DCC - Departamento de Ciência da Computação

Peer (P2P) applications are emerging technologies based on the same paradigm: the Peer-to-Peer paradigm. Motivated, respectively, by the necessity of executing applications in environments with no previous infra-structure and the demand for applications that share, ...

Mobile ad hoc networks (MANETs) and peer-to-peer (P2P) applications are emerging technologies based on the same paradigm: the P2P paradigm. Motivated, respectively, by the necessity of executing applications in environments with no previous infra-structure and the demand for applications that share files or distribute processing through the Internet, MANETs and P2P applications have received some interest from research community. As a characteristic of the distributed model, which they follow, such technologies face a difficult task of routing requests in a decentralized environment. In this paper, we conducted a detailed study of a Gnutella-like application running over a MANET where three different protocols were considered. The results show that each protocol that were analyzed performed well in under some conditions and for some metrics, while had drawbacks in others.

Both Mobile Ad Hoc Networks (MANETs) and Peer-to-Peer (P2P) Overlay Networks are decentralized and self-organizing networks with dynamic topology which route queries in a distributed environment. Nevertheless, MANETs are composed of resource-constrained devices susceptible to faults, whereas P2P networks are fault- tolerant. Thus, these systems are not only similar, but also complementary, and P2P networks are the ideal data sharing

Both Mobile Ad Hoc Networks (MANETs) and Peer-to-Peer (P2P) Overlay Networks are decentralized and self-organizing networks with dynamic topology and responsible for routing queries in a distributed environment. Nevertheless, MANETs are composed of resource-constrained devices susceptible to faults, whereas P2P networks are fault-tolerant. Thus, we argue that these systems are not only similar, but also complementary, and that P2P networks

Content dissemination over mobile ad hoc networks (MANETs) is usually performed using peer-to-peer (P2P) networks due to its increased resiliency and efficiency when compared to client-server approaches. P2P networks are usually divided into two types, structured and unstructured, based on their content discovery strategy. Unstructured networks use controlled flooding, while structured networks use distributed indexes. This article evaluates the performance of these two approaches over MANETs and proposes modifications to improve their performance. Results show that unstructured protocols are extremely resilient, however they are not scalable and present high energy consumption and delay. Structured protocols are more energy-efficient, however they have a poor performance in dynamic environments due to the frequent loss of query messages. Based on those observations, we employ selective forwarding to decrease the bandwidth consumption in unstructured networks, and introduce redundant query messages in structured P2P networks to increase their success ratio.

Due to the exploratory nature of DNNs, DL specialists often need to modify the input dataset, change a filter when preprocessing input data, or fine-tune the models’ hyperparameters, while analyzing the evolution of the training. However, the specialist may lose track of what hyperparameter configurations have been used and tuned if these data are not properly registered. Thus, these configurations must be tracked and made available for the user’s analysis. One way of doing this is to use provenance data derivation traces to help the hyperparameter’s fine-tuning by providing a global data picture with clear dependencies. Current provenance solutions present provenance data disconnected from W3C PROV recommendation, which is difficult to reproduce and compare to other provenance data. To help with these challenges, we present Keras-Prov, an extension to the Keras deep learning library to collect provenance data compliant with PROV. To show the flexibility of Keras-Prov, we extend a p...

In this paper, we propose a new approach for color image simplification in order to improve flame pixel classification. The fire detection performance depends critically on the performance of the flame pixel classifier. Color image simplification is the process of simplifying an image in order to decrease the number of colors while preserving, as much as possible, shapes. In this work, a hierarchical clustering method in a given color space is used to map the original colors into a smaller set of representative ones, allowing the use of a simple heuristic rule for classifying the clusters related to candidate flame colors. Using reverse mapping, we identify possible flame colors in the image. Main contributions of our work are the application of a simple hierarchical clustering method to color simplification, that decreases the number of possible flame colors, and a filtering methodology to reduce the influence of outliers. Several color spaces and distance measures were used to evaluate the proposed method. Experimental results demonstrate that color simplification is essential to successfully employ heuristic classification of flame colors.

This paper addresses transition detection which consists in identifying the boundary between consecutive shots. In this work, we propose an approach to cope with transition detection in which we define and use a new dissimilarity measure based on the size of the maximum cardinality matching calculated using a bipartite graph with respect to a sliding window. The experiments have used two video datasets which presents a variety of different video genres with 3079 transitions. Our method achieves performance measures similar to the best results found in the literature with a much simpler classification approach.

Hierarchical video segmentation provides region-oriented scale-space, i.e., a set of video segmentations at different detail levels in which the segmentations at finer levels are nested with respect to those at coarser levels. In this work, the hierarchical video segmentation is transformed into a graph partitioning problem in which each part corresponds to one supervoxel of the video, and we present a new methodology for hierarchical video segmentation which computes a hierarchy of partitions by a reweighting of the original graph using a simple dissimilarity measure in which a not too coarse segmentation can be easily inferred. We also provide an extensive comparative analysis, considering quantitative assessments showing accuracy, ease of use, and temporal coherence of our methods -p-HOScale, cp-HOScale and 2cp-HOScale. According to the experiments, the hierarchy inferred by our methods produces good quantitative results when applied to video segmentation. Moreover, unlike to other tested methods, space and time cost of our methods are not influenced by the number of supervoxels to be computed.

Currently there is a great advance in technology, especially in the social networks area, which allows us to access information recently released in a fascinating speed and comprehensive audience. Inserted in the business, the data mining area are gradually being used in the media, aiming to find groups of people who have an ideal standard of shopping for a particular company. Here in the following article a case study related to data mining over the social network Twitter. Resumo. Atualmente há um grande avanço da tecnologia, principalmente na área de redes sociais, que nos permite acessar informações recém-divulgadas em uma velocidade fascinante e um público-alvo abrangente. Inserido no âmbito empresarial, a área de mineração de dados vêm, aos poucos, sendo utilizada nos meios de comunicação social, com finalidade de encontrar grupos de pessoas que possuam um padrão ideal de compras para uma determinada empresa. Apresentamos no artigo a seguir um estudo de caso relacionado à mineração de dados em cima da rede social Twitter. 1. Introdução Um dos fatores mais importantes para o sucesso de uma empresa é conhecer seu público alvo. Isso porque ao se conhecer o cliente é possível direcionar campanhas de marketing específicas e adaptar os serviços prestados à vontade do mesmo. No entanto, avaliar a vontade e opinião dos clientes em relação aos produtos e serviços oferecidos pela empresa não é tarefa simples. Uma das formas utilizadas pelas empresas para coletar tais informações é a distribuição de questionários e realização de pesquisas de opinião, que visam possibilitar uma análise da aceitação do produto oferecido bem como possíveis questões a serem mudadas. Contudo o uso de questionários e pesquisas é, muitas vezes, ineficiente. Isso porque esta forma de coleta de dados possui um alcance relativamente reduzido, que varia de 1% em questionários de filantropia a 95% em questionários de censo, como citado por Courage e Baxter (2005), o que acaba prejudicando a precisão dos dados coletados. Outro motivo é que tais técnicas possuem um custo consideravelmente alto de aplicação. Esse custo ainda é maior quando tais questionários e pesquisas não são automatizados, isto é, são necessárias pessoas para suas aplicações e contabilizações. Tendo este cenário em mente surge a necessidade de uma forma alternativa de coleta de dados, que não necessite da procura do cliente e nem de seu tempo dedicado. Uma das alternativas que vem sendo estudadas atualmente é o uso dos dados coletados

RESUMO As instituições financeiras ao captarem recursos de poupadores e repassá-los aos agentes deficitários, mediante contratos de crédito, estão sujeitas ao risco de crédito, que se refere à incapacidade dos tomadores arcarem com suas dívidas. As cooperativas de crédito podem ter esse risco ampliado em virtude de um fator subjetivo envolvido nas concessões de crédito: as relações de proximidade existentes entre gestores da cooperativa e cooperados, o que diminui o rigor na avaliação do tomador de crédito. Nesse sentido, esse estudo objetivou avaliar os condicionantes do risco de crédito de uma cooperativa de crédito da região do Alto Paranaíba/Minas Gerais, utilizando dados de 2001 a 2010. Foi utilizado o modelo de regressão logit, sendo os contratos de crédito classificados como adimplentes e inadimplentes. Os resultados encontrados indicaram que quanto maior for a idade e os meses de fidelidade menor será a probabilidade de inadimplência na cooperativa e que a linha de crédito que amplia o risco está associada aos empréstimos. ABSTRACT During the process of obtaining resources from savers, and transferring them to deficit agents, by means of credit agreements, the financing institutions are subject to credit risk, which refers to the inability of credit borrowers in paying their debts. The credit cooperatives may have that risk increased due to a subjective factor involved in the credit extension, namely: the existing close-related relationships between cooperative managers and members, what reduces the accuracy in assessing the borrowers' credit. The purpose of this study was to assess determining factors of credit risk in a cooperative of Alto Paranaíba, State of Minas Gerais, Brazil, using data from the year of 2001 to 2010. The logic regression model was used, and credit agreements were classified as default and defaulters. According to results, as greater is the age and fidelity months, lesser will be the defaulters probability in the cooperative. Therefore, the credit line which increases the risk is associated with borrowings. Palavras-chave: Cooperativas de crédito, risco de crédito, Alto Paranaíba/Minas Gerais.

This paper describes the results and experiences of a teaching project using concepts of Computer Science Unplugged methods on a undergraduate database class. Four unplugged activities were developed and two database class were used in a quasi-experimental study in which we analyse differences in learning and satisfaction of the students between the two classes. One class used the unplugged activities and the other used traditional teaching methods. Results indicates that the learning process was not significantly affected but the satisfaction level of the students was higher on the class exposed to the unplugged activities. Resumo. Este trabalho descreve as experiências e resultados de um projeto de ensino em andamento desenvolvido para a disciplina de banco de dados, em nivel de graduação, aplicando atividades em sala de aula segundo os métodos da computação desplugada. Foram desenvolvidas quatro atividades e utiliza-das duas turmas em um estudo quasi-experimental para avaliar diferenças nos níveis de aprendizagem e satisfação com o método. A experiência ocorreu du-rante o segundo semestre letivo de 2013. Os resultados indicam que o método não teve impacto significativo na aprendizagem dos conceitos, mas teve dife-rença positiva e significativa em relação à satisfação dos alunos.

RESUMO Apesar da necessidade de gerenciamento de modelos nas organizações, um número reduzido de tomadores de decisões utiliza alguma ferramenta para gerenciamento de modelos. Algumas razões para este paradoxo são a complexidade inerente aos modelos e a falta de um ambiente amigável para os tomadores de decisões. Neste trabalho, é proposto um ambiente gráfico onde os tomadores de decisão possam formular, gerenciar e executar modelos de forma simples e objetiva. O ambiente possui suporte à SML (Structured Modeling Language), como linguagem de definição de modelos, e à SQL-M, uma extensão da SQL (Structured Query Language) para gerenciamento de modelos. As duas linguagens são utilizadas de forma visual. Os modelos são criados, visualmente, através de grafos de modelos e as instâncias de modelos são criadas através de um assistente que auxilia o tomador de decisão em todo o processo. A execução do modelo é feita por resolvedores externos ao ambiente, mas tudo de forma transparente para o tomador de decisão. Todas estas características formam um ambiente, onde o tomador de decisão pode gerenciar os modelos necessários à organização de forma simples, sem a necessidade de conhecer profundamente uma linguagem de definição ou de manipulação de modelos. ABSTRACT Despite the need for model use and management in organizations, a small number of decision makers use some tool for that purpose. Some reasons for this paradox are model complexity and the lack of a friendly environment for model construction, testing and management tailored for decision makers. In this work, we present GMMS (Graphical Model Management System), a graphical environment in which decision makers can formulate, manage and execute models in a simple and straight fashion. The environment provides support for SML (Structured Modeling Language), a model definition language, and for SQL-M, that is an SQL (Structured Query Language) extension for model management. Both languages are used through an easy to use graphical language. Models are created visually by dragging and dropping icons from a modeling palette, and model instances are created using a wizard that helps the decision maker along the process. Models are executed by using solvers external to the environment, in a way that is fully transparent to the decision maker. GMMS provides a friendly and easy to use environment where decision makers can play with models without having a deep technical knowledge on model management. This leaves spare time to get more concentrated on the problem and its domain. KEYWORDS Model Management, Decision Support Systems, Structured Modeling 1 Trabalho desenvolvido a partir de dissertação de Mestrado em Ciência da Computação através do convênio DCC-UFMG/DPI-UFV com financiamento parcial da FAPEMIG.