur Gl(d,C) quand d ! +¥. Biane a démontré que cette mesure est absolument continue par rapport à la mesurede Lebesgue et que son support est compact dans R+. Notre contribution consiste à redémontrer le résultat de Bianeen partant d’une représentation intégrale de la suite des moments sur une courbe de Jordon autour de l’origine etmoyennant des outils simples de l’analyse réelle et complexe. / My PhD work is composed of two parts, the first part is dedicated to the discrete-time stochastic analysis for obtuse random walks as to the second part, it is linked to free probability. In the first part, we present a construction of the stochastic integral of predictable square-integrable processes and the associated multiple stochastic integrals ofsymmetric functions on Nn (n_1), with respect to a normal martingale.[...] In a second step, we revisited thedescription of the marginal distribution of the Brownian motion on the large-size complex linear group. Precisely, let (Z(d)t )t_0 be a Brownian motion on GL(d,C) and consider nt the limit as d !¥ of the distribution of (Z(d)t/d)⋆Z(d)t/d with respect to E×tr.

Machine learning is evidenced as a research area with the main purpose of developing computational methods that are capable of learning with their previously acquired experiences. Although a large amount of machine learning techniques has been proposed and successfully applied in real systems, there are still many challenging issues, which need be addressed. In the last years, an increasing interest in techniques based on complex networks (large-scale graphs with nontrivial connection patterns) has been verified. This emergence is explained by the inherent advantages provided by the complex network representation, which is able to capture the spatial, topological and functional relations of the data. In this work, we investigate the new features and possible advantages offered by complex networks in the machine learning domain. In fact, we do show that the network-based approach really brings interesting features for supervised, semisupervised, and unsupervised learning. Specifically, we reformulate a previously proposed particle competition technique for both unsupervised and semisupervised learning using a stochastic nonlinear dynamical system. Moreover, an analytical analysis is supplied, which enables one to predict the behavior of the proposed technique. In addition to that, data reliability issues are explored in semisupervised learning. Such matter has practical importance and is found to be of little investigation in the literature. With the goal of validating these techniques for solving real problems, simulations on broadly accepted databases are conducted. Still in this work, we propose a hybrid supervised classification technique that combines both low and high orders of learning. The low level term can be implemented by any classification technique, while the high level term is realized by the extraction of features of the underlying network constructed from the input data. Thus, the former classifies the test instances by their physical features, while the latter measures the compliance of the test instances with the pattern formation of the data. Our study shows that the proposed technique not only can realize classification according to the semantic meaning of the data, but also is able to improve the performance of traditional classification techniques. Finally, it is expected that this study will contribute, in a relevant manner, to the machine learning area / Aprendizado de máquina figura-se como uma área de pesquisa que visa a desenvolver métodos computacionais capazes de aprender com a experiência. Embora uma grande quantidade de técnicas de aprendizado de máquina foi proposta e aplicada, com sucesso, em sistemas reais, existem ainda inúmeros problemas desafiantes que necessitam ser explorados. Nos últimos anos, um crescente interesse em técnicas baseadas em redes complexas (grafos de larga escala com padrões de conexão não triviais) foi verificado. Essa emergência é explicada pelas inerentes vantagens que a representação em redes complexas traz, sendo capazes de capturar as relações espaciais, topológicas e funcionais dos dados. Nesta tese, serão investigadas as possíveis vantagens oferecidas por redes complexas quando utilizadas no domínio de aprendizado de máquina. De fato, será mostrado que a abordagem por redes realmente proporciona melhorias nos aprendizados supervisionado, semissupervisionado e não supervisionado. Especificamente, será reformulada uma técnica de competição de partículas para o aprendizado não supervisionado e semissupervisionado por meio da utilização de um sistema dinâmico estocástico não linear. Em complemento, uma análise analítica de tal modelo será desenvolvida, permitindo o entendimento evolucional do modelo no tempo. Além disso, a questão de confiabilidade de dados será investigada no aprendizado semissupervisionado. Tal tópico tem importância prática e é pouco estudado na literatura. Com o objetivo de validar essas técnicas em problemas reais, simulações computacionais em bases de dados consagradas pela literatura serão conduzidas. Ainda nesse trabalho, será proposta uma técnica híbrica de classificação supervisionada que combina tanto o aprendizado de baixo como de alto nível. O termo de baixo nível pode ser implementado por qualquer técnica de classificação tradicional, enquanto que o termo de alto nível é realizado pela extração das características de uma rede construída a partir dos dados de entrada. Nesse contexto, aquele classifica as instâncias de teste segundo qualidades físicas, enquanto que esse estima a conformidade da instância de teste com a formação de padrões dos dados. Os estudos aqui desenvolvidos mostram que o método proposto pode melhorar o desempenho de técnicas tradicionais de classificação, além de permitir uma classificação de acordo com o significado semântico dos dados. Enfim, acredita-se que este estudo possa gerar contribuições relevantes para a área de aprendizado de máquina.