A classificação é uma tarefa do aprendizado de máquina e mineração de dados, na qual um classificador é treinado sobre um conjunto de dados rotulados de forma que as classes de novos itens de dados possam ser preditas. Tradicionalmente, técnicas de classificação trabalham por definir fronteiras de decisão no espaço de dados considerando os atributos físicos do conjunto de treinamento e uma nova instância é classificada verificando sua posição relativa a tais fronteiras. Essa maneira de realizar a classificação, essencialmente baseada nos atributos físicos dos dados, impossibilita que as técnicas tradicionais sejam capazes de capturar relações semânticas existentes entre os dados, como, por exemplo, a formação de padrão. Por outro lado, o uso de redes complexas tem se apresentado como um caminho promissor para capturar relações espaciais, topológicas e funcionais dos dados, uma vez que a abstração da rede unifica a estrutura, a dinâmica e as funções do sistema representado. Dessa forma, o principal objetivo desta tese é o desenvolvimento de métodos e heurísticas baseadas em teorias de redes complexas para a classificação de dados. As principais contribuições envolvem os conceitos de conformidade de padrão, caracterização de importância e otimização estrutural de redes. Para a conformidade de padrão, onde medidas de redes complexas são usadas para estimar a concordância de um item de teste com a formação de padrão dos dados, é apresentada uma técnica híbrida simples pela qual associações físicas e topológicas são produzidas a partir da mesma rede. Para a caracterização de importância, é apresentada uma técnica que considera a importância individual dos itens de dado para determinar o rótulo de um item de teste. O conceito de importância aqui é definido em termos do PageRank, algoritmo usado na engine de busca do Google para definir a importância de páginas da web. Para a otimização estrutural de redes, é apresentado um framework bioinspirado capaz de construir a rede enquanto otimiza uma função de qualidade orientada à tarefa, como, por exemplo, classificação, redução de dimensionalidade, etc. A última investigação apresentada no documento explora a representação baseada em grafo e sua habilidade para detectar classes de distribuições arbitrárias na tarefa de difusão de papéis semânticos. Vários experimentos em bases de dados artificiais e reais, além de comparações com técnicas bastante usadas na literatura, são fornecidos em todas as investigações. Em suma, os resultados obtidos demonstram que as vantagens e novos conceitos propiciados pelo uso de redes se configuram em contribuições relevantes para as áreas de classificação, sistemas de aprendizado e redes complexas. / Data classification is a machine learning and data mining task in which a classifier is trained over a set of labeled data instances in such a way that the labels of new instances can be predicted. Traditionally, classification techniques define decision boundaries in the data space according to the physical features of a training set and a new data item is classified by verifying its relative position to the boundaries. Such kind of classification, which is only based on the physical attributes of the data, makes traditional techniques unable to detect semantic relationship existing among the data such as the pattern formation, for instance. On the other hand, recent works have shown the use of complex networks is a promissing way to capture spatial, topological and functional relationships of the data, as the network representation unifies structure, dynamic and functions of the networked system. In this thesis, the main objective is the development of methods and heuristics based on complex networks for data classification. The main contributions comprise the concepts of pattern conformation, data importance and network structural optimization. For pattern conformation, in which complex networks are employed to estimate the membership of a test item according to the data formation pattern, we present, in this thesis, a simple hybrid technique where physical and topological associations are produced from the same network. For data importance, we present a technique which considers the individual importance of the data items in order to determine the label of a given test item. The concept of importance here is derived from PageRank formulation, the ranking measure behind the Googles search engine used to calculate the importance of webpages. For network structural optimization, we present a bioinspired framework, which is able to build up the network while optimizing a task-oriented quality function such as classification, dimension reduction, etc. The last investigation presented in this thesis exploits the graph representation and its hability to detect classes of arbitrary distributions for the task of semantic role diffusion. In all investigations, a wide range of experiments in artificial and real-world data sets, and many comparisons with well-known and widely used techniques are also presented. In summary, the experimental results reveal that the advantages and new concepts provided by the use of networks represent relevant contributions to the areas of classification, learning systems and complex networks.

A city is considered as a complex system. It consists of numerous interactivesub-systems and is affected by diverse factors including governmental landpolicies, population growth, transportation infrastructure, and market behavior.Land use and transportation systems are considered as the two most importantsubsystems determining urban form and structure in the long term. Meanwhile,urban growth is one of the most important topics in urban studies, and its maindriving forces are population growth and transportation development. Modelingand simulation are believed to be powerful tools to explore the mechanisms ofurban evolution and provide planning support in growth management. The overall objective of the thesis is to analyze and model urban growth basedon the simulation of land-use changes and the modeling of road networkexpansion. Since most previous urban growth models apply fixed transportnetworks, the evolution of road networks was particularly modeled. Besides,urban growth modeling is an interdisciplinary field, so this thesis made bigefforts to integrate knowledge and methods from other scientific and technicalareas to advance geographical information science, especially the aspects ofnetwork analysis and modeling. A multi-agent system was applied to model urban growth in Toronto whenpopulation growth is considered as being the main driving factor of urbangrowth. Agents were adopted to simulate different types of interactiveindividuals who promote urban expansion. The multi-agent model with spatiotemporalallocation criterions was shown effectiveness in simulation. Then, anurban growth model for long-term simulation was developed by integratingland-use development with procedural road network modeling. The dynamicidealized traffic flow estimated by the space syntax metric was not only used forselecting major roads, but also for calculating accessibility in land-usesimulation. The model was applied in the city centre of Stockholm andconfirmed the reciprocal influence between land use and street network duringthe long-term growth. To further study network growth modeling, a novel weighted network model,involving nonlinear growth and neighboring connections, was built from theperspective of promising complex networks. Both mathematical analysis andnumerical simulation were examined in the evolution process, and the effects ofneighboring connections were particular investigated to study the preferentialattachment mechanisms in the evolution. Since road network is a weightedplanar graph, the growth model for urban street networks was subsequentlymodeled. It succeeded in reproducing diverse patterns and each pattern wasexamined by a series of measures. The similarity between the properties of derived patterns and empirical studies implies that there is a universal growthmechanism in the evolution of urban morphology. To better understand the complicated relationship between land use and roadnetwork, centrality indices from different aspects were fully analyzed in a casestudy over Stockholm. The correlation coefficients between different land-usetypes and road network centralities suggest that various centrality indices,reflecting human activities in different ways, can capture land development andconsequently influence urban structure. The strength of this thesis lies in its interdisciplinary approaches to analyze andmodel urban growth. The integration of ‘bottom-up’ land-use simulation androad network growth model in urban growth simulation is the major contribution.The road network growth model in terms of complex network science is anothercontribution to advance spatial network modeling within the field of GIScience.The works in this thesis vary from a novel theoretical weighted network modelto the particular models of land use, urban street network and hybrid urbangrowth, and to the specific applications and statistical analysis in real cases.These models help to improve our understanding of urban growth phenomenaand urban morphological evolution through long-term simulations. Thesimulation results can further support urban planning and growth management.The study of hybrid models integrating methods and techniques frommultidisciplinary fields has attracted a lot attention and still needs constantefforts in near future. / <p>QC 20130514</p>

Complex networks have been used to represent the fundamental structure of a multitude of complex systems from various fields. In the network representation, the system is reduced to a set of nodes and links that denote the elements of the system and the connections between them respectively. Complex networks are commonly adaptive such that the structure of the network and the states of nodes evolve dynamically in a coupled fashion. Adaptive networks lead to peculiar complex dynamics and network topologies, which can be investigated by moment-closure approximations, a coarse-graining approach that enables the use of the dynamical systems theory. In this thesis, I study several contact processes in adaptive networks that are defined by the transmission of node states. Employing moment-closure approximations, I establish analytical insights into complex phenomena emerging in these systems. I provide a detailed analysis of existing alternative moment-closure approximation schemes and extend them in several directions. Most importantly, I consider developing analytical approaches for models with complex update rules and networks with complex topologies. I discuss four different contact processes in adaptive networks. First, I explore the effect of cyclic dominance in opinion formation. For this, I propose an adaptive network model: the adaptive rock-paper-scissors game. The model displays four different dynamical phases (stationary, oscillatory, consensus, and fragmented) with distinct topological and dynamical properties. I use a simple moment-closure approximation to explain the transitions between these phases. Second, I use the adaptive voter model of opinion formation as a benchmark model to test and compare the performances of major moment-closure approximation schemes in the literature. I provide an in-depth analysis that leads to a heightened understanding of the capabilities of alternative approaches. I demonstrate that, even for the simple adaptive voter model, highly sophisticated approximations can fail due to special dynamic correlations. As a general strategy for targeting such problematic cases, I identify and illustrate the design of new approximation schemes specific to the complex phenomena under investigation. Third, I study the collective motion in mobile animal groups, using the conceptual framework of adaptive networks of opinion formation. I focus on the role of information in consensus decision-making in populations consisting of individuals that have conflicting interests. Employing a moment-closure approximation, I predict that uninformed individuals promote democratic consensus in the population, i.e. the collective decision is made according to plurality. This prediction is confirmed in a fish school experiment, constituting the first example of direct verification for the predictions of adaptive network models. Fourth, I consider a challenging problem for moment-closure approximations: growing adaptive networks with strongly heterogeneous degree distributions. In order to capture the dynamics of such networks, I develop a new approximation scheme, from which analytical results can be obtained by a special coarse-graining procedure. I apply this analytical approach to an epidemics problem, the spreading of a fatal disease on a growing population. I show that, although the degree distribution has a finite variance at any finite infectiousness, the model lacks an epidemic threshold, which is a genuine adaptive network effect. Diseases with very low infectiousness can thus persist and prevail in growing populations.

In large and complex systems, failures can have dramatic consequences, such as black-outs, pandemics or the loss of entire classes of an ecosystem. Nevertheless, it is a centuries-old intuition that by using networks to capture the core of the complexity of such systems, one might understand in which part of a system a phenomenon originates. I investigate this intuition using spectral methods to decouple the dynamics of complex systems near stationary states into independent dynamical modes. In this description, phenomena are tied to a specific part of a system through localized eigenvectors which have large amplitudes only on a few nodes of the system's network. Studying the occurrence of localized eigenvectors, I find that such localization occurs exactly for a few small network structures, and approximately for the dynamical modes associated with the most prominent failures in complex systems. My findings confirm that understanding the functioning of complex systems generally requires to treat them as complex entities, rather than collections of interwoven small parts. Exceptions to this are only few structures carrying exact localization, whose functioning is tied to the meso-scale, between the size of individual elements and the size of the global network. However, while understanding the functioning of a complex system is hampered by the necessary global analysis, the prominent failures, due to their localization, allow an understanding on a manageable local scale. Intriguingly, food webs might exploit this localization of failures to stabilize by causing the break-off of small problematic parts, whereas typical attempts to optimize technological systems for stability lead to delocalization and large-scale failures. Thus, this thesis provides insights into the interplay of complexity and localization, which is paramount to ascertain the functioning of the ever-growing networks on which we humans depend.

Recent experimental techniques in biology made possible the acquisition of overwhelming amounts of data concerning complex biological networks, such as neural networks, gene regulation networks and protein-protein interaction networks. These techniques are able to record states of individual components of such networks (neurons, genes, proteins) for a large number of configurations. However, the most biologically relevantinformation lies in their connectivity and in the way their components interact, information that these techniques aren't able to record directly. The aim of this thesis is to study statistical methods for inferring information about the connectivity of complex networks starting from experimental data. The subject is approached from a statistical physics point of view drawing from the arsenal of methods developed in the study of spin glasses. Spin-glasses are prototypes of networks of discrete variables interacting in a complex way and are widely used to model biological networks. After an introduction of the models used and a discussion on the biological motivation of the thesis, all known methods of network inference are introduced and analysed from the point of view of their performance. Then, in the third part of the thesis, a new method is proposed which relies in the remark that the interactions in biology are not necessarily symmetric (i.e. the interaction from node A to node B is not the same as the one from B to A). It is shown that this assumption leads to methods that are both exact and efficient. This means that the interactions can be computed exactly, given a sufficient amount of data, and in a reasonable amount of time. This is an important original contribution since no other method is known to be both exact and efficient.

We are constantly struggling to understand how nature works, trying to identify recurrent events and looking for analogies and relations between objects or individuals. Knowing patterns of behavior is powerful and fundamental for survival of any species. In this thesis, datasets of diverse systems related to transportation, economics, sexual and social contacts, are characterized by using the formalisms of time series and network theory. Part of the results consists on the collection and analyzes of original network data, the rest focuses on the simulation of dynamical processes on these networks and to study how they are affected by the particular structures. The majority of the thesis is about temporal networks, i.e. networks whose structure changes in time. The new temporal dimension reveals structural dynamical properties that help to understand the feedback mechanisms responsible to make the network structure to adapt and to understand the emergence and inhibition of diverse phenomena in dynamic systems, as epidemics in sexual and contact networks. / Vi är ständigt kämpar för att förstå hur naturen fungerar, försöker identifier återkommande evenemang och söker analogier och relationer mellan objekt eller individer. Veta beteendemönster är kraftfull och grundläggande för överlevnad av arter. I denna avhandling, dataset av olika system i samband med transporter är ekonomi, sexuella och sociala kontakter, som kännetecknas av att använda formalismer av tidsserier och nätverk teori. En del av resultatet utgörs av insamling och analys av ursprungliga nätdata, fokuserar resten på simulering av dynamiska processer i dessa nätverk och att studera hur de påverkas av de särskilda strukturer. Huvuddelen av avhandlingen handlar om tidsmässiga nät, i.e. nät vars struktur förändringar i tid. Den nya tidsdimensionen avslöjar strukturella dynamiska egenskaper som hjälper till att förstå den feedback mekanismer som ansvarar för att göra nätverksstruktur att anpassa sig och förstå uppkomsten och hämning av olika företeelser i dynamiska system, epidemier i sexuella och kontaktnät. / Constantemente nos esforçamos para entender como a natureza funciona, tentando identificar eventos recorrentes e procurando por analogias e relações entre objetos ou indivíduos. Conhecer padrões de comportamento é algo poderoso e fundamental para a sobrevivência de qualquer espécie. Nesta tese, dados de sistemas diversos, relacionados a transporte, economia, contatos sexuais e sociais, são caracterizados usando o formalismo de séries temporais e teoria de redes. Uma parte dos resultados consiste na coleta e análise de dados de redes originais, a outra parte concentra-se na simulação de processos dinâmicos nessas redes e no estudo de como esses processos são afetados por determinadas estruturas. A maior parte da tese é sobre redes temporais, ou seja, redes cuja estrutura varia no tempo. A nova dimensão temporal revela propriedades estruturais dinâmicas que contribuem para o entendimento dos mecanismos de resposta responsáveis pela adaptação da rede, e para o entendimento da emergência e inibição de fenômenos diversos em sistemas dinâmicos, como epidemias em redes sexuais e de contato pessoal.

Cette thèse s'organise en deux parties : une première partie s'intéresse aux mesures de similarité entre sommets d'un graphe, une seconde aux méthodes de clustering de graphe biparti. Une nouvelle mesure de similarité entre sommets basée sur des marches aléatoires en temps courts est introduite. Cette méthode a l'avantage, en particulier, d'être insensible à la densité du graphe. Il est ensuite proposé un large état de l'art des similarités entre sommets, ainsi qu'une comparaison expérimentale de ces différentes mesures. Cette première partie se poursuit par la proposition d'une méthode robuste de comparaison de graphes partageant le même ensemble de sommets. Cette mesure est mise en application pour comparer et fusionner des graphes de synonymie. Enfin une application d'aide à la construction de ressources lexicales est présentée. Elle consiste à proposer de nouvelles relations de synonymie à partir de l'ensemble des relations de synonymie déjà existantes. Dans une seconde partie, un parallèle entre l'analyse formelle de concepts et le clustering de graphe biparti est établi. Ce parallèle conduit à l'étude d'un cas particulier pour lequel une partition d’un des groupes de sommets d’un graphe biparti peut-être déterminée alors qu'il n'existe pas de partitionnement correspondant sur l’autre type de sommets. Une méthode simple qui répond à ce problème est proposée et évaluée. Enfin Kodex, un système de classification automatique des résultats d'une recherche d'information est présenté. Ce système est une application en RI des méthodes de clustering vues précédemment. Une évaluation sur une collection de deux millions de pages web montre les avantages de l'approche et permet en outre de mieux comprendre certaines différences entre méthodes de clustering. / This thesis is organized in two parts : the first part focuses on measures of similarity (or proximity) between vertices of a graph, the second part on clustering methods for bipartite graph. A new measure of similarity between vertices, based on short time random walks, is introduced. The main advantage of the method is that it is insensitive to the density of the graph. A broad state of the art of similarities between vertices is then proposed, as well as experimental comparisons of these measures. This is followed by the proposal of a robust method for comparing graphs sharing the same set of vertices. This measure is shown to be applicable to the comparison and merging of synonymy networks. Finally an application for the enrichment of lexical resources is presented. It consists in providing candidate synonyms on the basis of already existing links. In the second part, a parallel between formal concept analysis and clustering of bipartite graph is established. This parallel leads to the particular case where a partition of one of the vertex groups can be determined whereas there is no corresponding partition on the other group of vertices. A simple method that addresses this problem is proposed and evaluated. Finally, a system of automatic classification of search results (Kodex) is presented. This system is an application of previously seen clustering methods. An evaluation on a collection of two million web pages shows the benefits of the approach and also helps to understand some differences between clustering methods.

Neste estudo são apresentados os resultados do trabalho sobre simulações de redes de conexões complexas. Foram simuladas redes regulares, intermediárias e aleatórias com o número de nós e de conexões variando entre 103 e 5x103 e entre 2x104 e 105, respectivamente, e com probabilidade variando de 0 a 1 com passo de 0.1, com o enfoque na Teoria Espectral. Utilizando a linguagem C e o software Matlab, as redes são representadas pela sua matriz adjacência, com o objetivo de observar-se o comportamento de seus autovalores através de histogramas. A finalidade é a caracterização de redes complexas. Observa-se que a distribuição dos autovalores segue a lei semicircular de Wigner. / This study presents the results of the work about simulations of networks of complex connections. They were simulate regular networks, middlemen and aleatory with the number of nodes and of connections varying between 103 and 5x104 and between 2x104 and 105, respectively, and with probability varying from 0 to 1 with step of 0.1, with the focus in the Spectral Theory. Using the language C and the software Matlab, the networks are represented by its adjacency matrix, with the objective of observing the behavior of its eigenvalues through histograms. The purpose is the characterization of complex networks. Its observed that the eigenvalues distribution follows the Wigners semicircular law.

Made available in DSpace on 2016-06-02T19:06:01Z (GMT). No. of bitstreams: 1 4845.pdf: 2916613 bytes, checksum: aa64366fdbe9befa8f14547fa4545763 (MD5) Previous issue date: 2012-08-18 / Multi-Electrode Array, MEA, was developed more than thirty years ago. This planar device of multiple microelectrodes has been used to detect local electric potential variations created by the ion movement through the protein channels that traverses the cell membranes in the near neighborhood. MEA offers the possibility of non invasive registering of the cell and their network s activities, allowing to know how the neurons start to connect through the synapses forming a network and generating spontaneous electrophysiological activities. In order to understand the neural network dynamics in face of the spontaneous activities and its evolution in dissociated hippocampal cell cultures, important properties in the synaptic plasticity, in this work it is proposed the analysis of the evolution and modeling of the hippocampal cell cultures in MEA using the theory of Partial Directed Coherence and Complex Networks. There were used the electrophysiological records obtained using MEA60 System, of the dissociated neurons of 18 days old Wistar rat embyo, in an experiment denoted as 371, realized at the University of Genoa, Italy. As the results obtained using the Partial Directed Coherence approach, it was verified that the method is capable to detect neuronal connectivity in the neuron cultures using MEA, even with the noisy signals. It was also verified that different time delays between signals during application of the PDC method do not affect directly on the results of the causality. PDC allowed to show that in MEA the amount of direct connections resulted is less than the amount of indirect connections, through the microelectrodes. This can indicate that the neurons prefer to communicate through existing connections than creating new connections. It was also observed that it is easier to lose direct connections than indirect connections between microelectrodes through the time. Through the experiments it can be observed that the culture in 25 DIV (Days In Vitro) developed more amount of connections between neighboring electrodes, with less overall connections than the culture in 46 DIV, that had more overall connections with less neighborhood connections. Since one of the PDC features is the directionality detection between connections, it was observed direction changes through the connections through the time, even though we do not know the physiological meaning of these changes in the cognitive process. It was also observed that the established connections do not follow random patterns, showing an indicative of a free scale network, although we used small statistical measures to characterize the networks. / Há mais de 30 anos a Matriz de microeletrodos (Multi-Electrode Array,MEA) foi desenvolvida. Este dispositivo planar de múltiplos microeletrodos tem permitido detectar as variações do potencial elétrico local que são criadas pelo movimento de íons através dos canais de proteínas que atravessam as membranas das células neuronais em sua vizinhança imediata. A MEA oferece a possibilidade de gravação não-invasiva da atividade das células e de redes de células, permitindo-nos conhecer como os neurônios se comunicam através de sinapses formando uma rede e disparando atividades eletrofisiológicas espontâneas evocadas. No sentido de poder entender melhor a dinâmica das redes neurais em termos de atividades espontâneas e sua evolução em culturas de células hipocampais dissociadas, que são as propriedades mais importantes na plasticidade sináptica, neste trabalho é proposta a análise da evolução e o modelamento de culturas de células hipocampais na MEA usando a teoria de Coerência Parcial Direcionada (PDC) e de Redes Complexas. Foram usados os registros eletrofisiológicos obtidos usando o sistema MEA60, de neurônios dissociados de embrião de rato Wistar, de 18 dias de vida em um experimento denotado como 371, realizado na Universidade de Gênova, Itália. Como resultados obtidos ao utilizar o método de Coerência Parcial Direcionada, verificou-se que o método é capaz de detectar conectividade neuronal nas culturas de neurônios da MEA, mesmo com sinais com presença de ruído. Também verificou-se que tempos de atraso diferentes na aplicação do PDC não tem uma influência direta nos resultados de causalidade. O PDC permitiu mostrar que na MEA a quantidade de conexões diretas estabelecidas foram sempre em menor quantidade do que as conexões indiretas, através de caminhos. Isso pode ser um indicativo de que os neurônios preferem estabelecer comunicações usando caminhos ja existentes do que criando novas conexões. Também observou-se que com o passar do tempo é mais fácil se perderem conexões diretas do que conexões intermediárias entre neurônios. Através dos experimentos realizados pode se obervar que a cultura no 25 DIV (dias in vitro) desenvolveu maior quantidade de conexões com nós vizinhos, possuindo menos conexões em total, do que a rede no 46 DIV que teve maior quantidade de conexões total mas menor quantidade de conexões com nós vizinhos. Como uma das características do PDC é a deteção da direcionalidade nas conexões, observouse mudanças no sentido das conexões ao longo do registro, embora não saibamos ainda o significado fisiológico dessas mudanças em processos cognoscitivos. Também observou-se que as conexões estabelecidas não seguiam um padrão aleatório, encontrando-se um indicativo de um comportamento de redes livres de escala, embora não possamos afirmar isso por ter utilizado poucas medidas estatísticas para caracterizar as redes.

Neste estudo são apresentados os resultados do trabalho sobre simulações de redes de conexões complexas. Foram simuladas redes regulares, intermediárias e aleatórias com o número de nós e de conexões variando entre 103 e 5x103 e entre 2x104 e 105, respectivamente, e com probabilidade variando de 0 a 1 com passo de 0.1, com o enfoque na Teoria Espectral. Utilizando a linguagem C e o software Matlab, as redes são representadas pela sua matriz adjacência, com o objetivo de observar-se o comportamento de seus autovalores através de histogramas. A finalidade é a caracterização de redes complexas. Observa-se que a distribuição dos autovalores segue a lei semicircular de Wigner. / This study presents the results of the work about simulations of networks of complex connections. They were simulate regular networks, middlemen and aleatory with the number of nodes and of connections varying between 103 and 5x104 and between 2x104 and 105, respectively, and with probability varying from 0 to 1 with step of 0.1, with the focus in the Spectral Theory. Using the language C and the software Matlab, the networks are represented by its adjacency matrix, with the objective of observing the behavior of its eigenvalues through histograms. The purpose is the characterization of complex networks. Its observed that the eigenvalues distribution follows the Wigners semicircular law.