Ronaldo R. Goldschmidt

Automatic feature selection for supervised learning in link prediction applications: a comparative study

For the last years, a considerable amount of attention has been devoted to the research about the link prediction (LP) problem in complex networks. This problem tries to predict the likelihood of an association between two not interconnected nodes in a network to appear in the future. One of the most important approaches to the LP problem is based on supervised machine learning (ML) techniques for classification. Although many works have presented promising results with this approach, choosing the set of features (variables) to train the classifiers is still a major challenge. In this article, we report on the effects of three different automatic variable selection strategies (Forward, Backward and Evolutionary) applied to the feature-based supervised learning approach in LP applications. The results of the experiments show that the use of these strategies does lead to better classification models than classifiers built with the complete set of variables. Such experiments were performed over three datasets (Microsoft Academic Network, Amazon and Flickr) that contained more than twenty different features each, including topological and domain-specific ones. We also describe the specification and implementation of the process used to support the experiments. It combines the use of the feature selection strategies, six different classification algorithms (SVM, K-NN, naïve Bayes, CART, random forest and multilayer perceptron) and three evaluation metrics (Precision, F-Measure and Area Under the Curve). Moreover, this process includes a novel ML voting committee inspired approach that suggests sets of features to represent data in LP applications. It mines the log of the experiments in order to identify sets of features frequently selected to produce classification models with high performance. The experiments showed interesting correlations between frequently selected features and datasets.

Using a Time based Relationship Weighting Criterion to Improve Link Prediction in Social Networks.

For the last years, a considerable amount of attention has been devoted to the research about the link prediction (LP) problem in complex networks. This problem tries to predict the likelihood of an association between two not interconnected nodes in a network to appear in the future. Various methods have been developed to solve this problem. Some of them compute a compatibility degree (link strength) between connected nodes and apply similarity metrics between non-connected nodes in order to identify potential links. However, despite the acknowledged importance of temporal data for the LP problem, few initiatives investigated the use of this kind of information to represent link strength. In this paper, we propose a weighting criterion that combines the frequency of interactions and temporal information about them in order to define the link strength between pairs of connected nodes. The results of our experiment with traditional weighted similarity metrics in ten co-authorship networks confirm our hypothesis that weighting links based on temporal information may, in fact, improve link prediction. Proposed criterion formulation, experimental procedure and results from the performed experiment are discussed in detail.

Dimensionality Reduction for Supervised Learning in Link Prediction Problems

In recent years, a considerable amount of attention has been devoted to research on complex networks and their properties. Collaborative environments, social networks and recommender systems are popular examples of complex networks that emerged recently and are object of interest in academy and industry. Many studies model complex networks as graphs and tackle the link prediction problem, one major open question in network evolution. It consists in predicting the likelihood of an association between two not interconnected nodes in a graph to appear. One of the approaches to such problem is based on binary classification supervised learning. Although the curse of dimensionality is a historical obstacle in machine learning, little effort has been applied to deal with it in the link prediction scenario. So, this paper evaluates the effects of dimensionality reduction as a preprocessing stage to the binary classifier construction in link prediction applications. Two dimensionality reduction strategies are experimented: Principal Component Analysis (PCA) and Forward Feature Selection (FFS). The results of experiments with three different datasets and four traditional machine learning algorithms show that dimensionality reduction with PCA and FFS can improve model precision in this kind of problem.

MEMORE - Um Ambiente Computacional para Apoio ao Acompanhamento do Programa Um Computador por Aluno

The keynotes goal is to reflect on emergent opportunities for human Discovery (in science), Creativity (in art & industry), and Learning (in education) as processes often occurring serendipitously in individuals and in communities empowered by dynamic Web connections in the global village. These reflections seem to fit best with the mandate of the CBIE Conference: sustainable education.Este trabalho apresenta uma evolucao do assistente inteligente SAE para inferir e fornecer orientacao pedagogica coerente com varias metodologias educacionais, superando uma dificuldade tradicional para este tipo de software, alem de fornecer mais apoio ao ensino-aprendizagem personalizado.O artigo fundamenta-se nos resultados do processo de desenvolvimento de um ambiente web de aprendizagem, ja modelado em dissertacao de mestrado (CABRAL,2006). O sistema tem por objetivo facilitar o acompanhamento das atividades que envolvem a definicao e a elaboracao de pre-projetos de Trabalhos de Conclusao de Curso (TCC). Para atingir tal objetivo o sistema e baseado em uma camada de agentes inteligentes que atuam como colaboradores para execucao de todas as atividades necessarias ao desenvolvimento do TCC. Para que esses agentes atuem de forma satisfatoria no ambiente e necessario o estabelecimento de um processo de comunicacao que possibilite aos agentes agirem em um nivel conceitual mais proximo do ser humano. Dessa forma, este artigo expoe os resultados obtidos ao longo do desenvolvimento dos agentes computacioniais e seus mecanismos de comunicacao.Este artigo descreve o projeto de desenvolvimento de jogo educativo baseado em metodologias participativas, pelas quais estudantes da rede publica de ensino medio idealizam, projetam e desenvolvem prototipos de um jogo sobre Sistema Imunologico e virus da Dengue. O objetivo da pesquisa e verificar os impactos do processo sobre a aprendizagem dos estudantes.O objetivo deste artigo e apresentar resultados de estudo de um caso exploratorio, realizado junto aos alunos de uma Instituicao de Ensino Superior. Como resultados, foram levantados indicadores de avaliacao do Sistema de EaD e identificados problemas, segundo a percepcao dos alunos. A partir dos resultados destaca-se que a Universidade desconhece a metodologia ITIL® e que seria fundamental que fosse implantada uma estrategia do servico (fase inicial da ITIL®), conduzindo a melhoria continua do servico e da infra-estrutura da Universidade.O facil acesso as informacoes devido a difusao da internet possibilita o enriquecimento intelectual, mas por vezes acaba por ser um meio de usufruir do conhecimento de outrem sem mencionar seus creditos/direitos autorais, o que por fim acaba na configuracao do plagio. O plagio no meio academico e uma tarefa dificil de ser controlada, devido o grande numero de trabalhos que sao feitos por uma vasta quantidade de alunos e tambem pelo excesso de tarefas dos professores e pelo pouco tempo que conseguem dedicar para o controle da qualidade e autenticidade dos trabalhos. Com isso, e importante contar com softwares que auxiliem no processo de verificacao de indicios de plagio, desta maneira o presente trabalho vislumbra desenvolver uma nova ferramenta de analise de indicios de plagio bem como aprimorar o metodo DIP – Detector de Indicios de Plagio para auxiliar o docente na verificacao da autenticidade dos trabalhos.A Neuropedagogia promove o confronto sucessivo e simultâneo entre concepcoes tacitas, crencas e valores normativos com esquemas inovadores, criando um espaco entropico e poetico no âmbito da experiencia humana. A natureza da ciencia neuropedagogica exige um modelo fractal de construcao de mundo, sustentado pelo prazer da perplexidade. Busca-se a geratriz e as leis de formacao para atingir a unanimidade sobre o conceito de educacao de pessoas com alta dotacao humana, impossivel de se concretizar, sem o recurso das ideologias filosoficas, cientificas e tecnologias aliadas as eticas.In this talk I will elaborate on the development, implementation and evaluation of the use of online tools for learning, with special attention to games and gamification, sharing and collaboration.

Relato de uma Experiência de Integração Universidade-Governo-Escola na Construção Coletiva de um Sistema de Monitoramento do Programa Um Computador por Aluno

A presente pesquisa tem como objetivo a avaliacao das consequencias da colaboracao entre alunos leigos em musica atraves de atividades educacionais envolvendo a composicao musical com suporte computacional. Foi realizado um estudo de caso que teve uma parte exploratoria e outra explanatoria. A exploratoria teve o objetivo de comparar o discurso entre os alunos de dois grupos na atividade de composicao musical colaborativa realizada. Ja a parte explanatoria teve o objetivo de entender a percepcao dos participantes sobre sua producao individual comparada a sua producao colaborativa. Um dos resultados parece indicar que atividades de aprendizagem de musica por composicao colaborativa sao melhores vistas pelos aprendizes, quando comparadas as mesmas atividades executadas de maneira individual. Alem disso, foi visto que os alunos parecem perceber o “discurso musical” de seus colegas e refletem sobre ele, adaptando suas contribuicoes individuais.Este artigo aborda a utilizacao das Tecnologias Educacionais em Rede, especialmente o recurso Wiki, na disciplina de Supervisao de Estagio em um Programa Especial de Graduacao de Formacao de Professores para a Educacao Profissional e Tecnologica. Analisou-se os registros feitos por vinte e nove estudantes e o respectivo orientador, referentes aos planos de aula e sua implementacao. As postagens geraram reflexoes no espaco Wiki, proporcionando categorias de elementos indicadores para a supervisao dos futuros estagios, na forma de rede de conversacao. A possibilidade de acompanhamentos, orientacoes, avaliacoes e aprendizagens tornou-se efetivamente um periodo de construcao de experiencias, alem do espaco academico.Este artigo descreve o modelo de trabalho adotado pelo Projeto Lotus para a transformacao de maquinas caca-niqueis em computadores com o sistema operacional Linux e a suite educacional GCompris como um esforco para melhorar a inclusao digital nas escolas municipais do ensino fundamental em Balneario Camboriu, SC. Os resultados obtidos entre 2009 e 2012 foram a instalacao de 12 laboratorios nos nucleos de educacao infantil, 122 computadores e um total de 2.433 estudantes favorecidos.O presente artigo tem como objetivo refletir sobre a insercao do Projeto Um Computador por Aluno na realidade escolar, apresentando experiencias de praticas pedagogicas utilizando-se do software Squeak Etoys, disponivel no laptop educacional. O trabalho apresenta um breve recorte historico a respeito do projeto citado, sobre sua insercao nas escolas publicas e da dinâmica da formacao continuada entre professores de algumas escolas de Sergipe contempladas pelo projeto. A pesquisa baseou-se em um estudo de caso, apresentando depoimentos obtidos a partir da formacao continuada do Projeto UCA, evidenciando praticas exitosas a partir da insercao do Programa no estado, que podem ser difundidas e aprimoradas pelos docentes.A utilizacao das novas tecnologias no contexto educacional inspira educadores no desenvolvimento de estrategias inovadoras e motivadoras nas diversas areas do conhecimento. No entanto e imprescindivel observar como os alunos interagem com essas tecnologias. A partir de observacoes do cotidiano escolar, promovemos o projeto Fa de Fanzine com o objetivo de proporcionar aos estudantes a oportunidade de produzir um fanzine, atraves do qual pudessem vivenciar a inedita e fascinante experiencia de integrar arte, leitura, producao de imagens e textos, em diversos generos, relacionados a um objeto de seu interesse, utilizando a tecnologia do computador, visando socializacao dos conteudos atraves de suporte impresso e virtual.Este artigo tem como finalidade explanar a implantacao por iniciativa e recursos proprios do Projeto UCA (Um Computador por Aluno) em um municipio da Regiao Metropolitana de Curitiba, capital do Estado do Parana, visando salientar a importância do Projeto como um todo, desde a estruturacao da escola para receber os Laptops Educacionais, a conscientizacao da comunidade escolar a respeito da utilizacao deste recurso e tambem a formacao continuada de todos os profissionais envolvidos no Projeto.

A Crowdsourcing Tool for Data Augmentation in Visual Question Answering Tasks

Visual Question Answering (VQA) is a task that connects the fields of Computer Vision and Natural Language Processing. Taking as input an image I and a natural language question Q about I, a VQA model must be able to produce a coherent answer R (also in natural language) to Q. A particular type of visual question is one in which the question is binary (i.e., a question whose answer belongs to the set {yes, no}). Currently, deep neural networks correspond to the state of the art technique for training of VQA models. Despite its success, the application of neural networks to the VQA task requires a very large amount of data in order to produce models with adequate precision. Datasets currently used for the training of VQA models are the result of laborious manual labeling processes (i.e., made by humans). This context makes relevant the study of approaches to augment these datasets in order to train more accurate prediction models. This paper describes a crowdsourcing tool which can be used in a collaborative manner to augment an existing VQA dataset for binary questions. Our tool actively integrates candidate items from an external data source in order to optimize the selection of queries to be presented to curators.

Context-based Time Score: An Effective Similarity Function for Link Prediction in Social Networks

Online social networks (OSNs) have become an extremely relevant way for modeling social interactions among people in a group or community. A comprehensive set of studies has analyzed how to predict which new interactions will occur between the participants from OSNs. The problem of predicting new interactions is formally stated as link prediction. Most studies related to link prediction is based on similarity functions that use data from different types such as topological (data about the network structure), temporal (chronological interaction data) or contextual (participant and link attributes), usually available on OSNs. However, none of those studies uses the different types of data simultaneously, leading to poor incorporation of the different aspects of the OSNs in link prediction. To address this issue, the present paper introduces a new similarity function called Context-based Time Score (CTS), which combines topological, temporal and contextual data to improve accuracy in predicting the occurrence of new connections. Experiments with ten different datasets revealed that CTS can outperform similarity functions that do not take the three types of data simultaneously.

Combining contextual, temporal and topological information for unsupervised link prediction in social networks

Abstract Understanding and characterizing the processes driving social interactions is one of the fundamental problems in social network analysis. In this context, link prediction aims to foretell whether two not linked nodes in a network will connect in the near future. Several studies proposed to solve link prediction compute compatibility degrees as link weights between connected nodes and, based on a weighted graph, apply weighted similarity functions to non-connected nodes in order to identify potential new links. The weighting criteria used by those studies were based exclusively on information about the existing topology (network structure). Nevertheless, such approach leads to poor incorporation of other aspects of the social networks, such as context (node and link attributes), and temporal information (chronological interaction data). Hence, in this paper, we propose three weighting criteria that combine contextual, temporal and topological information in order to improve results in link prediction. We evaluated the proposed weighting criteria with two popular weighted similarity functions (Adamic-Adar and Common Neighbors) in ten networks frequently used in experiments with link prediction. Results with the proposed criteria were statistically better than the ones obtained from the weighting criterion that is exclusively based on topological information.

Using a Time-Based Weighting Criterion to Enhance Link Prediction in Social Networks

Recently, the link prediction (LP) problem has attracted much attention from both scientific and industrial communities. This problem tries to predict whether two not linked nodes in a network will connect in the future. Several studies have been proposed to solve it. Some of them compute a compatibility degree (link strength) between connected nodes and apply similarity metrics between non-connected nodes in order to identify potential links. However, despite the acknowledged importance of temporal data for the LP problem, few initiatives investigated the use of this kind of information to represent link strength. In this paper, we propose a weighting criterion that combines the frequency of interactions and temporal information about them in order to define the link strength between pairs of connected nodes. The results of our experiment with weighted and non-weighted similarity metrics in ten co-authorship networks present statistical evidences that confirm our hypothesis that weighting links based on temporal information may, in fact, improve link prediction.

Distributed Data Clustering in the Context of the Internet of Things: A Data Traffic Reduction Approach

The Internet of Things (IoT) emerged with the objective to integrate physical objects into classical computer networks. These objects usually generate larges amount of data, transferring the bottleneck of data processing from sensors to communication systems. For example, analyzing IoT data often demands data centralization before running a mining algorithm. Thus, in order to reduce the data transference commonly required by the data clustering task, this paper proposes a grid-based data summarization approach. The proposed approach uses a single uniform grid to partition the space into cells and to summarize data before centralization. Summarization ensures the reduction of the amounts of data transferred. This approach also includes a data clustering algorithm that deals with the summarized and centralized data. Our preliminary experiments revealed good results in terms of data compression and quality of clustering with a two-dimensional benchmark dataset.