Tesis sobre el tema "Similarity of structure"

Aquest treball s'emmarca en l'àmbit del processament del llenguatge natural. Té l'objectiu d'explorar el potencial de la similitud verbal i, més concretament, de les classificacions verbals, a l'hora de capturar i modelitzar la informació bàsica relacionada amb l'expressió d'esdeveniments en espanyol. La tesi s'articula entorn de dos estudis que examinen la capacitat que té la similitud verbal de modelitzar la informació relativa als participants en esdeveniments. En primer lloc, elaborem una anàlisi de la similitud verbal amb relació a l'estructura argumental. Amb aquesta finalitat, partim de tres perspectives que tracten aquest tema: la lingüística teòrica, la lingüística de corpus i la psicolingüística, i analitzem com cadascuna defineix la similitud entre els verbs. Aquesta anàlisi ens serveix per a definir un conjunt de característiques lingüístiques i configuracions que s'apliquen en el segon estudi. Aquest estudi consisteix en la creació d'una classificació automàtica de sentits verbals amb l'ús d'un algorisme de clusterització (clustering). L'objectiu d'aquesta classificació és capturar l'estructura argumental dels verbs i reflectir-la en les classes, de tal manera que permeti modelitzar els participants en els esdeveniments expressats pels verbs. Les troballes ens permeten afirmar que la classificació verbal organitza la informació de manera que és capaç d'acomodar diferents aspectes de l'estructura argumental.<br>Este trabajo se enmarca dentro del ámbito del procesamiento del lenguaje natural. Su objetivo es explorar el potencial de la similitud verbal y, más concretamente, de las clasificaciones verbales, a la hora de capturar y modelizar la información básica relacionada con la expresión de acontecimientos en español. La tesis se articula en torno a dos estudios que examinan la capacidad de la similitud verbal para modelizar la información relativa a los participantes en acontecimientos. En primer lugar, elaboramos un análisis de la similitud verbal con relación a la estructura argumental. Para ello tomamos tres perspectivas que tratan este tema: la lingüística teórica, lingüística de corpus y la psicolingüística, y analizamos de qué modo cada una de ellas define la similitud entre los verbos. Este análisis nos sirve para definir un conjunto de características lingüísticas y configuraciones que se aplican en el segundo estudio. Este estudio consiste en la creación de una clasificación automática de sentidos verbales usando un algoritmo de agrupamiento (clustering). El objetivo de esta clasificación es capturar la estructura argumental de los verbos y reflejarla en las clases, de tal manera que permita modelizar a los participantes en los acontecimientos expresados por los verbos. Los hallazgos nos permiten afirmar que la clasificación verbal organiza la información de manera que es capaz de acomodar diferentes aspectos de la estructura argumental.<br>This paper aims to explore the potential of verb similarity, and more specifically of verb classifications, when it comes to capturing and modelling basic information related to events expressed in Spanish. The research is based on two studies that examine verb similarity's ability to model event participant information. We first perform a study of verb similarity with respect to argument structure, looking at its relevant characteristics through the lens of three different perspectives: linguistic theory, corpus linguistics and psycholinguistics. Based on this analysis, we choose the features and configurations to be explored in order to create an automatic classification of verb senses using a clustering algorithm. The aim of this automatic classification is to capture the argument structure of the verbs and apply it to the classifications in a way that allows us to adequately model the participants in the events expressed by those verbs. The evaluations carried out for this verb classification confirm automatic classifications' ability to capture and infer relevant information related to participants in events.

Um dos principais desafios remanescentes em Inteligência Artificial é como fazer sistemas inteligentes ancorar abstrações de alto nível na realidade. Pelo menos parte do problema vai em direção da questão de qual arcabouço de representação é mais apropriado de maneira que facilite o reconhecimento de objetos. A cognição em animais, particularmente em humanos, pode dar pistas de como tal arcabouço de representação se parece. Estudos na cognição do reconhecimento de objetos sugere que o problema da representação na cognição incorpora informações a respeito de similaridade e informação holística-estrutural (i.e. todo-parte) a respeito de conceitos. Temos a visão de que sistemas computacionais que lidam com relações de todo-parte deveriam representar informações holístico-estruturais e similaridade. No entanto, não existe arcabouço de representação que permite tais informações serem representadas de forma integrada. Esta tese propõe um arcabouço de representação de informação de similaridade entre estruturas de todo-parte. Nossa proposta é baseada na teoria dos espaços conceituais. Estes são espaços matemáticos onde regiões e pontos representam conceitos e objetos respectivamente, tal que a distância entre estas entidades denota a sua similaridade. Na nossa proposta, todos e partes são representados em espaços conceituais distintos, chamados espaços holísticos e estruturais. Espaços holísticos permitem o julgamento de similaridade sistemático entre todos. Por outro lado, espaços estruturais permitem o julgamento de similaridade sistemático entre estrutura de partes. Um ponto em um espaço estrutural denota uma estrutura particular de partes; regiões neste espaço representam diferentes tipos de estruturas de parte. Através da redefinição de espaços conceituais como um produto de espaços holísticos e estruturais, nosso objetivo é permitir a representação de informações de similaridade entre diferentes todos, levando em consideração a similaridade entre partes compartilhadas e suas configurações. Esta tese tem três contribuições principais: uma teoria geral sobre espaços holísticos e estruturais; uma formalização da teoria fundada em produto de espaços métricos; e um algoritmo genérico para reconhecimento de objetos, implementando processamento holístico-estrutural.<br>One of main remaining challenges in Artificial Intelligence is how to make intelligent systems to ground high-level abstractions in reality. At least part of this problem comes down to the question of what representation framework is best suited in a way that facilitates object recognition. Animal cognition, particularly in humans, can give a clue of how such representation framework looks like. Studies on the cognition of object recognition suggests that representation in cognition incorporates similarity and holistic-structural (i.e. whole-part) information about concepts. We hold the view that computer systems dealing with part-whole relations should be able to represent similarity and holistic-structural information. However, there exists no representation framework that allows such information to be represented in an integrated way. This thesis proposes a concept representation framework for representing information about similarity between part-whole structures. We base our proposal on the theory of conceptual spaces, which are mathematical spaces where regions and points represent concepts and objects respectively, such that their distance denotes their similarity. In our proposal, parts and wholes are represented in distinct conceptual spaces called holistic and structure spaces. Holistic spaces allow for systematic similarity judgements between wholes. On the other hand, structure spaces allow for systematic similarity judgement between structures of parts. A point in the structure space denotes a particular part structure; regions in the space represent different general types of part structures. By redefining conceptual spaces as a product of holistic and structure spaces, our goal is to allow one to represent similarity information between different wholes, taking into consideration the similarity between shared parts and their configurations. This thesis has three main contributions: a general theory about holistic and structure spaces; a formulation of the theory founded on products of metric spaces; and a generic algorithm for object recognition implementing holistic-structural processing.

The process of protein structure prediction is a crucial part of understanding the function of the building blocks of life. It is based on the approximation of a protein free energy that is used to guide the search through the space of protein structures towards the thermodynamic equilibrium of the native state. A function that gives a good approximation of the protein free energy should be able to estimate the structural distance of the evaluated candidate structure to the protein native state. This correlation between the energy and the similarity to the native is the key to high quality predictions. State-of-the-art protein structure prediction methods use very simple techniques to design such energy functions. The individual components of the energy functions are created by human experts with the use of statistical analysis of common structural patterns that occurs in the known native structures. The energy function itself is then defined as a simple weighted sum of these components. Exact values of the weights are set in the process of maximisation of the correlation between the energy and the similarity to the native measured by a root mean square deviation between coordinates of the protein backbone. In this dissertation I argue that this process is oversimplified and could be improved on at least two levels. Firstly, a more complex functional combination of the energy components might be able to reflect the similarity more accurately and thus improve the prediction quality. Secondly, a more robust similarity measure that combines different notions of the protein structural similarity might provide a much more realistic baseline for the energy function optimisation. To test these two hypotheses I have proposed a novel approach to the design of energy functions for protein structure prediction using a genetic programming algorithm to evolve the energy functions and a structural similarity consensus to provide a reference similarity measure. The best evolved energy functions were found to reflect the similarity to the native better than the optimised weighted sum of terms, and therefore opening a new interesting area of research for the machine learning techniques.

Thesis (MComm)--Stellenbosch University, 2012.<br>ENGLISH ABSTRACT: This study investigated race-group similarity effects as a form of interviewer bias in selection interview ratings. Social Identity Theory predicts that interviewers would assign higher ratings to interviewees of the same social group (the so-called in-group) primarily through the mechanism of similarity attraction. Research findings up to now have lent only partial support to this hypothesis. This study argues that interview structure may help to explain inconsistent research findings since structure could inhibit the functioning of the similarity-attraction mechanism. The present research pursued two objectives, namely (1) to determine the degree to which race-group similarity (between interviewer and interviewee) exerts a biasing effect on selection interview dimension ratings, (2) to determine whether same-group bias increases when interview structure is experimentally diminished. This experimental study manipulated the degree of structure in interviews (high- and low-structured conditions) and compared the degree to which race group similarity effects were evident under each condition. Interviews were simulated by showing video-taped interview segments to a sample of participants and asking them to rate interview dimensions on rating scales that had been compiled to reflect the degree of structure in each condition. The data were analysed using Hierarchical Linear Modelling (HLM) and multiple regression analysis to determine whether similarity effects were present in the interview rating data. The results support the hypothesis that racial similarity effects are found under low-structured conditions, as well as the hypothesis that interview structure moderates the influence of similarity effects. However, racial similarity effects were also found with the highly structured condition. Although these effects were smaller than in the low-structured condition, they were statistically significant. Future research should attempt to replicate this study as a field study to test the generalisability of the findings.<br>AFRIKAANSE OPSOMMING: Hierdie studie ondersoek onderhoudvoerdersydigheid in die vorm van rasgroepsoortgelykheidseffekte in seleksie-onderhoudbeoordelings. Sosiale Identiteitsteorie voorspel dat onderhoudvoerders diegene van dieselfde rasgroep (die sogenaamde ingroep) met hoër beoordelingstellings sal aanslaan, primêr deur die werking van die soortgelykheidaangetrokkendheidsmeganisme. Navorsingsresultate tot op hede leen slegs gedeeltelike steun aan hierdie hipotese. Hierdie studie argumenteer dat die rede vir teenstrydige navorsingbevindinge moontlik die gevolg van die bemiddelende effek van onderhoudstruktuur kan wees, aangesien struktuur moontlik die funksionering van die soortgelykheid-aangetrokkendheidsmeganisme kan inperk. Die studie streef dus twee doelwitte na, nl. (1) om die mate waartoe rasgroepooreenstemming tussen die onderhoudvoerder en onderhoudnemer ’n sydige invloed op onderhouddimensietellings uitoefen te bepaal en (2) om te bepaal of soortgelykheidseffekte toeneem namate onderhoudstruktuur eksperimenteel verlaag word. ’n Eksperimentele ontwerp is gebruik waarbinne onderhoudstruktuur (hoog- en laag gestruktuurde toestande) in video-opnames van onderhoude nageboots is. ’n Groep beoordelaars het hierdie stimilusmateriaal beoordeel aan die hand van beoordelingskriteria wat opgestel is om die mate van struktuur binne elke toestand te reflekteer. Gevolglik is die mate van rasgroepsoortgelykheidseffekte binne elke struktuurtoestand vergelyk. Die navorsingsdata is met gebruik van Hiërargiese Lineêre Modellering (HLM) en veelvoudige regressie ontleed om die teenwoordigheid van soortgelykheidsydigheid te bepaal. Die resultate steun die hipotese dat rassoortgelykheidseffekte onder laaggestruktuurde toestande voorkom, asook dat onderhoudstruktuur ’n modererende rol speel. Nietemin is soortgelyke effekte ook onder die hoog gestruktuurde toestand gevind. Alhoewel hierdie effekte kleiner as onder die laaggestruktuurde toestand was, was dit steeds statisties beduidend. Toekomstige navorsing kan poog om ‘n soortgelyke ondersoek as ‘n veldstudie te onderneem om die moontlikheid van veralgemening van die resultate te bepaal.

Lança-se aqui a hipótese de que é possível reproduzir o comportamento do protótipo a partir de modelos feitos de materiais diferentes dos utilizados na estrutura de referência. A influência de distorção de espessura, de diferentes densidades, tensões de escoamento, encruamento e viscoplasticidade são todas avaliadas através de equações analíticas e simulações das estruturas básicas como viga, placa e o impacto de uma estrutura de chapa dupla. Adicionalmente, são realizadas caracterizações quase-estáticas e dinâmicas de material e experimentos em placas circulares para três escalas diferentes (1/1, 2/3 e 1/3), quatro materiais (alumínio, titânio, aço inox e cobre) e duas condições de carregamento: impacto de uma massa a baixa velocidade (3,5m/s) e um projétil a alta velocidade (130m/s). Mostra-se que o uso da lei de Johnson-Cook para descrever o comportamento viscoplástico dos materiais permite obter boa aproximação para o cálculo da velocidade de impacto corrigida, sem que sejam necessárias informações adicionais do comportamento da estrutura. Além disso, os resultados revelam a importância de se manter a razão entre massa de impacto e massa da estrutura constante em modelo e protótipo, levando a desejável similaridade das distorções entre ambas. Assim, ao se confirmar a hipótese deste trabalho, mostra-se possível inferir o comportamento de um protótipo com uso de modelos em escala reduzida feitos a partir de materiais diferentes.<br>It is hypothesized that it is possible to reproduce the behaviour of the prototype by using models made of different materials. The influence of thickness, different densities, flow stresses, hardening and viscoplasticity are all evaluated through analytical equations and simulations of basic structures such as beam, plate and the impact of a double plate structure. In addition, quasi-static and dynamic materials characterization are conducted , with experiments being performed in circular plates for three different scales (1/1, 2/3 and 1/3), four materials (Aluminum, Titanium, Stainless Steel and Copper) and two conditions impact of a mass (3,5m/s) and a projectile at high speed (130m/s). It is shown that the use of Johnson-Cook\'s law to describe the viscoplastic behavior of the materials allows a satisfactory approximation to the calculation of the corrected impact velocity, without the need for additional information on the behavior of the structure. In addition, the results reveal the relevance of maintaining the ratio between impact mass and structure mass constant in model and prototype, leading to the desirable similarity of the distortions between the two. Thus, when confirming the hypothesis of this work, it is possible to infer the behavior of a prototype using small scale models made from different materials.

Three-dimensional ligand-based virtual screening methods have been used for many years in drug discovery, with a variable success depending on different factors, such as the complexity of the target system or the suitability of the molecular descriptors. New approaches are still necessary to cover the broad spectrum of relationships that a drug-like molecule may establish with the organism. In spite of the complexity of processes that modulate the activity of a drug, most tools are primarily focused on the use of shape or electrostatic descriptors. In contrast, since the importance of lipophilicity in pharmacodynamics and pharmacokinetics process, an exact representation of the 3D pattern of hydrophobic/hydrophilic regions can be a valuable guideline to enhance the molecular similarity studies. In this scenario, PharmScreen was conceived as a tool to exploit lipophilic 3D similarity. Exploiting the MST contributions to octanol/water partition coefficients, the capacity to perform correct molecular overlays and distinguish between active and inactive molecules is discussed. The overlap algorithm is validated against the AstraZeneca test, which comprises 121 experimentally derived sets of molecular overlays. The results point out the suitability of the MST-based hydrophobic parameters for generating molecular overlays, as correct predictions were obtained for 94%, 79%, and 54% of the molecules classified into easy, moderate, and hard sets, respectively. Moreover, the results point out that this accuracy is attained at a much lower degree of identity between the templates used by hydrophobic/HB fields and electrostatic/steric ones. On the other hand, the topological hydrophobic descriptors proposed are applied over 3D-QSAR models. In this context, the Miertus–Scrocco–Tomasi-derived hydrophobic descriptors have been shown to provide models for structure–activity relationships with a predictive accuracy comparable to traditional techniques based on electrostatic/steric parameters. The results reported support the assumption that lipophilicity, supplemented by HB acceptors/donors, provides a useful signature to enrich the information that can be retrieved from (i) molecular alignment and (ii) QSAR models, complementing the results obtained traditionally from electrostatic and steric properties. Taken together, lipophilicity is presented as a valuable alternative for the molecular similarity study. In addition, the applicability of our descriptors in structure-based methods has been explored in order to re-evaluate the complexes constituted by docking techniques (in our case, Glide). Since (de)solvation is fundamental for the establishment of the ligand-receptor complex, it can be expected that the docked ligands in the same pocket share lipophilic characteristics, even if there are several binding modes. However, approximations that affect solvation contribution are applied in the docking score functions, and by extension, some docking programs show problems performing VS especially in hydrophobic binding pockets. Specific binding typically requires the formation of key interactions between targets and ligands. Thus, 3D similarity relative to experimental binding modes could be sufficient to distinguish active compounds from decoys. In view of the results obtained the similarity descriptors proposed are introduced as a valid scoring function for discerning between active and inactive compounds. These findings support the usefulness of lipophilicity as driver descriptors in molecular similarity studies promoting their use in virtual screening campaigns considering LB approaches or in combination with SB. As conclusion, results obtained from the analysis of hydrophobic/hydrophilic descriptors presented in this thesis opens a new window to explore the vast chemical space, complementing the information derived from traditional descriptors in ligand- and structure-based approaches.<br>El fet d'assumir que molècules estructuralment semblants donaran lloc a activitats biològiques similars ha estat una idea àmpliament explotada en el disseny de fàrmacs. Aquesta premissa subjau en la majoria de les aplicacions pràctiques en recerca química i farmacèutica. No obstant això, el concepte de similitud molecular és subjectiu i la seva interpretació pot variar segons l’ús que se’n vulgui derivar. La quantificació d’aquesta mesura de semblança molecular depèn de la representació de les característiques químiques presents en l'estructura molecular mitjançant descriptors 1D, 2D o 3D, la ponderació d'aquests descriptors i l'expressió matemàtica de la funció de similitud. En l’àmbit de les característiques químiques utilitzades en els mètodes tridimensionals de similitud molecular, les propietats electrostàtiques i estèriques han estat dominants tradicionalment. Tanmateix, això oculta el paper fonamental exercit per altres contribucions a l'afinitat d'unió, com els canvis en la (de)solvatació del lligant i del receptor. Malgrat la seva rellevància, la lipofilicitat ocupa aparentment un paper secundari com a descriptor principal del reconeixement lligand-receptor. Sota aquesta premissa s’ha desenvolupat una eina de cribratge virtual 3D basada en lligands (PharmScreen) que explota les relacions de similitud entre topologies hidrofòbiques derivades del model continuo de solvatació Miertus – Scrocco – Tomasi (MST). Els estudis reportats al llarg d’aquesta tesis recolzen la utilitat de les contribucions atòmiques a la lipofilicitat com a descriptors fonamentals en estudis de similitud, complementant la informació derivada dels descriptors tradicional. PharmScreen es presenta, així, com una eina competitiva per aplicar en campanyes de cribratge virtual basada en lligand o en combinació amb tècniques basades en proteïna, obrint una nova finestra en l’ampli espai químic.

The aim of this thesis is to evaluate the sufficiency of Turkey for joining the Economic and Monetary Union of the European Union (EMU) in terms of similarity and convergence. The study has been conducted in a comparative and descriptive way. First, the similarity and convergence of Turkey to some selected countries are examined with respect to her macroeconomic position. When taking EMU as a benchmark and comparing the convergence of Turkey with the convergence of some of the countries and country groups, it is found that the macroeconomic deficiencies of Turkey are not in an extent that characterizes Turkey as a totally insufficient candidate for EMU. Next, whether there are similarity and convergence in trade structures of Turkey and the European Union of 15 member states (EU15) for the period between 1995 and 2008 is inspected. The results indicated that Turkish export structure is clearly converging to the export structure of EU15 in the course of time. In general, findings of the thesis indicated that there is mostly a continuous convergence in all of the indicators considered but still Turkey does not meet all of the convergence criteria, perfectly. Therefore, as a result of the examinations, some suggestions have been made which would facilitate EMU membership of Turkey.

Self-similar structure of the carbon thin films, obtained by magnetron sputtering is investigated numerically. Statistical parameters are calculated within two dimensional multifractal detrended fluctuation analysis. The numerical model for the surfaces under investigation was build from the SEM images of the carbon thin films. It is shown that the self-similarity in surface roughness preserves over all fragments of the sample, and through different resolutions of the SEM images. When you are citing the document, use the following link http://essuir.sumdu.edu.ua/handle/123456789/20616

The advent of high-throughput sequencing and structure determination techniques has had a tremendous impact on our quest in cracking the language of life. The genomic and protein data is now being accumulated at a phenomenal rate, with the motivation of deriving insights into the function, mechanism, and evolution of the biomolecules, through analysis of their similarities, differences, and interactions. The rapid increase in the size of the biomolecular databases, however, calls for development of new computational methods for sensitive and efficient management and analysis of this information. In this thesis, we propose and implement several approaches for accurate and highly efficient comparison and retrieval of protein sequences and structures. The observation that corresponding residues in related proteins share similar inter-residue contacts is exploited in derivation of a new set of biologically sensitive metric amino acid substitution matrices, yielding accurate alignment and comparison of proteins. The metricity of these matrices has allowed efficient indexing and retrieval of both protein sequences and structures. A landmark-guided embedding of protein sequences is developed to represent subsequences in a vector space for approximate, but extremely fast spatial indexing and similarity search. Whereas protein structure comparison and search tasks were hitherto handled separately, we propose an integrated approach that serves both of these tasks and performs comparable to or better than other available methods. Our approach hinges on identification of similar residue contacts using distance-based indexing and provides the best of the both worlds: the accuracy of detailed structure alignment algorithms, at a speed comparable to that of the structure retrieval algorithms. We expect that the methods and tools developed in this study will find use in a wide range of application areas including annotation of new proteins, discovery of functional motifs, discerning evolutionary relationships among genes and species, and drug design and targeting.

Recent work in microbial ecology has focused on elucidating controls over biogeographic patterns and connecting microbial community composition to ecosystem function. My objective was to investigate the relative influences of landscape legacies and contemporary environmental factors on the distribution of soil microbial communities and their contribution to ecosystem processes across a glacial till sequence in Taylor Valley, Antarctica. Within each till unit, I sampled from dry areas and areas with visible evidence of recent surface water movement generated by seasonal melting of ephemeral snow packs and hillslope ground ice. Using T-RFLP 16S rRNA gene profiles of microbial communities, I analyzed the contribution of till and environmental factors to community similarity, and assessed the functional potential of the microbial community using extracellular enzyme activity assays. Microbial communities were influenced by geochemical differences among both tills and local environments, but especially organized by variables associated with water availability as the first axis of an NMDS ordination was strongly related to shifts in soil moisture content. CCA revealed that tills explained only 3.4% of the variability in community similarity among sites, while geochemical variables explained 18.5%. Extracellular enzyme activity was correlated with relevant geochemical variables reflecting the influence of nutrient limitation on microbial activity. In addition, enzyme activity was related to changes in community similarity, particularly in wet environments with a partial Mantel correlation of 0.32. These results demonstrate how landscape history and environmental conditions can shape the functional potential of a microbial community mediated through shifts in microbial community composition.<br>Master of Science

Since multimedia data, such as images and videos, are way more expressive and informative than ordinary text-based data, people find it more attractive to communicate and express with them. Additionally, with the rising popularity of social networking tools such as Facebook and Twitter, multimedia information retrieval can no longer be considered a solitary task. Rather, people constantly collaborate with one another while searching and retrieving information. But the very cause of the popularity of multimedia data, the huge and different types of information a single data object can carry, makes their management a challenging task. Multimedia data is commonly represented as multidimensional feature vectors and carry high-level semantic information. These two characteristics make them very different from traditional alpha-numeric data. Thus, to try to manage them with frameworks and rationales designed for primitive alpha-numeric data, will be inefficient. An index structure is the backbone of any database management system. It has been seen that index structures present in existing relational database management frameworks cannot handle multimedia data effectively. Thus, in this dissertation, a generalized multidimensional index structure is proposed which accommodates the atypical multidimensional representation and the semantic information carried by different multimedia data seamlessly from within one single framework. Additionally, the dissertation investigates the evolving relationships among multimedia data in a collaborative environment and how such information can help to customize the design of the proposed index structure, when it is used to manage multimedia data in a shared environment. Extensive experiments were conducted to present the usability and better performance of the proposed framework over current state-of-art approaches.

The investigation of the hidden metric space behind complex network topologies is a fervid topic in current network science and the hyperbolic space is one of the most studied, because it seems associated to the structural organization of many real complex systems. The popularity-similarity-optimization (PSO) model simulates how random geometric graphs grow in the hyperbolic space, generating realistic networks with clustering, small-worldness, scale-freeness and rich-clubness. However, it misses to reproduce an important feature of real complex networks, which is the community organization. The geometrical-preferential-attachment (GPA) model was recently developed in order to confer to the PSO also a soft community structure, which is obtained by forcing different angular regions of the hyperbolic disk to have a variable level of attractiveness. However, the number and size of the communities cannot be explicitly controlled in the GPA, which is a clear limitation for real applications. Here, we introduce the nonuniform PSO (nPSO) model. Differently from GPA, the nPSO generates synthetic networks in the hyperbolic space where heterogeneous angular node attractiveness is forced by sampling the angular coordinates from a tailored nonuniform probability distribution (for instance a mixture of Gaussians). The nPSO differs from GPA in other three aspects: it allows one to explicitly fix the number and size of communities; it allows one to tune their mixing property by means of the network temperature; it is efficient to generate networks with high clustering. Several tests on the detectability of the community structure in nPSO synthetic networks and wide investigations on their structural properties confirm that the nPSO is a valid and efficient model to generate realistic complex networks with communities.

Esta dissertação avalia a aplicação de leis de similaridade distorcidas no contexto de impacto estrutural. A análise se apoia em um estudo teórico, numérico e experimental do impacto de um indentador contra um painel duplo. O modelo analítico descreve de forma simplificada o comportamento de partes desta estrutura e as simulações numéricas reproduzem os ensaios experimentais que utilizam um protótipo (tamanho real) e modelo (escala reduzida). A diferença nas propriedades mecânicas do material de construção do modelo e protótipo é considerada no procedimento de escalonamento, bem como os efeitos de escala por causa da taxa de deformação. Ainda, diante das limitações do aparato experimental, é desenvolvida uma formulação para as leis de similaridade que permite variações da massa impacto e da velocidade inicial do elemento impactante no ensaio. Dessa forma, apresenta-se um procedimento que permite inferir o comportamento de estruturas em tamanho real sob carregamento de impacto através do uso de estruturas em escala, mesmo com as limitações de aparato ou diferenças das propriedades mecânicas do material.<br>This work evaluates the distorted similarity laws applied in structural impact. The analysis is based in theoretical, numerical and experimental studies of the impact of an indenter to a reinforced panel. The theoretical approach describes, in a simplified manner, the behaviour of the structure components while the numerical analysis reproduces the experiments performed in two scales: prototype (large scale) and model (small scale). Although the panels are made of different materials, this mismatch in mechanical behaviour is taken into account in the scaling procedure, as well as the scale effects due to strain rate. A formulation that allows flexibility in experiment variables such as initial velocity and impact mass is developed due to experimental apparatus limitations. In general lines, is developed a procedure allowing to infer the behaviour of a large scale structure under impact load using scaled structures, while using different materials for prototype and model and respecting the experimental apparatus limits.

La comparaison de site peut-elle vérifier l’hypothèse: «Les origines biosynthétiques des produits naturels leurs confèrent des activités biologiques»? Pour répondre à cette question, nous avons développé un outil modélisant les propriétés accessibles au solvant des sites de liaison. La méthode a montré des aspects intéressants, mais elle souffre d’une sensibilité aux coordonnées atomiques. Cependant, des méthodes existantes nous ont permis de prouver que l’hypothèse est valide pour la famille des flavonoïdes. Afin d’étendre l’étude, nous avons développé un procédé automatique capable de rechercher des structures d’enzymes de biosynthèse de produits naturels disposant de sites actifs capables de lier une molécule de petite taille. Nous avons trouvé les structures de 117 enzymes.Les structures nous ont permis de caractériser divers modes de liaison substrat-enzyme, nous indiquant l’empreinte biologique des produits naturels ne correspond pas toujours au modèle « clé- serrure »<br>Can computational binding site similarity tools verify the hypothesis: “Biosynthetic moldings give potent biological activities to natural products”? To answer this question, we designed a tool modeling binding site properties according to solvent exposure. The method showed interesting characteristics but suffers from sensitivity to atomic coordinates. However, existing methods have delivered evidence that the hypothesis was valid for the flavonoid chemical class. In order to extend the study, we designed an automated pipeline capable of searching natural products biosynthetic enzyme structures embedding ligandable catalytic sites. We collected structures of 117 biosynthetic enzymes. Finally, according to structural investigations of biosynthetic enzymes, we characterized diverse substrate-enzyme binding-modes, suggesting that natural product biological imprints usually do not agree with the “key-lock” model

This study provides a systematic description of the nature and evolution of major transactions in the U.S. apparel sector, using a theory that applies across sectors. This research investigates the determinants of the existence and magnitude of major inter-firm transactions, relying on a unique longitudinal dataset of over 2,000 of the largest transactional (buy-sell) relations between publicly traded firms in the U.S. apparel sector. The results indicate the importance of inter-firm complementarity, rather than inter-firm similarity, in explaining the sector architecture; thus contributing to the future improvement of industry classification systems. This study also contributes to a deeper understanding of the apparel sector focusing on the change in the relative importance of manufacturing activities versus service activities and in the involvement of firms from the outside apparel sector. Implications of inter-firm transactions are discussed regarding industry policies, and human and environmental welfare in manufacturing and raw materials industries.

L’analogie est un mécanisme fondamental permettant d’interpréter une nouvelle situation à travers des expériences passées. La présente thèse vise à redonner toute sa force à ce postulat en mettant en avant la capacité à percevoir comme essentiellement similaires des expériences d’apparence pourtant différentes. Partant du principe que les analogies constituent un mécanisme naturel par lequel le système cognitif traite l’information nouvelle, un intérêt particulier est attribué à leur manifestation spontanée (i.e. sans incitation par un tiers à effectuer la comparaison), telles qu’elles apparaissent à travers l’assimilation de nouvelles expériences à des conceptions familières stockées en Mémoire à Long Terme (MLT). Ce mécanisme est envisagé comme un moteur du développement conceptuel chez le jeune enfant.Les trois premières études empiriques ont pour objectif de tester l’hypothèse selon laquelle des concepts abstraits familiers sont utilisées pour comprendre la structure profonde des situations rencontrées quotidiennement et évoquer des expériences passées en se basant sur des similitudes structurelles plutôt que superficielles. Les résultats issus de paradigmes expérimentaux de rappel d’histoires écrites, de rappel de situations filmées et d’évocation libre d’expériences personnelles valident notre hypothèse, dévoilant que les situations structurellement similaires sont plus fréquemment évoquées que les situations superficiellement similaires. Compte tenu du rôle des concepts abstraits dans la compréhension, la quatrième étude aborde la question de leur développement chez le jeune enfant. Nous faisons l’hypothèse que les processus cognitifs et neuronaux impliqués lors du traitement d’approximations sémantiques verbales (ex : « elle déshabille l’orange ») par le jeune enfant de 4 ans reflètent le mécanisme par lequel des catégories lexicales aux frontières immatures sont appliquées par analogie à de nouvelles situations. Conformément à notre prédiction, les Potentiels Évoqués (PE) indiquent que les jeunes enfants détectent l’incongruence (effet N400) de verbes inappropriés, mais pas celles des approximations sémantiques.Les implications et les perspectives émergeant de nos résultats sont discutées dans le cadre d’une approche plaçant la capacité à établir spontanément des rapprochements profonds au centre des mécanismes de compréhension de développement des catégories<br>Analogy is a crucial mechanism allowing one to interpret a new situation in the light of passed experiences. The present dissertation seeks to give this statement back its rightful strength by highlighting the ability to perceive as essentially similar experiences showing a different appearance. Based on the idea that analogies are a natural mechanism through which the cognitive system processes new information, a particular interest will be allotted to their spontaneous expression (i.e. without incitation by someone else to draw the comparison), as their appear through the assimilation of new experiences to familiar concepts stored in long term memory. This mechanism is considered as a motor for young children’s conceptual development.The purpose of the first three empirical studies is to test the hypothesis that familiar concepts are used to understand the deep structure of daily-life situations and to retrieve passed experiences based on structural rather than superficial similarities. Results from experimental paradigms of written story-recall, filmed situation-recall and free-recall of personal experiences confirm our hypothesis, revealing that structurally similar situations are more often retrieved than superficially similar ones. Given the role of abstract concepts in the understanding processes, the fourth study tackles the question of their development during early childhood. Our hypothesis is that cognitive and neural processes involved 4-year-olds’ processing of semantic approximations (e.g. “she is undressing the orange”) reflect the mechanism through which lexical categories with immature boundaries are applied by analogy to new situations. In line with our prediction, event-related potentials’ responses suggest that young children detect incorrect verbs incongruence (N400 effect), but not that of semantic approximations.We discuss implications and perspectives brought about by our results within a framework giving a central place to the ability to draw spontaneous comparisons in understanding processes and categories development

The efficiency of modern search engines depends on how well they present typo-corrected results to a user while typing. So-called fuzzy type-ahead search combines fuzzy string matching and search-as-you-type functionality, and creates a powerful tool for exploring indexed data. Current fuzzy type-ahead search algorithms work well on small data sets, but for big data of social networking services such as Facebook, e-commerce sites such as Amazon, or media streaming services such as YouTube, responsive fuzzy type-ahead search remains a great challenge. This thesis describes a method that enables responsive type-ahead search combined with fuzzy string matching on big data by keeping the search time optimal for human interaction at the expense of lower accuracy for less popular records when a query contains typos. This makes the method effective for e-commerce and media services where the popularity of search terms is a result of human behaviour and thus often follow a power-law distribution.<br>Effektiviteten hos moderna sökmotorer beror på hur väl de presenterar rättstavade resultat för en användare medan en sökning skrivs. Så kallad fuzzy type-ahead sök kombinerar approximativ strängmatchning och sök-medan-du-skriver funktionalitet, vilket skapar ett kraftfullt verktyg för att utforska data. Dagens algoritmer för fuzzy type-ahead sök fungerar väl för små mängder data, men för data i storleksordningen “big data” från t.ex sociala nätverkstjänster så som Facebook, e-handelssidor så som Amazon, eller media tjänster så som YouTube, är en responsiv fuzzy type-ahead sök ännu en stor utmaning. Denna avhandling beskriver en metod som möjliggör responsiv type-ahead sök kombinerat med approximativ strängmatchning för big data genom att hålla söktiden optimal för mänsklig interaktion på bekostnad av lägre precision för mindre populär information när en sök-förfrågan innehåller felstavningar. Detta gör metoden effektiv för e-handel och mediatjänster där populariteten av sök-termer är ett resultat av mänskligt beteende vilket ofta följer en potens-lag distribution.

Abstract Recognition of the importance of a regional perspective for understanding the structure and dynamics of local assemblages has stimulated the emergence of the field of macroecology. Most attention has been directed to terrestrial ecosystems, while large-scale patterns in biodiversity of aquatic organisms have received less attention. In this thesis I examined patterns of aquatic diversity across several geographic areas and scales, in an effort to understand some of the environmental and spatial factors determining species diversity in aquatic environments. The main objectives of this thesis were: (i) to examine the latitudinal diversity patterns of marine crustaceans and molluscs and their relationship to large scale environmental gradients, (ii) to study macroinvertebrate species richness in headwater streams at two spatial extents, within and across drainage systems, and assess the relative importance of local, landscape and regional variables, and (iii) to study diversity patterns of macroorganisms vs microorganism, comparing distance decay patterns of stream diatoms, macroinvertebrates and bryophytes. Latitudinal diversity patterns of crustaceans and molluscs were clearly related to larval developmental mode. An increase in species richness towards high latitudes was found for species with direct development, whereas richness of species with planktotrophic development decreased poleward. Sea surface temperature was the most important environmental gradient related to species richness of both phyla and each developmental mode, but with different effects on each mode. Stream macroinvertebrate species richness at the bioregion extent was negatively related to water humic content. Another factor related to species richness at the bioregion extent was elevation range, a variable linked to stream topographic heterogeneity. Local environmental variables explained most of the variation in species richness at the drainage system extent, however high among-region variability was evident. Patterns between macro- and microorganism may not be fundamentally different, but the level of environmental control varied, being strongest for diatoms, while some groups of benthic macroinvertebrates exhibited relatively strong dispersal limitation. The relative importance of niche vs. dispersal processes is not simply a function of organism size but other traits (e.g. life-history type, dispersal capacity) may obscure this relationship.

Cluster analysis deals with the problem of organization of a collection of patterns into clusters based on a similarity measure. Various distance functions can be used to define this measure. Clustering problems with the similarity measure defined by the squared Euclidean distance have been studied extensively over the last five decades. However, problems with other Minkowski norms have attracted significantly less attention. The use of different similarity measures may help to identify different cluster structures of a data set. This in turn may help to significantly improve the decision making process. High dimensional data visualization is another important task in the field of data mining and pattern recognition. To date, the principal component analysis and the self-organizing maps techniques have been used to solve such problems. In this thesis we develop algorithms for solving clustering problems in large data sets using various similarity measures. Such similarity measures are based on the squared L<br>Doctor of Philosophy

Drug repositioning applies existing drugs to new disease indications. A prerequisite for drug repurposing is drug promiscuity - a drug's ability to bind to several targets, possibly leading to side effects on the other hand. One reason for drug promiscuity is binding site similarity between (otherwise unrelated) proteins. In this thesis, a new algorithm for remote binding site similarity assessment and its application to the whole of the Protein Data Bank (PDB) is presented, forming the base for off-target identification and drug repositioning. The present thesis contributes to a long-standing debate on the reasons for drug promiscuity, being one of the pioneer studies investigating these from a protein structural point of view. Except for a small influence of flexibility, the analysis of all promiscuous drugs in the PDB revealed that drug properties are of minor importance. However, a strong correlation between promiscuity and binding site similarity of protein targets is found (r = 0.81), suggesting binding site similarity as the main reason for drug promiscuity. For 71 % of the promiscuous drugs at least one pair of their targets' binding sites is similar and for 18 % all are similar. In order to overcome issues in detection of remotely similar binding sites, a score for binding site similarity is developed: LigandRMSD measures the similarity of the aligned ligands and uncovers remote local similarities in proteins. It can be applied to arbitrary binding site alignments and also works on distinct ligands on a structural proteome scale. To answer the question on which other targets might be hit when targeting a particular protein, an all-to-all binding site alignment of 32,202 protein structures is analyzed. Of the hundreds of million possible protein pairs, 0.27 % were found to have similar binding sites. Extrapolating to the human proteome, for one human protein are 54 proteins with a similar binding site expected on average. Clearly, this is in contrast to the one drug-one target paradigm in drug development. Based on these data, disadvantageous off-targets can be uncovered and drug-repositioning candidates inferred. The enormous potential is demonstrated with the example of Viagra, proposing it for repositioning to Alzheimer's disease and prostate cancer. The findings in this thesis question the established single-target dogma in drug discovery. Drugs are triggered to modulate multiple targets simultaneously by the widespread binding site similarity. With the presented pipeline, drug targets can be reliably predicted: Starting from a target protein, additional targets are predicted based on binding site similarity and prioritized according to the resulting ligand structural overlap. Identifying drug targets helps to understand severe side effects and opens the door for drug repositioning.

3D city models have been widely used in various applications such as urban planning, traffic control, disaster management etc. Efficient visualisation of 3D city models in different levels of detail (LODs) is one of the pivotal technologies to support these applications. In this thesis, a framework is proposed to visualise the 3D city models online. Then, generalisation methods are studied and tailored to create 3D city scenes in different scales dynamically. Multiple representation structures are designed to preserve the generalisation results on different level. Finally, the quality of the generalised 3D city models is evaluated by measuring the visual similarity with the original models.   In the proposed online visualisation framework, City Geography Makeup Language (CityGML) is used to represent city models, then 3D scenes in Extensible 3D (X3D) are generated from the CityGML data and dynamically updated to the user side for visualisation in the Web-based Graphics Library (WebGL) supported browsers with X3D Document Object Model (X3DOM) technique. The proposed framework can be implemented at the mainstream browsers without specific plugins, but it can only support online 3D city model visualisation in small area. For visualisation of large data volumes, generalisation methods and multiple representation structures are required.   To reduce the 3D data volume, various generalisation methods are investigated to increase the visualisation efficiency. On the city block level, the aggregation and typification methods are improved to simplify the 3D city models. On the street level, buildings are selected according to their visual importance and the results are stored in the indexes for dynamic visualisation. On the building level, a new LOD, shell model, is introduced. It is the exterior shell of LOD3 model, in which the objects such as windows, doors and smaller facilities are projected onto walls.  On the facade level, especially for textured 3D buildings, image processing and analysis methods are employed to compress the texture.   After the generalisation processes on different levels, multiple representation data structures are required to store the generalised models for dynamic visualisation. On the city block level the CityTree, a novel structure to represent group of buildings, is tested for building aggregation. According to the results, the generalised 3D city model creation time is reduced by more than 50% by using the CityTree. Meanwhile, a Minimum Spanning Tree (MST) is employed to detect the linear building group structures in the city models and they are typified with different strategies. On the building level and the street level, the visible building index is created along the road to support building selection. On facade level the TextureTree, a structure to represent building facade texture, is created based on the texture segmentation.   Different generalisation strategies lead to different outcomes. It is critical to evaluate the quality of the generalised models. Visually salient features of the textured building models such as size, colour, height, etc. are employed to calculate the visual difference between the original and the generalised models. Visual similarity is the criterion in the street view level building selection. In this thesis, the visual similarity is evaluated locally and globally. On the local level, the projection area and the colour difference between the original and the generalised models are considered. On the global level, the visual features of the 3D city models are represented by Attributed Relation Graphs (ARG) and their similarity distances are calculated with the Nested Earth Mover’s Distance (NEMD) algorithm.   The overall contribution of this thesis is that 3D city models are generalised in different scales (block, street, building and facade) and the results are stored in multiple representation structures for efficient dynamic visualisation, especially for online visualisation.<br>QC 20111116<br>ViSuCity

Thesis in theoretical part deals with the procedures used in image databases searching. There are discussed two basic possible approaches - text based searching and content based searching. In next section there are described methods for image similarity detection. Practical part deals with detailed description and implementation of three selected image features used for image searching. In third part there are presented testing procedure for implemented algorithms and test results. In conclusion implementation of Rapidminer operator are described. This operator uses all implemented algorithms and allows image similarity matching, searching for most similar images in database, and copy these images to output folder.

Cette thèse traite de la reconnaissance d'entités dans les documents océrisés guidée par une base de données. Une entité peut être, par exemple, une entreprise décrite par son nom, son adresse, son numéro de téléphone, son numéro TVA, etc. ou des méta-données d'un article scientifique tels que son titre, ses auteurs et leurs affiliations, le nom de son journal, etc. Disposant d'un ensemble d'entités structurées sous forme d'enregistrements dans une base de données et d'un document contenant une ou plusieurs de ces entités, nous cherchons à identifier les entités contenues dans le document en utilisant la base de données. Ce travail est motivé par une application industrielle qui vise l'automatisation du traitement des images de documents administratifs arrivant en flux continu. Nous avons abordé ce problème comme un problème de rapprochement entre le contenu du document et celui de la base de données. Les difficultés de cette tâche sont dues à la variabilité de la représentation d'attributs d'entités dans la base et le document et à la présence d'attributs similaires dans des entités différentes. À cela s'ajoutent les redondances d'enregistrements et les erreurs de saisie dans la base de données et l'altération de la structure et du contenu du document, causée par l'OCR. Devant ces problèmes, nous avons opté pour une démarche en deux étapes : la résolution d'entités et la reconnaissance d'entités. La première étape consiste à coupler les enregistrements se référant à une même entité et à les synthétiser dans un modèle entité. Pour ce faire, nous avons proposé une approche supervisée basée sur la combinaison de plusieurs mesures de similarité entre attributs. Ces mesures permettent de tolérer quelques erreurs sur les caractères et de tenir compte des permutations entre termes. La deuxième étape vise à rapprocher les entités mentionnées dans un document avec le modèle entité obtenu. Nous avons procédé par deux manières différentes, l'une utilise le rapprochement par le contenu et l'autre intègre le rapprochement par la structure. Pour le rapprochement par le contenu, nous avons proposé deux méthodes : M-EROCS et ERBL. M-EROCS, une amélioration/adaptation d'une méthode de l'état de l'art, consiste à faire correspondre les blocs de l'OCR avec le modèle entité en se basant sur un score qui tolère les erreurs d'OCR et les variabilités d'attributs. ERBL consiste à étiqueter le document par les attributs d'entités et à regrouper ces labels en entités. Pour le rapprochement par les structures, il s'agit d'exploiter les relations structurelles entre les labels d'une entité pour corriger les erreurs d'étiquetage. La méthode proposée, nommée G-ELSE, consiste à utiliser le rapprochement inexact de graphes attribués modélisant des structures locales, avec un modèle structurel appris pour cet objectif. Cette thèse étant effectuée en collaboration avec la société ITESOFT-Yooz, nous avons expérimenté toutes les étapes proposées sur deux corpus administratifs et un troisième corpus extrait du Web<br>This thesis focuses on entity recognition in documents recognized by OCR, driven by a database. An entity is a homogeneous group of attributes such as an enterprise in a business form described by the name, the address, the contact numbers, etc. or meta-data of a scientific paper representing the title, the authors and their affiliation, etc. Given a database which describes entities by its records and a document which contains one or more entities from this database, we are looking to identify entities in the document using the database. This work is motivated by an industrial application which aims to automate the image document processing, arriving in a continuous stream. We addressed this problem as a matching issue between the document and the database contents. The difficulties of this task are due to the variability of the entity attributes representation in the database and in the document and to the presence of similar attributes in different entities. Added to this are the record redundancy and typing errors in the database, and the alteration of the structure and the content of the document, caused by OCR. To deal with these problems, we opted for a two-step approach: entity resolution and entity recognition. The first step is to link the records referring to the same entity and to synthesize them in an entity model. For this purpose, we proposed a supervised approach based on a combination of several similarity measures between attributes. These measures tolerate character mistakes and take into account the word permutation. The second step aims to match the entities mentioned in documents with the resulting entity model. We proceeded by two different ways, one uses the content matching and the other integrates the structure matching. For the content matching, we proposed two methods: M-EROCS and ERBL. M-EROCS, an improvement / adaptation of a state of the art method, is to match OCR blocks with the entity model based on a score that tolerates the OCR errors and the attribute variability. ERBL is to label the document with the entity attributes and to group these labels into entities. The structure matching is to exploit the structural relationships between the entity labels to correct the mislabeling. The proposed method, called G-ELSE, is based on local structure graph matching with a structural model which is learned for this purpose. This thesis being carried out in collaboration with the ITESOFT-Yooz society, we have experimented all the proposed steps on two administrative corpuses and a third one extracted from the web

Promoter classification is the task of separating promoter from non promoter sequences. Determining promoter regions where the transcription initiation takes place is important for several reasons such as improving genome annotation and defining transcription start sites. In this study, various promoter prediction methods called ProK-means, ProSVM, and 3S1C are proposed. In ProSVM and ProK-means algorithms, structural features of DNA sequences are used to distinguish promoters from non promoters. Obtained results are compared with ProSOM which is an existing promoter prediction method. It is shown that ProSVM is able to achieve greater recall rate compared to ProSOM results. Another promoter prediction methods proposed in this study is 3S1C. The difference of the proposed technique from existing methods is using signal, similarity, structure, and context features of DNA sequences in an integrated way and a hierarchical manner. In addition to current methods related to promoter classification, the similarity feature, which compares the promoter regions between human and other species, is added to the proposed system. We show that the similarity feature improves the accuracy. To classify core promoter regions, firstly, signal, similarity, structure, and context features are extracted and then, these features are classified separately by using Support Vector Machines. Finally, output predictions are combined using multilayer perceptron. The result of 3S1C algorithm is very promising.

People have complex thoughts, and they often express their thoughts with complex sentences using natural languages. This complexity may facilitate efficient communications among the audience with the same knowledge base. But on the other hand, for a different or new audience this composition becomes cumbersome to understand and analyze. Analysis of such compositions using syntactic or semantic measures is a challenging job and defines the base step for natural language processing. In this dissertation I explore and propose a number of new techniques to analyze and visualize the syntactic and semantic patterns of unstructured English texts. The syntactic analysis is done through a proposed visualization technique which categorizes and compares different English compositions based on their different reading complexity metrics. For the semantic analysis I use Latent Semantic Analysis (LSA) to analyze the hidden patterns in complex compositions. I have used this technique to analyze comments from a social visualization web site for detecting the irrelevant ones (e.g., spam). The patterns of collaborations are also studied through statistical analysis. Word sense disambiguation is used to figure out the correct sense of a word in a sentence or composition. Using textual similarity measure, based on the different word similarity measures and word sense disambiguation on collaborative text snippets from social collaborative environment, reveals a direction to untie the knots of complex hidden patterns of collaboration.

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO<br>COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR<br>PROGRAMA DE EXCELENCIA ACADEMICA<br>O presente trabalho apresenta os resultados que obtivemos com a aplicação de grande número de modelos e algoritmos em um determinado conjunto de experimentos de agrupamento de texto. O objetivo de tais testes é determinar quais são as melhores abordagens para processar as grandes massas de informação geradas pelas crescentes demandas de data quality em diversos setores da economia. O processo de deduplicação foi acelerado pela divisão dos conjuntos de dados em subconjuntos de itens similares. No melhor cenário possível, cada subconjunto tem em si todas as ocorrências duplicadas de cada registro, o que leva o nível de erro na formação de cada grupo a zero. Todavia, foi determinada uma taxa de tolerância intrínseca de 5 porcento após o agrupamento. Os experimentos mostram que o tempo de processamento é significativamente menor e a taxa de acerto é de até 98,92 porcento. A melhor relação entre acurácia e desempenho é obtida pela aplicação do algoritmo K-Means com um modelo baseado em trigramas.<br>This document reports our findings on a set of text clusterig experiments, where a wide variety of models and algorithms were applied. The objective of these experiments is to investigate which are the most feasible strategies to process large amounts of information in face of the growing demands on data quality in many fields. The process of deduplication was accelerated through the division of the data set into individual subsets of similar items. In the best case scenario, each subset must contain all duplicates of each produced register, mitigating to zero the cluster s errors. It is established, although, a tolerance of 5 percent after the clustering process. The experiments show that the processing time is significantly lower, showing a 98,92 percent precision. The best accuracy/performance relation is achieved with the K-Means Algorithm using a trigram based model.

Les modèles d'occupation de niche au sein des communautés locales, la variabilité spatiale de la biodiversité le long des gradients environnementaux du stress et des perturbations, et les processus de succession végétale sont plusieurs sujets fondamentaux en écologie. Récemment, l'approche basée sur les traits est apparue comme un moyen prometteur de comprendre les processus structurant les communautés végétales et cette approche a même été proposée comme méthode pour reconstruire l'écologie communautaire en fonction des traits fonctionnels. Par conséquent, lier ces thèmes fondamentaux en utilisant une lentille fonctionnelle devrait nous donner un aperçu de certaines questions fondamentales en écologie et sera l'objectif principal de ma thèse. En général, mon projet de doctorat vise à étudier les structures de l'occupation de l'espace fonctionnel dans les gradients spatio-temporels. Plus précisément, l'objectif du chapitre 2 est (i) d'étudier les modèles d'occupation de la niche fonctionnelle en calculant trois métriques clés de niche (le volume total de niche fonctionnelle , le chevauchement des niches fonctionnelles et le volume de niche fonctionnel moyen) des communautés pauvres en espèces aux communautés riches en espèces et (ii) de déterminer le principal facteur de la structure observée de l'occupation de la niche fonctionnelle dans les communautés végétales du monde entier. Dans le chapitre 3, je vise à prédire et à expliquer la variation de la richesse en espèces selon les gradients de stress et de perturbation, en reliant le modèle d'équilibre dynamique et l'occupation de la niche fonctionnelle en fonction du cadre développé au chapitre 2. L'objectif du chapitre 4 est de tester expérimentalement l’application d'une méthode d'ordination CSR évaluée globalement en fonction de trois traits de feuilles (surface foliaire, teneur en matière sèche des feuilles et surface foliaire spécifique) dans les études locales. Enfin, l'objectif du chapitre 5 est de tester expérimentalement les hypothèses qui concilient les points de vue déterministes et historiquement contingents de la succession végétale, en étudiant la variation des divergences taxonomiques et fonctionnelles entre les communautés selon des gradients de stress et de perturbation. L'étude globale (chapitre 2) est basée sur une collection de 21 jeux de données, couvrant les biomes tropicaux et tempérés, et se compose de 313 communautés végétales représentant différentes formes de croissance. Les études locales (chapitre 3, 4 et 5) sont basées sur le même système expérimental constitué de 24 mésocosmes présentant différents niveaux de stress et de perturbation. L'expérience a commencé en 2009 avec le même mélange de graines de 30 espèces herbacées semées sur les 24 mésocosmes et s'est terminée en 2016. Nous avons permis la colonisation naturelle de graines de la banque commune de graines de sol et de l'environnement pendant la succession de sept ans. Dix traits ont été mesurés sur cinq individus (échantillonnés directement à partir des mésocosmes) par espèce par mésocosme en 2014 (chapitre 3 et 4). Un autre ensemble de traits (16 traits, y compris certains traits qui ne pouvaient pas être mesurés directement dans les mésocosmes), ont été mesurés au niveau de l'espèce (valeurs moyennes des traits) pour les 34 espèces les plus abondantes (certaines espèces disparues dans les mésocosmes) au cours des sept Ans, en les regroupant séparément pour une saison de croissance. Au chapitre 2, nous avons constaté que les communautés étaient plus diverses en termes fonctionnels (une augmentation du volume fonctionnel total) dans les communautés riches en espèces et que les espèces se chevauchaient davantage au sein de la communauté (augmentation du chevauchement fonctionnel), mais ne divisaient pas plus finement l'espace fonctionnel (aucune réduction du volume fonctionnel moyen). En outre, le filtrage de l'habitat est un processus répandu qui conduit à la caractérisation de l'occupation de niche fonctionnelle dans les communautés végétales. Dans le chapitre 3, nous avons trouvé un modèle similaire d'occupation de niche fonctionnelle sur un système expérimental avec une taille spatiale communautaire constante et un effort d'échantillonnage des traits, qui, avec le chapitre 2, nous a fourni une image plus complète et plus solide de l'occupation de niche fonctionnelle dans les communautés végétales. De plus, nous avons réussi à relier le modèle de l'occupation de la niche fonctionnelle et le modèle d'équilibre dynamique et avons constaté que le filtrage concurrentiel était le processus dominant qui détermine le mode d'occupation de la niche fonctionnelle et la richesse des espèces le long du stress et de la perturbation des gradients. Au chapitre 4, nous fournissons un soutien empirique à une méthode d'ordination CSR calibrée globalement en montrant une relation entre l'abondance relative d'espèces en croissance dans les mésocosmes ayant différents niveaux de fertilité du sol et mortalité indépendante de la densité et leur classification CSR. Au chapitre 5, nous avons montré que la succession d'installations au cours de sept ans dans ces mésocosmes était plus déterministe d'un point de vue fonctionnel, mais plus historiquement contingent d'un point de vue taxonomique et que l'importance relative de la contingence historique a diminué à mesure que l'environnement devenait plus stressé ou perturbé. En conclusion, les structures de l'occupation de l'espace fonctionnel dans (le volume fonctionnel total, le chevauchement fonctionnel et le volume fonctionnel moyen, les Chapitre 2 et 3) ou entre les communautés locales (dissimilarité fonctionnelle, chapitre 5) sont déterministes plutôt que neutres (ou contingence historique ). Les espèces tolératrices de stress sont plus avantagées dans les mésocosmes moins fertiles tandis que les espèces rudérales sont plus avantagées dans les mésocosmes avec plus de mortalité indépendante de la densité.<br>Abstract : The patterns of niche occupancy within local communities, the spatial variability of biodiversity along environmental gradients of stress and disturbance, and the processes of plant succession are several fundamental topics in ecology. Recently, the trait-based approach has emerged as a promising way to understand the processes structuring plant communities and has even been proposed as a method to rebuild community ecology based on functional traits. Therefore, linking these fundamental themes through a functional lens should give us more insight into some basic questions in ecology and will be the main objective of my thesis. Generally, my PhD project is to investigate the structures of functional space occupancy along both spatial and temporal gradients. Specifically, the objective of Chapter 2 is to investigate the patterns of functional niche occupancy by calculating three key niche metrics (the total functional niche volume, the functional niche overlap and the average functional niche volume) from speciespoor communities to species-rich communities and to determine the main driver of the observed pattern of functional niche occupancy across plant communities worldwide. In Chapter 3, I aim to predict and explain the variation of species richness along gradients of stress and disturbance, by linking the dynamic equilibrium model and functional niche occupancy based on the framework developed in Chapter 2. The objective of Chapter 4 is to experimentally test the application of a globally calibrated CSR ordination method based on three leaf traits (leaf area, leaf dry matter content and specific leaf area) in local studies. Finally, the aim of Chapter 5 is to experimentally test the hypotheses reconciling the deterministic and historically contingent views of plant succession, by investigating the variation of taxonomic and functional dissimilarities between communities along gradients of stress and disturbance. The global study (Chapter 2) is based on a collection 21 trait datasets, spanning tropical to temperate biomes, and consisting of 313 plant communities representing different growth forms. The local studies (Chapter 3, 4 and 5) are based on the same experimental system consisting of 24 mesocosms experiencing different levels of stress and disturbance. The experiment started in 2009 with the same seed mixture of 30 herbaceous species broadcast over the 24 mesocosms and ended in 2016. We allowed natural colonization of seeds from the common soil seed bank and from the surroundings during the seven-year succession. Ten traits were measured on five individuals (sampled directly from the mesocosms) per species per mesocosms in 2014 (Chapter 3 and 4). Another set of traits (16 traits including some traits that were not able to measured directly in the mesocosms) were measured at the species level (species mean traits values) for the 34 most abundant species (some species disappeared in the mesocosms) over the seven years, by regrowing them separately for one growing season. In Chapter 2, we found communities were more functionally diverse (an increase in total functional volume) in species-rich communities, and species overlapped more within the community (an increase in functional overlap) but did not more finely divide the functional space (no decline in average functional volume). Moreover, habitat filtering is a widespread process driving the pattern of functional niche occupancy across plant communities. In Chapter 3, we found a similar pattern of functional niche occupancy on an experimental system with a constant community spatial size and trait-sampling effort, which together with Chapter 2 provided us a more comprehensive and robust picture of functional niche occupancy across plant communities. In addition, we succeeded in linking the pattern of functional niche occupancy and the dynamic equilibrium model and found that habitat filtering was the dominant process determining the pattern of functional niche occupancy and species richness along the gradients stress and disturbance. In Chapter 4, we provide empirical support for a globally calibrated CSR ordination method by showing a relationship between the relative abundance of species growing in mesocosms having different levels of soil fertility and density-independent mortality and their CSR classification. In Chapter 5, we showed that plant succession over seven years in these mesocosms was more deterministic from a functional perspective but more historically contingent from a taxonomic perspective, and that the relative importance of historical contingency decreased as the environment became more stressed or disturbed. In conclusion, the structures of functional space occupancy within (the total functional volume, the functional overlap and the average functional volume; Chapter 2 and 3) or between local communities (functional dissimilarity, Chapter 5) are deterministic rather than neutral (or historical contingency). Stress-tolerators were more favored in high stress communities, while ruderals are more favored in high disturbed mesocosms (Chapter 4).

Learning to write software requires much practice and frequent assessment. Consequently, the use of computers to assist in the assessment of computer programs has been important in supporting large classes at universities. The main approaches to the problem are dynamic analysis (testing student programs for expected output) and static analysis (direct analysis of the program code). The former is very sensitive to all kinds of errors in student programs, while the latter has traditionally only been used to assess quality, and not correctness. This research focusses on the application of static analysis, particularly structural similarity, to marking student programs. Existing traditional measures of similarity are limiting in that they are usually only effective on tree structures. In this regard they do not easily support dependencies in program code. Contemporary measures of structural similarity, such as similarity flooding, usually rely on an internal normalisation of scores. The effect is that the scores only have relative meaning, and cannot be interpreted in isolation, ie. they are not meaningful for assessment. The SimRank measure is shown to have the same problem, but not because of normalisation. The problem with the SimRank measure arises from the fact that its scores depend on all possible mappings between the children of vertices being compared. The main contribution of this research is a novel graph similarity measure, the Weighted Assignment Similarity measure. It is related to SimRank, but derives propagation scores from only the locally optimal mapping between child vertices. The resulting similarity scores may be regarded as the percentage of mutual coverage between graphs. The measure is proven to converge for all directed acyclic graphs, and an efficient implementation is outlined for this case. Attributes on graph vertices and edges are often used to capture domain specific information which is not structural in nature. It has been suggested that these should influence the similarity propagation, but no clear method for doing this has been reported. The second important contribution of this research is a general method for incorporating these local attribute similarities into the larger similarity propagation method. An example of attributes in program graphs are identifier names. The choice of identifiers in programs is arbitrary as they are purely symbolic. A problem facing any comparison between programs is that they are unlikely to use the same set of identifiers. This problem indicates that a mapping between the identifier sets is required. The third contribution of this research is a method for applying the structural similarity measure in a two step process to find an optimal identifier mapping. This approach is both novel and valuable as it cleverly reuses the similarity measure as an existing resource. In general, programming assignments allow a large variety of solutions. Assessing student programs through structural similarity is only feasible if the diversity in the solution space can be addressed. This study narrows program diversity through a set of semantic preserving program transformations that convert programs into a normal form. The application of the Weighted Assignment Similarity measure to marking student programs is investigated, and strong correlations are found with the human marker. It is shown that the most accurate assessment requires that programs not only be compared with a set of good solutions, but rather a mixed set of programs of varying levels of correctness. This research represents the first documented successful application of structural similarity to the marking of student programs.

This thesis addresses problems involving structural alignments for similarity detection between entities. In the general computational context, a structural alignment is defined as an optimization problem where representative inputs are assigned to relative positions subject to the minimization of some objective function. The output is an inferred relationship based upon the resultant value of the objective function, and/or the arrangement of aligned positions. Two bioinformatics similarity detection applications were used as case studies within this work, the structural alignment of biomolecular proteins and the document similarity detection problem in biomedical literature. The structural alignment of protein biomolecules involves generating residue pair correspondences of maximal overlap with minimal geometric divergence using each protein’s set of three-dimensional atomic coordinates. As protein structure decides its functionality, similarity in structure usually implies similarity in function. During the investigation of this structural alignment problem, it became apparent that a fast and approximate asymmetric linear sum assignment algorithm was required. Accordingly, a new heuristic algorithm, Asymmetric Greedy Search (AGS), was developed. Extensive computational experiments using a range of model graphs demonstrated the effectiveness of the algorithm. In addition, a new type of deterministic model graph that is suitable for reproducible benchmarking of these types of algorithms was also developed. Incorporating AGS, a new non-sequential protein structure alignment method, SPalignNS, was then developed. As compared to existing methods, SPalignNS achieved greater alignment accuracy with commonly used protein alignment test datasets, and also achieved the highest agreement with manually curated reference alignments. The document similarity detection problem is a fundamental application of natural language processing, and constitutes the basis of information retrieval systems. Document matching systems for locating relevant literature have mostly relied on methods developed over a decade ago, largely due to the unavailability of a common evaluation framework. A database of relevance annotations for over 180,000 PubMed-listed document pairs was developed with a subsequent application in training a sentence-based transferred learning model, HuBERT (Hierarchical PubMed BERT). When applied to relevant biomedical literature searches in PubMed, the new HuBERT method produced superior results compared to those attained by the baseline methods from existing document matching systems.<br>Thesis (PhD Doctorate)<br>Doctor of Philosophy (PhD)<br>School of Info & Comm Tech<br>Science, Environment, Engineering and Technology<br>Full Text

Una de les àrees més prometedores en recerca biomèdica i farmacèutica és el disseny molecular computacional, que intenta establir relacions entre propietats físico-químiques i activitat biològica. <br/>L'èxit d'aquestes tècniques depen críticament de la qualitat de la descripció molecular. En aquest sentit, metodologies basades en potencials d'interacció molecular (MIP) són eines útils per la comparació de compostos que presenten comportaments biològics semblants. <br/>Aquest projecte desenvolupa eines per comparar molècules basades en la caracterització de llurs MIPs. El programa de similaritat molecular MIPsim ha estat desenvolupat i aplicat a diferents problemes biològics. <br/>Aquesta tesi consisteix en quatre estudis científics que mostren l'ús del MIPSim en aliniament molecular, catalisi enzimàtica, en acoratge de molècules dins el lligand i en estudis 3D-QSAR.<br>One of the most promising areas in biomedical and pharmaceutical research is computer assisted molecular design, which tries to stablish relationships between physicochemical properties and biological activity. <br/>The success of these techniques depends critically on the quality of the molecular description. In this sense, methodologies based on molecular interaction potentials (MIP) are useful tools for the comparison of compounds displaying related biological behaviours. <br/>This project aims to develop tools to compare 'molecules based on the characterization 'of their MIPs. To this end, the molecular similarity program MIPSim has been further developed and applied to different biological problems. <br/>This thesis consists on four scientific studies showing the use of MIPSim for molecular alignment, enzymatic catalysis, ligand-protein docking and 3D-QSAR analyses.

Ma thèse introduit cinq logiciels de trois différents domaines: le calcul parallèle et distribué, la biologie structurale computationnelle et la chémoinformatique. Le logiciel pour le calcul parallèle et distribué s'appelle PAR. PAR permet d'exécuter des expériences indépendantes de manière parallèle et distribuée. Les logiciels pour la biologie structurale computationnelle sont Durandal, EleKit et Fragger. Durandal exploite la propagation de contraintes géométriques afin d'accélérer l'algorithme de partitionnement exact pour des modèles de protéines. EleKit permet de mesurer la similarité électrostatique entre une petite molécule et la protéine qu'elle est conçue pour remplacer sur une interface protéine-protéine. Fragger est un cueilleur de fragments de protéines permettant de sélectionner des fragments dans la banque de protéines mondiale. Enfin, le logiciel de chémoinformatique est ACPC. ACPC permet l'encodage fin, d'une manière rotation-translation invariante, d'une molécule dans un ou une combinaison des trois espaces chimiques (électrostatique, stérique ou hydrophobe). ACPC est un outil de criblage virtuel qui supporte les requêtes consensus, l'annotation de la molécule requête et les processeurs multi-coeurs<br>This thesis introduces five software useful in three different areas : parallel and distributed computing, computational structural biology and chemoinformatics. The software from the parallel and distributed area is PAR. PAR allows to execute independent experiments in a parallel and distributed way. The software for computational structural biology are Durandal, EleKit and Fragger. Durandal exploits the propagation of geometric constraints to accelerate the exact clustering algorithm for protein models. EleKit allows to measure the electrostatic similarity between a chemical molecule and the protein it is designed to replace at a protein-protein interface. Fragger is a fragment picker able to select protein fragments in the whole protein data-bank. Finally, the chemoinformatics software is ACPC. ACPC encodes in a rotation-translation invariant way a chemical molecule in any or a combination of three chemical spaces (electrostatic, steric or hydrophobic). ACPC is a ligand-based virtual screening tool supporting consensus queries, query molecule annotation and multi-core computers

In this thesis, we propose many developments in the context of Structural Similarity. We address both node (local) similarity and graph (global) similarity. Concerning node similarity, we focus on improving the diffusive process leading to compute this similarity (e.g. Commute Times) by means of modifying or rewiring the structure of the graph (Graph Densification), although some advances in Laplacian-based ranking are also included in this document. Graph Densification is a particular case of what we call graph rewiring, i.e. a novel field (similar to image processing) where input graphs are rewired to be better conditioned for the subsequent pattern recognition tasks (e.g. clustering). In the thesis, we contribute with an scalable an effective method driven by Dirichlet processes. We propose both a completely unsupervised and a semi-supervised approach for Dirichlet densification. We also contribute with new random walkers (Return Random Walks) that are useful structural filters as well as asymmetry detectors in directed brain networks used to make early predictions of Alzheimer's disease (AD). Graph similarity is addressed by means of designing structural information channels as a means of measuring the Mutual Information between graphs. To this end, we first embed the graphs by means of Commute Times. Commute times embeddings have good properties for Delaunay triangulations (the typical representation for Graph Matching in computer vision). This means that these embeddings can act as encoders in the channel as well as decoders (since they are invertible). Consequently, structural noise can be modelled by the deformation introduced in one of the manifolds to fit the other one. This methodology leads to a very high discriminative similarity measure, since the Mutual Information is measured on the manifolds (vectorial domain) through copulas and bypass entropy estimators. This is consistent with the methodology of decoupling the measurement of graph similarity in two steps: a) linearizing the Quadratic Assignment Problem (QAP) by means of the embedding trick, and b) measuring similarity in vector spaces. The QAP problem is also investigated in this thesis. More precisely, we analyze the behaviour of $m$-best Graph Matching methods. These methods usually start by a couple of best solutions and then expand locally the search space by excluding previous clamped variables. The next variable to clamp is usually selected randomly, but we show that this reduces the performance when structural noise arises (outliers). Alternatively, we propose several heuristics for spanning the search space and evaluate all of them, showing that they are usually better than random selection. These heuristics are particularly interesting because they exploit the structure of the affinity matrix. Efficiency is improved as well. Concerning the application domains explored in this thesis we focus on object recognition (graph similarity), clustering (rewiring), compression/decompression of graphs (links with Extremal Graph Theory), 3D shape simplification (sparsification) and early prediction of AD.<br>Ministerio de Economía, Industria y Competitividad (Referencia TIN2012-32839 BES-2013-064482)

La comparación de imágenes es un ingrediente fundamental en muchos problemas de procesamiento de imagen y visión por computador. Esta tesis aborda el problema de la comparación de entornos locales en imágenes, o patches, por medio de medidas de similitud (o funciones distancia). En particular, estudiamos el problema de la comparación invariante afín de imágenes a partir de sus patches, lo cual abre la puerta a un análisis más profundo de la estructura de similitud y auto-similitud existente en imágenes naturales. Nuestro trabajo parte de una aproximación axiomática reciente a las medidas de similitud entre imágenes definidas en variedades de Riemann. Empezamos obteniendo y estudiando medidas de similitud afín invariantes para después construir con ellas dos nuevos métodos. El objetivo del primero de ellos es la reconstrucción o completación plausible de regiones de una imagen donde la información se ha perdido, dañado o está oculta. El modelo propuesto es capaz de reconstruir texturas con distorsión perspectiva o incluso más compleja. El segundo método extiende la aproximación denominada de Non-Local Means para el problema de eliminación de ruido en imágenes aprovechando la auto-similitud invariante afín de lasimágenes reales. Nuestra extensión es comparada con éxito con el método original, tanto cualitativa como cuantitativamente, y se obtienen resultados prometedores en comparación con los métodos del estado del arte.<br>Image comparison is a main ingredient in many image processing and computer vision problems and applications, and not surprisingly it is a very diverse topic. The subject of this thesis is the comparison of local patches of images by means of similarity measures (or distance functions). In particular, we are interested in affine invariant patch-wise image comparison which opens the door to a more thorough analysis of similarities and self-similarities present in natural images. Our work is based on a recently proposed axiomatic framework for similarity measures between images defined on Riemannian manifolds. At the beginning we derive and study some affine invariant similarity measures and then present two novel methods built around them. The first method for exemplar-based image inpainting is aimed at the recovery of occluded, missing or corrupted parts of an image, in such a way that the reconstructed image looks natural. It is capable of reconstructing textures under perspective or even more complex distortions. The second method extends the well-known Non-Local Means approach for image denoising by taking advantage of affine invariant self-similarities of real images. Our extension improves the original method in both quantitative and qualitative assessments, and the results are promising when compared with state-of-the-art methods.

The study aimed to characterize the tree-shrub community of a riparian forest, one of the last remnatns on the Araguari River, at Uberlândia, Triângulo Mineiro region, Minas Gerais State. The composition, horizontal and vertical structure were investigated by phytosociological survey of 110 plots of 10m x 10m (1,1ha). All trees with CAP ` 15 cm were registered. There were sampled 1.393 trees distributed in 89 species and 36 families. Hirtella gracilipes (Hook.f.) Prance was the most abundant specie, with the highest frequency, density and VI. It is a small tree specie, predominant in the understory of the forest. The Shannon index was H = 3,65 nats.ind-1. The most common soil in the study area was classified as Litholic Neosol dystrophic, with gneiss rock outcrops, and in some parts of the forest there were soils of the Haplic Cambisol Tb dystrophic type. The floristic comparisons indicated the riparian forest of Araguari river is more similar to the Triângulo Mineiro forests, to riparian forest in Três Marias, MG, and to gallery forests of the Central Brazil, than to riparian forests of southsoutheast of Minas Gerais and São Paulo. The analysis of maximum, medium and minimum heights of the species showed the forest has three vertical strata. The stratum 1 (1,5m > h ` 10m) had the highest richness, density and was represented by a group shade-tolerant and light-demanding understory species. There were also individuals juvenile of species belonging to the higher strata, classified as light-demanding canopy species. The stratum 2 (10m > h ` 15m) occupied the largest part of the study area, and was considered mostly a intermediate stratum, including canopy species as well as some understory mature trees. The stratum 3 (h > 15m) had the lowest density and the species which form this part of the forest have far greater number of individuals in inferior strata. This indicates the forest is recovering from anthropic disturbance suffered in the past. The presence of a major quantity of animaldispersed species and individuals emphasizes the important preservation of this riparian forest for the fauna of Araguari River Valley and whole region.<br>O estudo objetivou caracterizar a comunidade arbóreo-arbustiva de um dos últimos remanescentes de mata ciliar do rio Araguari, situado no município de Uberlândia MG, região do Triângulo Mineiro. A composição, estrutura horizontal e vertical foi investigada pelo levantamento sociológico de 110 parcelas de 10 m x 10 m (1,1 ha). Todos os indivíduos com CAP ` 15 cm foram registrados. Foram amostrados 1.393 indivíduos distribuídos em 89 espécies e 36 famílias. Hirtella gracilipes (Hook.f.) Prance, com pequeno porte e predominante no sub-bosque da mata foi a espécie mais abundante, com os maiores valores de freqüência, densidade e VI. O índice de Shannon foi de H = 3,65 nats/ind. O solo com maior ocorrência na área de estudo foi classificado como Neossolo Litólico distrófico, com afloramentos de rocha do tipo gnaisse, e em alguns trechos da mata ocorreram solos do tipo Cambissolo Háplico Tb distrófico. As comparações florísticas indicam que a mata ciliar do rio Araguari é mais semelhante às florestas do Triângulo Mineiro, juntamente com a mata ciliar de Três Marias, MG e com as matas de galeria do Centro-Oeste do Brasil, do que as florestas ciliares do sul-sudeste de Minas Gerais e do Estado de São Paulo. A análise das alturas máxima, média e mínima das espécies mostrou que a floresta possui três estratos verticais. O estrato 1 (1,5m > h ` 10m) teve a maior riqueza, densidade e foi representado por um conjunto de espécies umbrófilas e heliófilas de sub-bosque, além de indivíduos de espécies que alcançaram os estratos superiores da floresta, classificadas como heliófilas de dossel . O estrato 2 (10m > h ` 15m) ocupa a maior parte da área de estudo e foi considerado, principalmente, um estrato de passagem, composto por espécies que chegam aos estratos superiores como também por alguns indivíduos de espécies do sub-bosque que possuem maior porte. O estrato 3 ( h > 15m) teve a menor densidade e as espécies que compõe esta camada da mata possui um grande número de indivíduos nos estratos inferiores. Isto indica que a floresta está se recuperando de perturbações antrópicas sofridas no passado. A ocorrência de um grande número de espécies e indivíduos com dispersão zoocórica ressalta a importância da preservação desta mata ciliar para a fauna do vale do rio Araguari e região.<br>Mestre em Ecologia e Conservação de Recursos Naturais