Made available in DSpace on 2015-04-11T14:03:19Z (GMT). No. of bitstreams: 1 Vitor Bremgartner da Frota.pdf: 2212741 bytes, checksum: d795bab086a9fe7a83cb7bac13963dc1 (MD5) Previous issue date: 2012-02-29 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Com a disseminação dos cursos de Educação a Distância, um problema cada vez mais frequente é a falta de um acompanhamento personalizado ao aluno e a demora em responder por parte de mediadores e demais colegas às dúvidas e requisições dos alunos em Ambientes Virtuais de Aprendizagem (AVAs), normalmente postadas em fóruns de discussão ou manifestada via e-mails. A abordagem adotada para solução deste problema apresentada nesta dissertação baseia-se em Sistemas Multiagente e em uma ontologia estendida da especificação IMS Learner Information Package (LIP) a partir da teoria de habilidades e competências desenvolvida pelo sociólogo suíço Phillipe Perrenoud. Por meio dos agentes e da ontologia são detectadas as eventuais dúvidas dos alunos e estas são direcionadas aos membros da comunidade que possuem o perfil mais adequado a solucioná-la, isto é, para aqueles que possuem as habilidades e competências adequadas, diminuindo a demora de resposta às dúvidas dos alunos. Dessa forma, os alunos poderão interagir entre si com o propósito de solucionar eventuais dúvidas ou erros em determinadas atividades, compartilhando conhecimentos. A solução descrita nesta dissertação se baseia na aprendizagem colaborativa pela interação entre aprendizes, na qual é uma estratégia empregada com o propósito de construir conhecimento de maneira mais significativa em um AVA. O processo de recomendação personalizada é realizado por agentes que utilizam a ontologia. Para validação do sistema, foi desenvolvida uma Rede de Petri que representa as interações entre os agentes, além de simular a correta ordem de execução ou o paralelismo entre eles. A partir dessas concepções, foi implementado um SMA que utiliza a ontologia de modelo de aluno desenvolvida e este foi empregado para auxiliar alunos e tutores em um curso de Cálculo Numérico que faz uso do AVA Moodle. Os resultados obtidos com os testes aplicados no sistema comprovam a validade e a viabilidade da solução encontrada, composta de duas etapas: uma simulação em uma turma fictícia, a fim de avaliar a eficácia dos agentes e da ontologia e outro teste em uma turma real, com o objetivo de avaliar a eficácia do processo de recomendação entre os alunos. Questionários de pesquisa de opinião foram passados em sala de aula a fim de obter e avaliar as impressões dos alunos quanto ao recurso disponível aos mesmos no AVA.

A livre disponibilizaÃÃo de conteÃdos digitais vem crescendo de maneira expressiva na Web. Muitos desses conteÃdos podem ser modificados, reutilizados e adaptados para propÃsitos especÃficos. Embora muitos recursos para armazenamento e disponibilizaÃÃo de conteÃdos sejam bastante difundidos, nÃo se percebem na Internet repositÃrios que possuam instrumentos apropriados para o controle de versÃes de conteÃdo binÃrio (CB). A maioria das soluÃÃes encontradas para a persistÃncia de dados na Web à baseada em Bancos de Dados Relacionais (BDR). A simples adiÃÃo de modificaÃÃes aos dados originais e armazenamento em novos registros de tabelas em um BD pode ser ineficiente devido a uma quantidade potencialmente grande de informaÃÃes redundantes. à possÃvel, para este fim, adaptar ferramentas de versionamento conhecidas por Sistemas de Controle de VersÃo (SCV), ferramentas especializadas no armazenamento de regiÃes modificadas de documentos. Entretanto, um SCV à menos eficiente do que os Sistemas Gerenciadores de Bancos de Dados (SGBD) em tempo de acesso e recuperaÃÃo de informaÃÃes, o que pode comprometer o desempenho da aplicaÃÃo, caso seja adotado como soluÃÃo para a persistÃncia de dados. Visando conceber um repositÃrio de CBs versionados na Web, com gestÃo eficiente tanto para o acesso como para o armazenamento de CBs, neste trabalho à analisado o desempenho dos SGBDs gratuitos mais utilizados na atualidade e de um SVC que se revelou o mais adequado ao escopo deste trabalho. Os resultados obtidos serviram de base para a especificaÃÃo da arquitetura de um repositÃrio que se apoia em uma abordagem hÃbrida, com o uso simultÃneo de um SGBD e um SVC. Em seguida, foi realizada a modelagem por Redes de Petri Coloridas, o que permitiu a simulaÃÃo e a anÃlise da arquitetura concebida, demonstrando-se a maior eficiÃncia da arquitetura proposta em relaÃÃo a uma abordagem de armazenamento tradicional. / The free availability of digital content has fairly increased on the Web. Many of these contents can be modified, reused and adapted for specific purposes. Although diferent resources for providing and storing contents are widely available, there is a lack of tools suitable to versioning control of binary content on the Internet. In addition, most solutions to persistent data on the Web are based on Relational DataBase (RDB). As a simple solution, we can modify the original data and store new records in the tables of a database. However, this leads to a possible ineficiency due to a potentially large amount of redundant information. Nevertheless, in order to overcome this issue, it is possible to adapt versioning tools, also known as Control Version System (CVS), to the solution. These tools are specialized in storage of the modified regions of documents. This process is known as "deltification". However, regarding access time and gathering data, a CVS is less eficient than a DataBase Management System (DBMS). Therefore, if a CVS is used as a solution for data persistence, this can reduce the overall performance of the application.Aiming at designing a repository of versioned binary content on theWeb, with eficient management for both accessing and storing binary data, this work analyzes the performance of free DBMSs most frequently used and a CVS, which we consider being the most suitable to the addressed repository. The attained results were the basis for the specification of the architecture of a repository that relies on a hybrid approach. The resulting approach lies in the simultaneous use of a DBMS and a CVS. We took into account features and runtime performance of both tools for each operation required in the final application. Then, we designed models on Colored Petri Nets, which allowed the simulation and analysis of the aimed architecture. As a result, we present the eficiency of the proposed architecture against a traditional storage approach.

Apresenta uma análise da estimação de custos em sistemas de manufatura utilizando rede de Petri (PN), ferramenta gráfica e matemática para modelagem e simulação. Esta análise permite a apresentação de um sistema aplicável à programação e otimização de processos de fabricação aliadas à estimação dos custos ao longo do processo. Para uma determinada peça, a partir de conceitos sobre custeio de produtos, é proposto um modelo que visa conectar as transições da PN com as informações de custos de produção. O estudo de caso mostra que informações importantes à tomada de decisão podem ser obtidas através do uso de uma metodologia que incorpore planejamento de processos, métodos de custeio e redes de Petri. Finalmente, traz sugestão de um trabalho subseqüente: a partir do modelo proposto, implementar um template no editor/simulador de redes de Petri, Petri Net Tools 2000, capaz de realizar os cálculos necessários para estimativas de custo e fornecer resultados relevantes para um planejamento real. / This work presents a cost estimation analysis in manufacturing systems based in Petri net (PN), which is a graphical and mathematical tool. This analysis allows the presentation of a system, applicable to the programming and optimizing of manufacturing processes joined with cost estimation. A cost model is proposed using concepts of product costing. It seeks for connecting PN transitions to the data manufacturing cost of a certain part. The case study shows that important information to decision support can be found through the use of a methodology which joins process planning, costing systems and Petri nets. There is also a suggestion of a later work: a template implementation, on Petri Net Tools 2000 editor/simulator of PN models, which is able to accomplish the necessary calculations for cost estimates and, to supply important results to a real planning.

O design de sistemas automatizados inteligentes é fortemente dependente da etapa inicial de análise de requisitos, que além de suprimir possíveis incongruências - próprias desta fase inicial - provê um modelo inicial e funcional do sistema a ser implementado, capaz de orientar a definição dos parâmetros (design parameters) e em seguida a própria implementação. Um grande esforço tem sido empregado na área de Inteligência Artificial para definir planejadores automáticos confiáveis que possam ser usados na solução de problemas reais, que geralmente possuem um número elevado de parâmetros. Isto leva a uma situação onde métodos formais, geralmente aplicados em abordagens independentes de domínio, precisam ser aliados a métodos mais pragmáticos para produzir bons resultados. Seguindo esse princípio,o presente trabalho propõe um método de projeto no qual o usuário podem fazer a aquisição de conhecimento, modelar o domínio (tanto o domínio de aplicação quanto o domínio do problema de planejamento), fazer uma análise dinâmica do modelo e eventualmente verificá-lo usando linguagens conhecidas como UML, as Redes de Petri, e HTN, mantendo as características hierárquicas do problema. Esta abordagem demanda novos planejadores automáticos que consideram a abstração do sistema hierárquico, que é derivado de um modelo hierárquico de requisitos e de uma análise unificada feita também em redes de Petri hierárquicas. / The design of intelligent systems is strongly dependent of the requirements analysis initial stage, wich can remove possible inconsistencies - own this early stage - providing a primary functional model of the system being implemented. Besides that, it can be able to guide the definition of the parameters (design parameters) and then the implementation itself. A lot of effort has been employed in Artificial Intelligence field to define reliable automated planning systems that can be used to solve real problems, which generally have a large number of parameters. This leads to a situation where formal methods, generally applied to solve domain-independent problems, need to be combined with more pragmatic methods to produce good results. Following this principle, this work proposes a design method in which the user can make the acquisition of knowledge, model the domain (both the work domain and the planning problem domain), make a dynamic analysis of the model and eventually simulate it using well-kown language as UML, Petri Nets, and HTN, maintaining the hierarchical characteristics of the problem. This approach demands new automated planners who consider the abstraction of hierarchical system, which is derived from a hierarchical model requirements and an unified analysis made in hierarchical Petri nets.

O comércio de retalhos e resíduos têxteis está localizado nas ruas do Brás, antigo bairro industrial e operário da cidade de São Paulo. Esse bairro, hoje, concentra um grande número de indústrias e lojas de confecções, que vendem no atacado e no varejo para \"sacoleiras\" de todas as partes do Brasil. O comércio de retalhos vai nascer nos interstícios das antigas indústrias têxteis e, posteriormente, se alimentar do rejeito da indústria da confecção que fornece, diariamente, toneladas de resíduos e retalhos para serem comercializados pelos \"retalheiros\". Uma parte desses retalhos e rejeitos é comprada por costureiras da Grande São Paulo e até mesmo por \"sacoleiras\", sendo que a maior parte dessa mercadoria é enviada para Santa Cruz do Capibaribe - cidade do interior pernambucano, que constitui em pólo de confecções de \"sulanca\". São vestuários de qualidade considerada inferior, consumidos, predominantemente, por uma população de baixa renda. O nordestino de menor poder aquisitivo passa a ser consumidor do rejeito das confecções do Centro Sul. O comércio de retalhos e resíduos é controlado, predominantemente, por migrantes nordestinos, formando uma rede de pessoas e mercadorias em torno dessa atividade. Desta forma, procuramos estudar como tais migrantes foram mobilizados para trabalhar nesse comércio nas ruas do Brás. Estudamos as transformações do bairro, desde a ocupação dos italianos até a chegada dos nordestinos, no sentido de entender a presença do comércio em suas ruas. Ao identificar as conexões que esse espaço mantém com outros espaços, construindo uma malha de homens e mercadorias, optamos por seguir a circulação da mercadoria - retalho até o seu destino final, qual seja, as feiras da sulanca de Pernambuco. Dentre as múltiplas problematizações que a pesquisa traz à tona, destacamos as redes sociais como um instrumento mobilizador do migrante comerciante autônomo de retalhos contribuindo para a ampliação do capital / Trading cloth patches and textile remainders is carried out in the streets of the Brás neighborhood, an industrial and working class old district in São Paulo. It now concentrates several industries as well as wholesale and retail clothing stores meant for \"sacoleiras\" coming from all over Brazil. The patch trade originates in the interstices of old textile industries. Later it feeds on refuses of clothing industries that daily supply tons of remainders and patches to be traded by the \"retalheiros\". These patches and refuses are partly bought both by sewing women in greater São Paulo and by \"sacoleiras\". Most goods are sent to Santa Cruz do Capibaribe, in the hinterlands of Pernambuco State, which is a \"sulanca\" manufacturing pole. It merely consists of poorer quality clothing, mostly consumed by a lower income population. The lower income northeastern citizen becomes a buyer of fabric refuses from the Southern Center. The patch and remainder trade is mostly controlled by northeastern migrants, who form a net of people and merchandise based on this activity. Hence, this dissertation aimed at studying how these migrants were mobilized to work in this trade in the streets of Brás. In order to understand this process, the work also investigated the changes undergone by the district, from the Italian to the northeastern Brazilian occupation on the area. Once the connection among this and other areas was identified by the people/goods net evolution the patch-goods circulation was followed to one of its ends, namely the \"sulanca\" fairs in Pernambuco. Among the manifold questions raised by this research, attention should be drawn to the social nets as a mobilizing means of the free-lance migrant, who trades patches that contributes to capital increase

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-04-25T11:55:26Z No. of bitstreams: 1 wolmararaujoneto.pdf: 9045744 bytes, checksum: 685167b5aabc68d680681643c44b7a4e (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-04-25T12:12:34Z (GMT) No. of bitstreams: 1 wolmararaujoneto.pdf: 9045744 bytes, checksum: 685167b5aabc68d680681643c44b7a4e (MD5) / Made available in DSpace on 2017-04-25T12:12:34Z (GMT). No. of bitstreams: 1 wolmararaujoneto.pdf: 9045744 bytes, checksum: 685167b5aabc68d680681643c44b7a4e (MD5) Previous issue date: 2014-08-25 / CNPq - Conselho Nacional de Desenvolvimento Científico e Tecnológico / A presente dissertação tem como objetivo principal desenvolver uma metodologia capaz de auxiliar no projeto e implementação de controles para sistemas a eventos discretos (SED) de forma a otimizar a modelagem, documentação e implantação da planta estudada. Para tanto a definição tradicional de autômato foi alterado para incluir um novo subconjunto de estados obrigatórios, isto é, estados que o sistema deve passar antes de chegar a um estado marcado. Seguindo a modelagem tradicional, cada subsistema é modelado através desta nova abordagem. Em uma segunda etapa, uma nova iteração de análise de cada subsistema é realizada para restringir situações indesejadas do sistema. Com esta fase finalizada os subsistemas são mesclados através de uma técnica tradicional de autômatos em paralelo. O sistema final resultante é apresentado a um otimizador, baseado em PROLOG, que busca a melhor sequência de eventos obedecendo a todas as restrições modeladas e termine no estado marcado informado pelo projeto. Este resultado fornece a Rede de Petri ideal do sistema e sua programação em GRAFCET. Tais métodos serão testados em componentes de um sistema de manufatura flexível real. / The present master thesis has as its main goal to develop a methodology able to assist in design and implementation of discrete event system (DES) controllers in order to optimize the modeling, documentation and deployment of plant in study. To do so the automaton traditional concept was changed to include a new subset of mandatory states, i.e. states that the system should pass before reach a marked state. Following the traditional modeling, each subsystem is modeled using this new approach. In a second step, a new iteration of analysis of each subsystem is performed to restrain system’s unwanted situations. At the end of this step the subsystems are composed by a traditional parallel automatons technique. The resulting system is presented to an optimizer, based in PROLOG, which searches the best event sequence obeying to all modeled restrictions and ends in the marked state defined in design. This result provides the ideal Petri Net of the system and its GRAFCET programming. Such methods will be tested in a flexible real manufacturing system.

In this thesis we consider extensions and generalizations of some classical problems in Combinatorial Geometry. Our work is an offshoot of four classical problems in Combinatorial Geometry. A fundamental assumption in these problems is that the underlying point set is R2. Two fundamental themes entwining the problems considered in this thesis are: “What happens if we assume that the underlying point set is finite?”, “What happens if we assume that the underlying point set has a special structure?”. Let P ⊂ R2 be a finite set of points in general position. It is reasonable to expect that if |P| is large then certain ‘patterns’ in P always occur. One of the first results was the Erd˝os-Szekeres Theorem which showed that there exists a f(n) such that if |P| ≥ f(n) then there exists a convex subset S ⊆ P, |S| = n i.e., a subset which is a convex polygon of size n. A considerable number of such results have been found since. Avis et al. in 2001 posed the following question which we call the n-interior point problem: Is there a finite integer g(n) for every n, such that, every point set P with g(n) interior points has a convex subset S ⊆ P with n interior points. i.e. a subset which is a convex polygon that contains exactly n interior points. They showed that g(1) = 1, g(2) = 4. While it is known that g(3) = 9, it is not known whether g(n) exists for n ≥ 4. In the first part of this thesis, we give a positive solution to the n-interior point problem for point sets with bounded number of convex layers. We say a family of geometric objects C in Rd has the (l, k)-property if every subfamily C′ ⊆ C of cardinality at most l is k-piercable. Danzer and Gr¨unbaum posed the following fundamental question which can be considered as a generalised version of Helly’s theorem: For every positive integer k, does there exist a finite g(k, d) such that if any family of convex objects C in Rd has the (g(k, d), k)-property, then C is k-piercable? Very few results(mostly negative) are known. Inspired by the original question of Danzer and Gr¨unbaum we consider their question in an abstract set theoretic setting. Let U be a set (possibly infinite). Let C be a family of subsets of U with the property that if C1, . . . ,Cp+1 ∈ C are p + 1 distinct subsets, then |C1 ∩ · · · ∩Cp+1| ≤ l. In the second part of this thesis, we show in this setting, the first general positive results for the Danzer Grunbaum problem. As an extension, we show polynomial sized kernels for hitting set and covering problems in our setting. In the third part of this thesis, we broadly look at hitting and covering questions with respect to points and families of geometric objects in Rd. Let P be a subset of points(possibly infinite) in Rd and C be a collection of subsets of P induced by objects of a given family. For the system (P, C), let νh be the packing number and νc the dual packing number. We consider the problem of bounding the transversal number τ h and the dual transversal number τ c in terms of νh and νc respectively. These problems has been well studied in the case when P = R2. We systematically look at the case when P is finite, showing bounds for intervals, halfspaces, orthants, unit squares, skylines, rectangles, halfspaces in R3 and pseudo disks. We show bounds for rectangles when P = R2. Given a point set P ⊆ Rd, a family of objects C and a real number 0 < ǫ < 1, the strong epsilon net problem is to find a minimum sized subset Q ⊆ P such that any object C ∈ C with the property that |P ∩C| ≥ ǫn is hit by Q. It is customary to express the bound on the size of the set Q in terms of ǫ. Let G be a uniform √n × √n grid. It is an intriguing question as to whether we get significantly better bounds for ǫ-nets if we restrict the underlying point set to be the grid G. In the last part of this thesis we consider the strong epsilon net problem for families of geometric objects like lines and generalized parallelograms, when the underlying point set is the grid G. We also introduce the problem of finding ǫ-nets for arithmetic progressions and give some preliminary bounds.

Cette thèse a pour objectif la génération de séquences de franchissement dans les Réseaux de Petri Temporisés (RdPT) en utilisant une approche incrémentale. Le verrou principal auquel est confronté ce travail est l’explosion combinatoire qui résulte de la construction classique du graphe d’accessibilité du RdPT. Nous proposons d’utiliser la notion de séquence de steps temporisés, afin d’exprimer progressivement l’ensemble des séquences de franchissements permettant de passer d’un état courant à un état cible. La notion de step temporisé correspond à une abstraction logique du comportement du système considéré. Le caractère incrémental de l’approche a pour objectif de gagner en efficacité. En effet, il consiste à exprimer tout nouvel état de la résolution par rapport à une profondeur K+1, en fonction d’un état atteint à la profondeur K. Ainsi, nous proposons plusieurs algorithmes de recherche incrémentale permettant d'améliorer l'efficacité de la résolution des problèmes d'accessibilité. Nous utilisons ensuite la programmation par contraintes pour modéliser le problème de recherche d’accessibilité dans un RdPT et mettre en œuvre notre approche incrémentale. Notre approche permet également d’ajouter des contraintes spécifiques à un contexte de résolution. Nous avons notamment utilisé cette possibilité pour proposer des techniques d'identification des jetons dans un RdPT borné, dans le cadre de la reconfiguration des systèmes manufacturiers. Nous concluons par l’évaluation de différentes applications constituant des « benchmarks » permettant d’illustrer l'efficacité des approches proposées / This PhD thesis is dedicated to the generation of firing sequences in Timed Petri Net (TPN) using an incremental approach. To reduce the influence of the well-known combinatorial explosion issue, a unique sequence of timed steps is introduced to represent implicitly the underlying reachability graph of the TPN, without needing its whole construction. This sequence of timed steps is developed based on the logical abstraction technique. The advantage of the incremental approach is that it can express any state just from the last step information, instead of representing all states before.Several incremental search algorithms are introduced to improve the efficiency of our methodology. Constraint programming techniques are used to model and solve our incremental model, in which search strategies are developed that can search for solutions more efficiently. Our methodology can be used to add specific constraints to model realistic systems. Token identification techniques are developed to handle token confusion issues that appear when addressing the reconfiguration of manufacturing systems. Experimental benchmarks illustrate the effectiveness of approaches proposed in this thesis

Deep neural networks with millions of parameters are at the heart of many state of the art computer vision models. However, recent works have shown that models with much smaller number of parameters can often perform just as well. A smaller model has the advantage of being faster to evaluate and easier to store - both of which are crucial for real-time and embedded applications. While prior work on compressing neural networks have looked at methods based on sparsity, quantization and factorization of neural network layers, we look at the alternate approach of pruning neurons. Training Neural Networks is often described as a kind of `black magic', as successful training requires setting the right hyper-parameter values (such as the number of neurons in a layer, depth of the network, etc ). It is often not clear what these values should be, and these decisions often end up being either ad-hoc or driven through extensive experimentation. It would be desirable to automatically set some of these hyper-parameters for the user so as to minimize trial-and-error. Combining this objective with our earlier preference for smaller models, we ask the following question - for a given task, is it possible to come up with small neural network architectures automatically? In this thesis, we propose methods to achieve the same. The work is divided into four parts. First, given a neural network, we look at the problem of identifying important and unimportant neurons. We look at this problem in a data-free setting, i.e; assuming that the data the neural network was trained on, is not available. We propose two rules for identifying wasteful neurons and show that these suffice in such a data-free setting. By removing neurons based on these rules, we are able to reduce model size without significantly affecting accuracy. Second, we propose an automated learning procedure to remove neurons during the process of training. We call this procedure ‘Architecture-Learning’, as this automatically discovers the optimal width and depth of neural networks. We empirically show that this procedure is preferable to trial-and-error based Bayesian Optimization procedures for selecting neural network architectures. Third, we connect ‘Architecture-Learning’ to a popular regularize called ‘Dropout’, and propose a novel regularized which we call ‘Generalized Dropout’. From a Bayesian viewpoint, this method corresponds to a hierarchical extension of the Dropout algorithm. Empirically, we observe that Generalized Dropout corresponds to a more flexible version of Dropout, and works in scenarios where Dropout fails. Finally, we apply our procedure for removing neurons to the problem of removing weights in a neural network, and achieve state-of-the-art results in scarifying neural networks.

Trends towards ‘informalization’ are looming large in the world today. African cities have long been characterised by the presence of an ‘informal sector’ but are now experiencing new waves of ‘informalization’. Policies of liberalisation and structural adjustment are both changing the conditions under which urban dwellers make a living and encouraging states to abdicate from responsibilities for popular welfare. In this context, urbanites increasingly rely on informal ways of income earning and of social security provisioning. This book is about processes of ‘informalization’ in the West African city of Bissau in Guinea-Bissau. It begins with a historical account of the way conditions of informality have evolved through the encounter of locally specific forms of informal relations with colonialism and the socialist era. This is followed by an analysis of how disadvantaged groups who rely on informal ways of provisioning are faring in the context of contemporary changes. The study looks at both the informal income-generating activities and the social networks that urbanites engage in to sustain their income activities and their consumption. It seeks to assess whether these groups are coping with these wider changes or are becoming marginalised from networks of assistance and from activities that provide sufficient incomes. The social relations pervading access to support and livelihood resources as well as the informal rules governing such access are in focus. Forms of regulation in the informal sphere are also discussed. / <p>This thesis won the prize of “Best doctoral thesis in the Social Sciences at Stockholm University in 2001-2002”. Författaren är numera verksam vid Nordiska Afrikainstitutet</p>