Search

The contemporary education has a lot of challenges and among them is the adaptation of using new technologies with classical education paradigm. It hasn\'t been different with distance education. In this context, this work proposes to develop a hybrid tutoring system model with decisions based on the teachers knowledge and help from Self Organizing Maps (SOM) or Kohonen\'s Maps neural networks. The proposed system has a initial teaching method that is set by specialist teacher up and while system is being running this pedagogic method is refined by the neural networks, which use patterns extracted from students that has used the system. The model proposes the utilization of a basic neural network structure with automated training which is capable of train several networks and define the one which represents results that is more coherent with the pattern\'s set, dismissing the intervention of a specialist on the evaluation of the network training performance. The system has adaptive and reactive features related to the apprentice, being able to offer to the students a personalized and dynamic learning. The system was developed in a web environment aiming avail the advantages of this technology. At this work, besides the proposed model developing it also were performed a data gathering with fresh students from integrated learning technical of Federal Institution of Goiás, Luziânia, Goiás, Brazil, to evaluate system\'s applicability. This thesis presents the fundamentals theorists of the virtual education environment, as also the artificial neural networks SOM, used on proposed model. Likewise, it shows the system developing process, the automated training build, in addition with the system tutor structure. The knowledge\'s transmission is inspired in the content\'s didactic transposition, with organization didactic units in levels that aim develop distinct skills. The SOM networks analysis indicate that the automated train was able to train several networks and identify a network with best topologic order. Moreover, this work presents a comparison between students performance when submitted to learn using the system with purely specialized orientation and hybrid orientation. The outcomes of this evaluation points out the viability of the proposed model, since the system has shown to be able to learn from students and adapt the teacher learning method. The apprentices that studied utilizing the system had amplified theirs grades on the learning system evaluations and the hybrid tutor was capable of take decisions which magnify the acceptation of the tutor learning indication. A educação contemporânea tem muitos desafios, dentre eles o de adequar a utilização de novas tecnologias aos paradigmas clássicos de ensino. Com a educação a distância não tem sido diferente. Nesse contexto, este trabalho apresenta o processo de desenvolvimento de um sistema tutor inteligente híbrido com decisões baseadas no conhecimento do professor, subsidiadas pelas decisões das redes neurais Self Organizing Maps (SOM) ou Mapas de Kohonen. O sistema proposto tem a sua estratégia de ensino inicial estabelecida pelo professor especialista e à medida em que o sistema é utilizado a estratégia pedagógica é refinada pelas redes neurais, que utilizam padrões extraídos dos estudantes que usaram o tutor. O modelo propõe a utilização de uma estrutura de rede neural com treinamento automatizado que é capaz de treinar diferentes redes e definir a que apresente o resultado mais coerente com conjunto de padrões, dispensando a intervenção de um especialista na avaliação do desempenho das redes. O sistema tem característica adaptativa e reativa ao aprendiz, capaz de oferecer ao estudante um ensino personalizado e dinâmico. O sistema foi desenvolvido para ambiente web com o objetivo de obter vantagens que essa tecnologia oferece. Neste trabalho, além do desenvolvimento do modelo proposto também foram realizadas coletas de dados com alunos dos anos iniciais do ensino técnico integrado do Instituto Federal de Goiás Campus Luziânia para avaliar a aplicabilidade do sistema. A tese apresenta os fundamentos teóricos de ambientes virtuais de educação, bem como das redes neurais artificiais SOM, utilizadas no modelo proposto. Também é apresentado o processo de desenvolvimento do sistema, a construção do treinamento automatizado e a estruturação do sistema tutor. A transmissão de conhecimentos é inspirada na transposição didática de conteúdos, com unidades didáticas organizadas em níveis que tem como objetivo desenvolver diferentes competências. As análises das redes SOM indicaram que o treinamento automatizado foi capaz de treinar diversas redes e identificar a rede com melhor ordenação topológica. Também são apresentadas comparações entre os desempenhos dos estudantes quando submetidos a estudos nos sistemas com orientação puramente especialista e híbrido. Os resultados dessa avaliação indicam a viabilidade do modelo proposto, pois o sistema mostrou-se capaz de aprender com os estudantes e ajustar as estratégias de ensino do professor. Os aprendizes que estudaram no sistema ampliaram suas médias nas avaliações de aprendizado do sistema e o tutor híbrido foi capaz de tomar decisões que ampliaram a aceitação das indicações de estudos do tutor. Doutor em Ciências

Orientadores: Marcio Luiz de Andrade Netto, Jose Alfredo Ferreira Costa Acompanha Anexo A: Midia com informações adicionais em CD-R Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação Resumo: Esta tese propõe novas metodologias de classificação não-supervisionada de imagens de sensoriamento remoto que particularmente exploram as características e propriedades do Mapa Auto-organizável de Kohonen (SOM - Self-Organizing Map). O ponto chave dos métodos de classificação propostos é realizar a análise de agrupamentos das imagens através do mapeamento produzido pelo SOM, ao invés de trabalhar diretamente com os padrões originais das cenas. Tal estratégia reduz significativamente a complexidade da análise dos dados, tornando possível a utilização de técnicas normalmente consideradas computacionalmente inviáveis para o processamento de imagens de sensoriamento remoto, como métodos de agrupamentos hierárquicos e índices de validação de agrupamentos. Diferentemente de outras abordagens, nas quais o SOM é utilizado como ferramenta de auxílio visual para a detecção de agrupamentos, nos métodos de classificação propostos, mecanismos para analisar de maneira automática o arranjo de neurônios de um SOM treinado são aplicados e aprimorados com o objetivo de encontrar as melhores partições para os conjuntos de dados das imagens. Baseando-se nas propriedades estatísticas do SOM, modificações nos cálculos de índices de validação agrupamentos são propostas com o objetivo de reduzir o custo computacional do processo de classificação das imagens. Técnicas de análise de textura em imagens são aplicadas para avaliar e filtrar amostras de treinamento e/ou protótipos do SOM que correspondem a regiões de transição entre classes de cobertura terrestre. Informações espaciais a respeito dos protótipos do SOM, além das informações de distância multiespectral, também são aplicadas em critérios de fusão de agrupamentos procurando facilitar a discriminação de classes de cobertura terrestre que apresentam alto grau de similaridade espectral. Resultados experimentais mostram que os métodos de classificação propostos apresentam vantagens significativas em relação às técnicas de classificação não-supervisionada mais freqüentemente utilizadas na área de sensoriamento remoto. Abstract: This thesis proposes new methods of unsupervised classification for remotely sensed images which particularly exploit the characteristics and properties of the Kohonen Self-Organizing Map (SOM). The key point is to execute the clustering process through a set of prototypes of SOM instead of analyzing directly the original patterns of the image. This strategy significantly reduces the complexity of data analysis, making it possible to use techniques that have not usually been considered computationally viable for processing remotely sensed images, such as hierarchical clustering methods and cluster validation indices. Unlike other approaches in which SOM is used as a visual tool for detection of clusters, the proposed classification methods automatically analyze the neurons grid of a trained SOM in order to find better partitions for data sets of images. Based on the statistical properties of the SOM, clustering validation indices calculated in a modified manner are proposed with the aim of reducing the computational cost of the classification process of images. Image texture analysis techniques are applied to evaluate and filter training samples and/or prototypes of the SOM that correspond to transition regions between land cover classes. Spatial information about the prototypes of the SOM, in addition to multiespectral distance information, are also incorporated in criteria for merging clusters with aim to facilitate the discrimination of land cover classes which have high spectral similarity. Experimental results show that the proposed classification methods present significant advantages when compared to unsupervised classification techniques frequently used in remote sensing. Doutorado Engenharia de Computação Doutor em Engenharia Elétrica

This paper reviews the first block of our unsupervised deep collaborative recommendation (UDCF) system and proposes a platform whose goal is to try to find the adequate parameters of the Kohonen maps, to create homogeneous clusters in profile data and results, the homogeneity is verified thanks to the very low variance rate of the results obtained by the cluster population and a second criterion which is the high prediction rate of collaborative recommendation. Although the revision concerns only the clustering block, and the use of a symmetrical autoencoder without searching for its optimization, the result obtained (82.33%) for the optimal configurations with high homogeneity of the Kohonen map is equivalent to the optimized result of the UDCF and even better than the classical recommendation methods.

La conducción autónoma es uno de los temas en los que ha habido más investigación en los últimos años ya que llegar a conseguir un vehículo totalmente autónomo sería un gran avance para cualquier sociedad. Uno de los problemas de este campo de investigación es que generar nuevos agentes de conducción autónoma es lento y tedioso, impidiendo un desarrollo rápido. En este trabajo se propone un nuevo enfoque que intenta paliar este problema y que consiste en la utilización de imágenes realistas sintetizadas con Redes Generativas Adversarias para entrenar agentes de conducción autónoma que sean potencialmente aplicables en entornos reales. Siguiendo este enfoque, no solo se podrían generar imágenes sintéticas en entornos variados realistas, sino que también se podrían obtener imágenes realistas etiquetadas. Esto es posible ya que, como las imágenes sintéticas se generan a partir de un simulador, se pueden obtener de este las etiquetas como si de un vehículo real se tratara, pero de una forma mucho más flexible y rápida. Para desarrollar el sistema, se toma como base el simulador SDSandbox y se aplican una serie de modificaciones entre las cuales destaca la generación de circuitos aleatorios con mapas de Kohonen. Para producir las imágenes realistas se hace una extensa experimentación visitando las arquitecturas que conforman el estado del arte y utilizando conjuntos de datos tanto creados de forma artesanal como mundialmente conocidos. El objetivo de esta experimentación es obtener un modelo multimodal capaz de generar carreteras naturales nocturnas, de tarde y de día. Para comprobar que las imágenes generadas son realistas se hace una encuesta que implica a individuos reales, llegando a la conclusión de que las imágenes son lo suficientemente reales como para ser utilizadas para el entrenamiento de agentes de conducción autónoma. Finalmente, se crea un agente de conducción autónoma que utiliza como única entrada las imágenes producidas por las Redes Generativas Adversarias, añadiendo las líneas de la carretera. Autonomous driving is one of the subjects on which there has been more research in recent years, as achieving a totally autonomous vehicle would be a great advance for any society. One of the problems in this field of research is that generating new autonomous driving agents is slow and tedious, preventing the rapid development of autonomous driving. This paper proposes a new approach that tries to alleviate this problem and that consists in the use of synthetic realistic images with adversarial generative networks to train autonomous driving agents that are potentially applicable in real environments. Following this approach, not only could synthetic images be generated in various realistic environments, but also labelled realistic images could be obtained. This would be possible because, as the synthetic images are generated from a simulator, the labels could be obtained from it as if it was a real vehicle. To develop the system, the simulator SDSandbox is taken as a basis and a series of modifications are applied among which the generation of random circuits with Kohonen maps stands out. In order to produce the realistic images, extensive experimentation is performed by visiting the architectures that make up the state of the art and using both handmade and world-famous data sets. The objective of this experimentation is to create a multimodal model capable of generating natural roads at night, in the evening and during the day. In order to verify that the images generated are realistic, a survey is made involving real individuals, reaching the conclusion that the images are real enough to be used for the training of autonomous driving agents. Finally, an autonomous driving agent is created that uses as the only input the images produced by the generative adversarial networks, including the road lines. Ingenieria Informática

Artificial neural networks represent an alternative to traditional multivariate techniques, such as principal component and discriminant analysis, which rely on hypotheses regarding the normal distribution of the data and homoscedasticity. They also may be a powerful tool for multivariate modeling of systems that do not present linear correlation between variables, as well as to visualize high-dimensional data in bi- or trivariate structures. One special kind of neural network of interest in archaeometric studies is the Self-Organizing Map (SOM). SOMs can be distinguished from other neural networks for preserving the topological features of the original multivariate space. In this study, the self-organizing maps were applied to concentration data of chemical elements measured in archaeological ceramics from Central Amazon using instrumental neutron activation analysis (INAA). The main objective was testing the chemical patterns previously identified using cluster and principal component analysis, forming groups of ceramics according the multivariate chemical composition. It was verified by statistical tests that the chemical elemental data was not normally distributed and did not present homogeneity of covariance matrices for different groups, as requested by principal component analysis and other multivariate techniques. The maps obtained were consistent with the patterns identified by cluster and principal component analysis, forming two chemical groups of pottery shards for each archaeological site tested. Finally, it was verified the potential of SOMs for testing if failures in underlying hypotheses of traditional multivariate techniques might be critically influencing the results and subsequent archaeological interpretation of archaeometric data.

Este artículo presenta una metodología basada en redes neuronales para realizar priorización de zonas de mercado visto desde un enfoque empresarial. En esta investigación se intenta dar solución a la incertidumbre que existe en la mayoría de las organizaciones en torno a la prioridad que tiene una zona de mercado; para ello se hace una búsqueda de los criterios más relevantes que las empresas tienen en cuenta para asignar prioridades a ciertos clientes. La problemática se sustenta por la ausencia de herramientas que permitan determinar la prioridad de una zona de mercado y la falta de una interrelación efectiva entre los departamentos de logística y mercadeo. Para ello se ocupan los mapas de Kohonen que son un tipo de red neuronal que facilitan el agrupamiento de clientes y permiten determinar cuáles de ellos son los que impactan con mayor frecuencia los criterios de priorización previamente establecidos. Finalmente, se presentan tres escenarios con fin de validar la propuesta formulada y ver que comportamiento tienen las redes neuronales en temas de priorización de zonas de mercado. Este artículo presenta una metodología basada en redes neuronales para realizar priorización de zonas de mercado visto desde un enfoque empresarial. En esta investigación se intenta dar solución a la incertidumbre que existe en la mayoría de las organizaciones en torno a la prioridad que tiene una zona de mercado; para ello se hace una búsqueda de los criterios más relevantes que las empresas tienen en cuenta para asignar prioridades a ciertos clientes. La problemática se sustenta por la ausencia de herramientas que permitan determinar la prioridad de una zona de mercado y la falta de una interrelación efectiva entre los departamentos de logística y mercadeo. Para ello se ocupan los mapas de Kohonen que son un tipo de red neuronal que facilitan el agrupamiento de clientes y permiten determinar cuáles de ellos son los que impactan con mayor frecuencia los criterios de priorización previamente establecidos. Finalmente, se presentan tres escenarios con fin de validar la propuesta formulada y ver que comportamiento tienen las redes neuronales en temas de priorización de zonas de mercado.

This letter contains an approach to improve the Kohonen phonetic typewriter [1] by enlarging the phonetic objects which are used to label the phonetic map: the transitions between phonemes are introduced as new labels of neurons.

Chapitre 5, à la suite de la conférence ACSEG 2001 à Rennes; In the present article, we attempt to devise a typology of forms of part-time employment by applying a widely used neuronal methodology called Kohonen maps. Starting out with data that we describe using category-specific variables, we show how it is possible to represent observations and the modalities of the variables that define them simultaneously, on a single map. This allows us to ascertain, and to try to describe, the main categories of part-time employment.

In the analysis of poverty and social exclusion, indicators of living conditions are some interesting non-monetary complements to the usual measurements in terms of current or annual income. Living conditions depend in fact on longer term factors than income, and provide further information on households' actual resources that allow to compare more accurately between living standards. But in counterpart, a difficulty comes from the qualitative nature of the information, and the large number of dimensions and items that may be taken into account; in other words, living conditions are difficult to "measure". A consequence is that very often, the information is either used only partly, or reduced into a global score of (bad) living conditions, that results from counting "negative" items, and the qualitative dimension is lost. In this paper, we propose to use the Kohonen algorithm first to describe how the elements of living conditions are combined, and secondly to classify households according to their living conditions. The main interest of a classification is to make appear not only quantitative differences in the "levels" of living conditions, but also qualitative differences within similar "levels"., Comment: Extended version of a paper presented in ACSEG 2000 conference in Paris (dec 2000)

The knowledge-based Toxtree expert system (SAR approach) was integrated with the statistically based counter propagation artificial neural network (CP ANN) model (QSAR approach) to contribute to a better mechanistic understanding of a carcinogenicity model for non-congeneric chemicals using Dragon descriptors and carcinogenic potency for rats as a response. The transparency of the CP ANN algorithm was demonstrated using intrinsic mapping technique specifically Kohonen maps. Chemical structures were represented by Dragon descriptors that express the structural and electronic features of molecules such as their shape and electronic surrounding related to reactivity of molecules. It was illustrated how the descriptors are correlated with particular structural alerts (SAs) for carcinogenicity with recognized mechanistic link to carcinogenic activity. Moreover, the Kohonen mapping technique enables one to examine the separation of carcinogens and non-carcinogens (for rats) within a family of chemicals with a particular SA for carcinogenicity. The mechanistic interpretation of models is important for the evaluation of safety of chemicals.

Article de synthèse sur les applications de l'algorithme de Kohonen pour la visualisation et l'analyse de données; This paper shows how to use the Kohonen algorithm to represent multidimensional data, by exploiting the self-organizing property. It is possible to get such maps as well for quantitative variables as for qualitative ones, or for a mixing of both. The contents of the paper come from various works by SAMOS-MATISSE members, in particular by E. de Bodt, B. Girard, P. Letrémy, S. Ibbou, P. Rousset. Most of the examples have been studied with the computation routines written by Patrick Letrémy, with the language IML-SAS, which are available on the WEB page http://samos.univ-paris1.fr.