Na natureza e na ciência, dados e informações que desviam significativamente da média frequentemente possuem grande relevância. Esses dados são usualmente denominados na literatura como outliers. A identificação de outliers é importante em muitas aplicações reais, tais como detecção de fraudes, diagnóstico de falhas, e monitoramento de condições médicas. Nos últimos anos tem-se testemunhado um grande interesse na área de Redes Complexas. Redes complexas são grafos de grande escala que possuem padrões de conexão não trivial, mostrando-se uma poderosa maneira de representação e abstração de dados. Embora um grande montante de resultados tenham sido reportados nesta área de pesquisa, pouco tem sido explorado acerca de detecção de outliers em redes complexas. Considerando-se a dinâmica de uma caminhada aleatória, foram propostos neste trabalho uma medida de distância e um método de ranqueamento de outliers. Através desta técnica, é possível detectar como outlier não somente nós periféricos, mas também nós centrais (hubs), depedendo da estrutura da rede. Também foi identificado que existem características bem definidas entre os nós outliers, relacionadas a funcionalidade dos mesmos para a rede. Além disso, foi descoberto que nós outliers têm papel importante para a rotulação a priori na tarefa de detecção de comunidades semi-supervisionada. Isto porque os nós centrais são bons difusores de informação e os nós periféricos encontram-se em regiões de borda de comunidade. Baseado nessa observação, foi proposto um método de detecção de comunidades semi-supervisionado. Os resultados de simulações mostram que essa abordagem é promissora / In nature and science, information and data that deviate significantly from the average value often have great relevance. These data are often called in literature as outliers. Outlier identification is important in many real applications, such as fraud detection, fault diagnosis, monitoring of medical conditions. In recent years, it has been witnessed a great interest in the area of Complex Networks. Complex networks are large-scale graphs with non-trivial connection patterns, proving to be a powerful way of data representation and abstraction. Although a large amount of results have been reported in this research area, little has been explored about the outlier detection in complex networks. Considering the dynamics of a random walk, we proposed in this paper a distance measure and a outlier ranking method. By using this technique, we can detect not only peripheral nodes, but also central nodes (hubs) as outliers, depending on the network structure. We also identified that there are well defined relationship between the outlier nodes and the functionality of the same nodes for the network. Furthermore, we found that outliers play an important role to label a priori nodes in the task of semi-supervised community detection. This is because the hubs are good information disseminators and peripheral nodes are usually localized in the regions of community edges. Based on this observation, we proposed a method of semi-supervised community detection. The simulation results show that this approach is promising

On choisit un groupe algébrique G, un sous-groupe algébrique H de G ; on choisit une mesure de probabilité borélienne μ sur G. On considère alors la chaîne de Markov sur l’espace homogène X = G/H de probabilité de transition Px = μ * δx pour x ε X. Dans cette thèse, on étudie les propriétés de récurrence de ces marches aléatoires.On s’intéresse à deux types de récurrence : la récurrence presque-sûre (toute trajectoire revient presque-sûrement infiniment souvent dans un compact) et la récurrence en loi (il existe une mesure de probabilité μ stationnaire sur X .On s’intéresse également aux éventuelles propriétés de transience presque-sûre (toute trajectoire quitte presque-sûrement définitivement tout compact).On construira d’abord un exemple où on n’a ni récurrence presque-sûre en tout point, ni transience presque-sûre en tout point. On montrera ensuite un critère de récurrence presque-sûre dans le cas où G est un groupe de Lie semi-simple ; on a en fait dans ce cas une dichotomie : soit tous les points sont récurrents,soit tous les points sont transients.Dans le cas où G est le groupe affine GL(d,ℝ) α ℝd,on donnera un critère de récurrence en loi sur les Grassmanniennes affines, et, dans un dernier chapitre, on donnera quelques résultats partiels d'un projet en cours,permettant de donner des résultats pour le groupe SO(p, p+1) α ℝ2p+1. / Choose an algebraic group G, and an algebraic subgroup H. Choose a Borel probability measure μ on G. Consider the Markov chain on the G-space X = G/H with transition probability Px = μ * δx for x ε X.The point of this dissertation is the study of the recurrence properties of such a random walk.We consider two types of recurrence : almost-certain recurrence (i.e. almost-every trajectory enters some compact set infinitely often) and the associated almost-certain transience (where almost-every trajectory eventually leaves every compact set) and recurrence in law (i.e. there exists a μ stationary probability measure on X).First, we show that, in general, there is no dichotomy between almost-certain recurrence and transience by constructing an example with both almost-certainly recurrent and almost-certainly transient points.We then prove a criterion for almost-certain recurrence when G is a semi-simple Lie group and X is a G-space. In fact, in this case, we have a dichotomy where either every point of X is almost-certainly recurrent, or every point of X is almost certainly transient.When G is the affine group GL(d,ℝ) α ℝd, we give a criterion for recurrence in law on the affine Grassmannians.In the final chapter, we give some partial results from an ongoing project,which give a criterion for recurrence in law the group SO(p,p+1)α ℝ2p+1.

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第21542号 / 理博第4449号 / 新制||理||1639(附属図書館) / 京都大学大学院理学研究科数学・数理解析専攻 / (主査)教授 熊谷 隆, 准教授 福島 竜輝, 教授 牧野 和久 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DGAM

There has been an explosion of network data in the physical, chemical, biological, computational, and social sciences in the last few decades. Node embeddings, i.e., Euclidean-space representations of nodes in a network, make it possible to apply to network data, tools and algorithms from multivariate statistics and machine learning that were developed for Euclidean-space data. Random walk node embeddings are a class of recently developed node embedding techniques where the vector representations are learned by optimizing objective functions involving skip-bigram statistics computed from random walks on the network. They have been applied to many supervised learning problems such as link prediction and node classification and have demonstrated state-of-the-art performance. Yet, their properties remain poorly understood. This dissertation studies random walk based node embeddings in an unsupervised setting within the context of capturing hidden block structure in the network, i.e., learning node representations that reflect their patterns of adjacencies to other nodes. This doctoral research (i) Develops VEC, a random walk based unsupervised node embedding algorithm, and a series of relaxations, and experimentally validates their performance for the community detection problem under the Stochastic Block Model (SBM). (ii) Characterizes the ergodic limits of the embedding objectives to create non-randomized versions. (iii) Analyzes the embeddings for expected SBM networks and establishes certain concentration properties of the limiting ergodic objective in the large network asymptotic regime. Comprehensive experimental results on real world and SBM random networks are presented to illustrate and compare the distributional and block-structure properties of node embeddings generated by VEC and related algorithms. As a step towards theoretical understanding, it is proved that for the variants of VEC with ergodic limits and convex relaxations, the embedding Grammian of the expected network of a two-community SBM has rank at most 2. Further experiments reveal that these extensions yield embeddings whose distribution is Gaussian-like, centered at the node embeddings of the expected network within each community, and concentrate in the linear degree-scaling regime as the number of nodes increases. / 2023-09-24T00:00:00Z

Problembakgrund & problemdiskussion: Att som professionell investerare lyckas överavkasta sitt jämförelseindex år efter år är lättare sagt än gjort. Oavsett vilken strategi en förvaltare använder sig av kan det vara svårt att kontinuerligt lyckas prestera bättre än marknaden. Författarna av denna studie har blivit inspirerade av anomalin PEAD, postearnings-announcement-drift. Den visar att om ett bolag överraskar positivt (negativt) har aktiekursen en tendens att stiga (falla) ett tag efter att den nya informationen har presenterats.   Problemformulering: Är det möjligt att kombinera indexplacering med PAD i samband med rapporter för att nå överavkastning gentemot OMXS30-index?  Vilken tidsperiod av tre, fem eller tio dagar är det i genomsnitt mest lukrativt att hålla aktien under rådande strategi? Syfte: Studiens huvudsyfte är att se om en kombination av aktiv förvaltning grundat på PEAD tillsammans med en passiv förvaltning i en börsfond som följer OMXS30, skulle lyckas överavkasta mot OMXS30. Studiens delsyfte är att se vilken av tidsperioderna tre, fem eller tio dagar som kan generera största avkastning. Författarna vill med denna studie kunna ge professionella investerare en stabil strategi som över tid visar sig överavkasta jämförelsebart index på ett kostnadseffektivt sätt och utan att ta allt för stor risk.   Teori: Studien behandlar teorier kopplat till Post Earnings Announcement Drift (PEAD) som sedan sätts i kontext med en genomgång av de etablerade teorierna om den effektiva marknadshypotesen och random walk. Som i sin utformning motsätter sig att en strategi byggd på PEAD ska kunna bringa överavkastning.    Metod: Studien använder sig av en kvantitativ metod med en deduktiv ansats. Aktiedata är inhämtad från bolag som ingått i OMXS30 under perioden 2010–2019. Studien har testat om strategin har överavkastat gentemot studiens utvalda referensindex.    Empiri/analys: Studiens resultat har ett empiriskt stöd för att PAD3 har en signifikant överavkastning gentemot OMXS30. Medan PAD5 och PAD10 visade sig inte ha en statistiskt säkerställd överavkastning mot OMXS30. Författarna har vidare funnit intressanta resultat som starkt pekar mot att strategin presterar starkt under en negativ marknad.   Slutsats: Strategin visade sig ha starka bevis att motsäga random walk då ett återupprepande prismönster gick att finna i studiens resultat. Trots att majoriteten av resultaten av de statistiska testerna inte var signifikanta ställer sig författarna även tveksamma till den effektiva marknadshypotesen. Strategin med den kortaste aktiva förvaltningen överavkastade index med över 150% vilket bör kunna ses som ett tecken på att marknaden inte till fullo lyckas prisa in den nya information som presenteras i samband med då bolagen släpper rapporter.

The purpose of this study is to examine whether the returns on the Stockholm Stock Exchange and its different sectors is mean reverting during the period 2003–2019. In addition to the examination of the entire period, the study also examines the periods before, during and after the global financial crisis. Daily, weekly and monthly data is used in combination with three different statistical methods in the form of ADF-tests, KPSS-tests and GPH-tests. The previous research conducted within the area yield different results but the tendency to find support for mean reversion increases during periods of economic uncertainty. The main standpoint used in previous literature and also this study is based on Famas (1970) publication about the efficient market hypothesis (EMH).  The results from the entire period are in line with the EMH where no robust indications for mean reversion is found neither for the specific sectors or the index itself. However, when the different periods are examined, in the period following the global financial crisis the healthcare and real estate sector show signs of mean reversion. In summary, the results show more support for mean reversion when the sectors are examined in relation to the entire Stockholm Stock Exchange. The conclusion based on these results is that an investor, in contrast to the EMH, can use the serial correlation in returns for these sectors as an indicator for when to buy and sell assets during periods as the one following the global financial crisis.

Using the ARFIMA-FIGARCH model, this dissertation examines the efficiency of Botswana's bond market. It focuses on the properties of the return and volatility of the Fleming Asset Bond Index (the main aggregate fixed income benchmark index in Botswana) over the period September 2009 to May 2019. The weak-form version of efficient market hypothesis (EMH) is used as a criterion to investigate the existence of long memory in both bond returns and volatility. The results of our study indicate that the Botswana bond market data follow, to a great extent, the long-range dependence which negates the precepts of the efficient market hypothesis. Furthermore, policy reforms intended to stimulate bond market reform and related efficiency gains appear not to have produced the desired outcomes as the existence of long memory is found across all sample periods. Further remedial policies are suggested to enhance market dynamism.

In the course of evolution, genomes diverge from their ancestors either via global mutations and by rearrangement of their chromosomal segments, or through local mutations within their genes. In this thesis (Chapters: 2, 3 and 4) we analyze the evolution of genomes based on different rearrangement operations including: in Chapter 2 both restricted and unrestricted double-cut-and-join (DCJ) operations, in Chapter 3 both internal and general reversal and translocation (IRT and HP, respectively) operations, and in Chapter 4 translocation, weighted reversal (WR) and maximum length reversal (MLR) operations. Based on the rearrangement operation chosen we can model the evolution of genomes as a discrete or continuous-time Markov chain process on the space of signed genomes. For each model of evolution, we study the stochastic process by investigating the time up to which the difference between the number of operations along the evolutionary trajectory and the edit distance of the genome from its ancestor is negligible, as soon as these two values starts diverging drastically from one another we say the process escapes from parsimony. One of the major parameters in the known edit distance formulas between any two genomes (such as reversal, DCJ, IRT, HP and translocation) is the number of cycles in their breakpoint graph. For DCJ, IRT and HP models by adopting the method elaborated by Berestycki and Durret, we estimate the number of cycles in the breakpoint graph of the genome at time t and its ancestor by the number of tree components of the random graph constructed from the model of evolution at time t, which is an Erdös-Rényi. We also proved that for each of the DCJ, IRT and HP models of evolution, the process on a genome of size n is bound to its parsimonious estimate up to t ≈ n/2 steps. Since the random graph constructed from the models of evolution for the translocation, WR and MLR processes are not Erdös-Rényi, the proofs of their parsimony- bound require more advanced mathematical tools, however our simulation shows for the translocation, two types of WR, and MLR (except for reversals with very short maximum length) models, the escape from parsimony do not occur before n/2 steps, where n is the number of genes in the genome. A basic result in this field is due to Berestycki and Durrett, from 2006, who found that a random transposition (pairwise exchange of the elements in the corresponding permutation of the genome) evolves along its parsimonious path of evolution up to n/2 steps, where n is the number of the genes. Although, this transposition model is applicable solely for evolution of a unichromosomal ancestor which remains unichromosomal at each step t of the process; however for the DCJ, IRT, HP and translocation models the genomes are multichromosomal which increases the difficulty of the problem at hand. The models studied in Chapters 2 - 4 are all based on signed permutation representations of genomes, where each "gene" occurs exactly once, with either positive or negative polarity. The same genes occur in all the genomes being considered. There is no distinction between the same gene in two different genomes. In Chapter 5 we generalize our representation to genes that may have several copies of a gene, which differ only by a few point mutations. This leads to problems of identifying copies in two genomes that are primary orthologs, under the assumptions of differentials in point mutation rate. We provide algorithms, software and test examples.

In a stopping rule problem, a real-time player decides to stop or continue at stage n based on the observations up to that stage, but in a k-step look-ahead stopping rule problem, we suppose the player knows k steps ahead. The aim of this Ph.D. dissertation is to study this type of prophet problems for simple random walk, determine the optimal stopping rule and calculate the expected return for them. The optimal one-step look-ahead stopping rule for a finite simple random walk is determined in this work. We also study two infinite horizon stopping rule problems, sum with negative drift problems and discounted sum problems. The optimal one, two and three-step look-ahead stopping rules are introduced for the sum with negative drift problem for simple random walk. We also compare the maximum expected returns and calculate the upper bound for the advantage of the prophet over the decision maker. The last chapter of this dissertation concentrates on the discounted sum problem for simple random walk. Optimal one-step look-ahead stopping rule is defined and lastly we compare the optimal expected return for one-step look-ahead prophet with a real-time decision maker.

Functioning of the living cell is controlled by a complex network of interactions among genes, proteins, and other molecules. A major goal of systems biology is to understand and explain the mechanisms by which these interactions govern the cell's response to various conditions. Molecular interaction networks have proven to be a powerful representation for studying cellular behavior. Numerous algorithms have been developed to unravel the complexity of these networks. Our work addresses the drawbacks of existing techniques. This thesis includes three related research efforts that introduce network-based approaches to bridge current methodological gaps in systems biology. i. Functional enrichment methods provide a summary of biological functions that are overrepresented in an interesting collection of genes (e.g., highly differentially expressed genes between a diseased cell and a healthy cell). Standard functional enrichment algorithms ignore the known interactions among proteins. We propose a novel network-based approach to functional enrichment that explicitly accounts for these underlying molecular interactions. Through this work, we close the gap between set-based functional enrichment and topological analysis of molecular interaction networks. ii. Many techniques have been developed to compute the response network of a cell. A recent trend in this area is to compute response networks of small size, with the rationale that only part of a pathway is often changed by disease and that interpreting small subnetworks is easier than interpreting larger ones. However, these methods may not uncover the spectrum of pathways perturbed in a particular experiment or disease. To avoid these difficulties, we propose to use algorithms that reconcile case-control DNA microarray data with a molecular interaction network by modifying per-gene differential expression p-values such that two genes connected by an interaction show similar changes in their gene expression values. iii. Top-down analyses in systems biology can automatically find correlations among genes and proteins in large-scale datasets. However, it is often difficult to design experiments from these results. In contrast, bottom-up approaches painstakingly craft detailed models of cellular processes. However, developing the models is a manual process that can take many years. These approaches have largely been developed independently. We present Linker, an efficient and automated data-driven method that analyzes molecular interactomes. Linker combines teleporting random walks and k-shortest path computations to discover connections from a set of source proteins to a set of target proteins. We demonstrate the efficacy of Linker through two applications: proposing extensions to an existing model of cell cycle regulation in budding yeast and automated reconstruction of human signaling pathways. Linker achieves superior precision and recall compared to state-of-the-art algorithms from the literature. / Ph. D.