<p>The aim of the project was to identify biologically relevant novel gene clusters by using combined genomic data instead of using only gene expression data in isolation. The clustering algorithm based on self-organizing maps (Kasturi et al., 2005) was extended and implemented in order to use gene location data together with the gene expression and the motif occurrence data for gene clustering. A distance function was defined to be used with gene location data. The algorithm was also extended in order to use vector angle distance for gene expression data. <em>Arabidopsis thaliana</em> is chosen as a data source to evaluate the developed algorithm. A test data set was created by using 100 Arabidopsis genes that have gene expression data with seven different time points during cold stress condition, motif occurrence data which indicates the occurrence frequency of 614 different motifs and the chromosomal location data of each gene. Gene Ontology (http://www.geneontology.org) and TAIR (http://arabidopsis.org) databases were used to find the <em>molecular function</em> and <em>biological process</em> information of each gene in order to examine the biological accuracy of newly discovered clusters after using combined genomic data. The biological evaluation of the results showed that using combined genomic data to cluster genes resulted in new biologically relevant clusters.</p>

M.Ing. / This project investigates the use of Self-Organizing Maps (SOM) for feature recognition and analysis in 3D objects. Object data was generated to simulate data obtained from 3D scanning and trained using SOM. The trained data was analysed using speci cally developed software. The feature recognition and analysis process can be summarized as follows: a 3D object le is converted to a pure 3D data le, this data le is trained using the SOM algorithm after which the output is analyzed using a 3D object viewer and SOM data display.

Considering the commonality of breast cancer among women in the United States and the increasing popularity of precision medicine and data analytics in healthcare, the aim of this study was to use self-organizing maps (SOM) to profile and make decisions about breast cancer patients. Breast cancer mass measurements were combined with nine non-medical attributes—family income, history of cancer, level of education, preference of probability level, presence of dependents, employment status, marital status, age, and location—that were randomly generated based on recent population statistics and fed into a SOM. The SOM’s accuracy was evaluated at around 80%. To show the decision-making capabilities of the SOM, a subset of the patients were treated as new patients and placed on the map. Profiles of these clusters were created to show how decisions made about patients’ diagnosis, delivery, and treatment differed based on the cluster to which they belonged.

No description available.

Projetar máquinas capazes identificar pessoas é um problema cuja solução encontra uma grande quantidade de aplicações. Implementações em software de sistemas baseados em medições de características físicas pessoais (biométricos), estão começando a ser produzidos em escala comercial. Nesta categoria estão os sistemas de Reconhecimento Automático de Locutor, que se usam da voz como característica identificadora. No presente momento, os métodos mais populares são baseados na extração de coeficientes mel-cepstrais (MFCCs) das locuções, seguidos da identificação do locutor através de Hidden Markov Models (HMMs), Gaussian Mixture Models (GMMs) ou quantização vetorial. Esta preferência se justifica pela qualidade dos resultados obtidos. Fazer com que estes sistemas sejam robustos, mantendo sua eficiência em ambientes ruidosos, é uma das grandes questões atuais. Igualmente relevantes são os problemas relativos à degradação de performance em aplicações envolvendo um grande número de locutores, e a possibilidade de fraude baseada em vozes gravadas. Outro ponto importante é embarcar estes sistemas como sub-sistemas de equipamentos já existentes, tornando-os capazes de funcionar de acordo com o seu operador. Este trabalho expõe os conceitos e algoritmos envolvidos na implementação de um software de Reconhecimento Automático de Locutor independente de texto. Inicialmente é tratado o processamento dos sinais de voz e a extração dos atributos essenciais deste sinal para o reconhecimento. Após isto, é descrita a forma pela qual a voz de cada locutor é modelada através de uma rede neural de arquitetura Self-Organizing Map (SOM) e o método de comparação entre as respostas dos modelos quando apresentada uma locução de um locutor desconhecido. Por fim, são apresentados o processo de construção do corpus de vozes usado para o treinamento e teste dos modelos, as arquiteturas de redes testadas e os resultados experimentais obtidos numa tarefa de identificação de locutor. / The design of machines that can identify people is a problem whose solution has a wide range of applications. Software systems, based on personal phisical attributes measurements (biometrics), are in the beginning of commercial scale production. Automatic Speaker Recognition systems fall into this cathegory, using voice as the identifying attribute. At present, the most popular methods are based on the extraction of mel-frequency cepstral coefficients (MFCCs), followed by speaker identification by Hidden Markov Models (HMMs), Gaussian Mixture Models (GMMs) or vector quantization. This preference is motivated by the quality of the results obtained by the use of these methods. Making these systems robust, able to keep themselves efficient in noisy environments, is now a major concern. Just as relevant are the problems related to performance degradation in applications with a large number of speakers involved, and the issues related to the possibility of fraud by the use of recorded voices. Another important subject is to embed these systems as sub-systems of existing devices, enabling them to work according to the operator. This work presents the relevant concepts and algorithms concerning the implementation of a text-independent Automatic Speaker Recognition software system. First, the voice signal processing and the extraction of its essential features for recognition are treated. After this, it is described the way each speaker\'s voice is represented by a Self-Organizing Map (SOM) neural network, and the comparison method of the models responses when a new utterance from an unknown speaker is presented. At last, it is described the construction of the speech corpus used for training and testing the models, the neural network architectures tested, and the experimental results obtained in a speaker identification task.

Made available in DSpace on 2017-07-21T14:19:23Z (GMT). No. of bitstreams: 1 Kelly Lais.pdf: 3909972 bytes, checksum: a9e068fb926de9ca233857005e400c84 (MD5) Previous issue date: 2014-12-16 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / O Sensoriamento Remoto (SR) dispõe de tecnologias em constante crescimento e com grande potencial para a agricultura, tanto no gerenciamento de culturas, manejo de solo, bem como discriminações de feições da terra. Atualmente, há muitos métodos de análise e categorização de paisagens, que com a integração de dados de SR e técnicas de Sistemas de Informação Geográfica (SIG) apresentam alternativa promissora. Isto é, proporcionam maior facilidade na manipulação de dados geográficos, bem como otimização da validação a campo. Neste contexto, esta pesquisa foi realizada utilizando classificação digital não-supervisionada pela Rede Neural Artificial (RNA) Self-Organizing Maps (SOM) no reconhecimento de padrões de uso e cobertura da terra em um recorte de imagem orbital de alta resolução Rapideye pertencente à Bacia do Pitangui, o qual abrange o município de Ponta Grossa, localizado a centro-leste do Estado do Paraná. Primeiramente aplicouse a técnica NDVI (Índice de Vegetação por Diferença Normalizada) para estimular a separação das classes, principalmente os diferentes tipos de cultivos agrícolas, bem como cobertura florestal. A imagem orbital e NDVI foram segmentadas por meio de Análise de Imagem Baseada em Objeto (GEOBIA), gerando descritores com propriedades espaciais, espectrais e de textura, culminando no banco de dados relacional (BDR) com tais descritores. Mediante Análise de Componentes Principais (ACP) reduziu-se a dimensionalidade dos dados do BDR, selecionando os descritores mais significativos. A dimensionalidade foi reduzida, sem perda de informação, de 42 descritores para 21, a saber 6 espaciais, 12 espectrais e 3 de textura. Após esta preparação dos dados, utilizou-se a RNA SOM para o reconhecimento dos padrões pré-determinados a campo. As classes de uso e cobertura da terra discriminadas pela RNA SOM foram cultivos (cultivo 1, 2, 3 e 4), estradas e construções, cobertura florestal e corpos d’água. A RNA SOM culminou no agrupamento das classes cultivos inclusive em relação ao seu ciclo fenológico. A associação da banda artificial NDVI, com seus descritores às bandas espectrais, incrementou a separabilidade entre classes, tais como cobertura florestal e corpos d’água. As classes de uso e cobertura da terra foram validadas a campo, a exatidão global foi de 91% de acerto, com índice kappa de 0,9, considerado resultado excelente em valores de referência. Também foi realizado o teste estatístico F, o qual satisfez as hipóteses de nulidade nas áreas analisadas.Conclui-se que os métodos utilizados apresentaram eficácia, agilidade e baixo custo no mapeamento d o uso e cobertura da terra em escala detalhada. / Remote Sensing (RS) uses steadily growing technologies and presents great potential for agriculture, e.g. in crop and land management, as well as for discrimination of land features. Currently, there are many methods of analysis and landscape categorization that when integrated with RS data and Geographic Information Systems (GIS) techniques stage as promising alternatives. That is, they provide greater ease in handling spatial data as well as optimizing validation on the field. In this context, this study was carried out using unsupervised digital classification with Artificial Neural Network (ANN) Self- Organizing Maps (SOM), in order to recognize patterns of land cover and land use in part of a high-resolution Rapideye orbital image belonging to the Pitangui River Basin, which encompasses the city of Ponta Grossa, located in the Central-Eastern portion of the State of Paraná. Initially, NDVI (Normalized Difference Vegetation Index) technique was applied to stimulate the separation of classes, especially to evidence different types of agricultural crops and forest cover. The orbital image and the NDVI were segmented through Geographic Object-Based Image Analysis (GEOBIA), generating descriptors with spatial, spectral and textural properties, culminating in the relational database (RDB) with such descriptors. With Principal Component Analysis (PCA) dimensionality of the BDR data was reduced, selecting the most significant descriptors. Dimensionality was reduced without information loss, from 42 descriptors to 21, namely 6 spatial, 12 spectral and 3 textural. After this data preparation, ANN SOM was used to recognize predetermined patterns in the field. The classes of land cover and land use discriminated by ANN SOM were crops (crop 1, 2, 3 and 4), roads and buildings; forest cover and water bodies. The ANN SOM culminated in the grouping of crop classes including in relation to its phenological cycle. The association of the NDVI artificial band with descriptors to spectral bands, increased the separability between classes, such as forest cover and water bodies. Classes of land cover and land use were validated in the field, the global accuracy was 91%, with kappa index of 0.9 and considered to be excellent as reference values. F statistical test was also carried out and showed satisfiability in the analyzed areas. It is concluded that the methods used were effective, agile and low-cost in detailed scale mapping of land use and coverage.

Projetar máquinas capazes identificar pessoas é um problema cuja solução encontra uma grande quantidade de aplicações. Implementações em software de sistemas baseados em medições de características físicas pessoais (biométricos), estão começando a ser produzidos em escala comercial. Nesta categoria estão os sistemas de Reconhecimento Automático de Locutor, que se usam da voz como característica identificadora. No presente momento, os métodos mais populares são baseados na extração de coeficientes mel-cepstrais (MFCCs) das locuções, seguidos da identificação do locutor através de Hidden Markov Models (HMMs), Gaussian Mixture Models (GMMs) ou quantização vetorial. Esta preferência se justifica pela qualidade dos resultados obtidos. Fazer com que estes sistemas sejam robustos, mantendo sua eficiência em ambientes ruidosos, é uma das grandes questões atuais. Igualmente relevantes são os problemas relativos à degradação de performance em aplicações envolvendo um grande número de locutores, e a possibilidade de fraude baseada em vozes gravadas. Outro ponto importante é embarcar estes sistemas como sub-sistemas de equipamentos já existentes, tornando-os capazes de funcionar de acordo com o seu operador. Este trabalho expõe os conceitos e algoritmos envolvidos na implementação de um software de Reconhecimento Automático de Locutor independente de texto. Inicialmente é tratado o processamento dos sinais de voz e a extração dos atributos essenciais deste sinal para o reconhecimento. Após isto, é descrita a forma pela qual a voz de cada locutor é modelada através de uma rede neural de arquitetura Self-Organizing Map (SOM) e o método de comparação entre as respostas dos modelos quando apresentada uma locução de um locutor desconhecido. Por fim, são apresentados o processo de construção do corpus de vozes usado para o treinamento e teste dos modelos, as arquiteturas de redes testadas e os resultados experimentais obtidos numa tarefa de identificação de locutor. / The design of machines that can identify people is a problem whose solution has a wide range of applications. Software systems, based on personal phisical attributes measurements (biometrics), are in the beginning of commercial scale production. Automatic Speaker Recognition systems fall into this cathegory, using voice as the identifying attribute. At present, the most popular methods are based on the extraction of mel-frequency cepstral coefficients (MFCCs), followed by speaker identification by Hidden Markov Models (HMMs), Gaussian Mixture Models (GMMs) or vector quantization. This preference is motivated by the quality of the results obtained by the use of these methods. Making these systems robust, able to keep themselves efficient in noisy environments, is now a major concern. Just as relevant are the problems related to performance degradation in applications with a large number of speakers involved, and the issues related to the possibility of fraud by the use of recorded voices. Another important subject is to embed these systems as sub-systems of existing devices, enabling them to work according to the operator. This work presents the relevant concepts and algorithms concerning the implementation of a text-independent Automatic Speaker Recognition software system. First, the voice signal processing and the extraction of its essential features for recognition are treated. After this, it is described the way each speaker\'s voice is represented by a Self-Organizing Map (SOM) neural network, and the comparison method of the models responses when a new utterance from an unknown speaker is presented. At last, it is described the construction of the speech corpus used for training and testing the models, the neural network architectures tested, and the experimental results obtained in a speaker identification task.

This diploma thesis occupies itself with the subject of near synonymy, concretely with adjectives. On the basis of corpus linguistic methods two pairs of near synonyms have been researched - verschieden/unterschiedlich and bedeutend/bedeutsam. The 15 primary collocators (according to the syntactic position of each adjective) have been examined using the InterCorp parallel corpus methods in order to find out the most frequent Czech equivalence. Keywords: lexical-semantic relations, near synonymy, lexicography, corpora, cooccurrence analysis, Self Organizing Maps, CCDB

We used the self-organizing maps (SOMs) method on eight models that participated in the Coupled model intercomparison project phase 5 (CMIP5) and two different greenhouse gases (GHG) concentration experiments. The SOMs were created from the winter 500 hPa horizontal temperature flux for each model. The clustering by the SOM revealed that in addition to the three flux pathways found in reanalyses (Pacific, Atlantic and Siberian/continental pathway), superpositions of these occur for the free running climate models, which develop their dynamic more freely than the reanalyses. It was found that the general structure of fluxes is indirectly dependent on the GHG concentrations, as the derived results from SOM patterns are different between the two GHG concentrations. It is suggested that flux patterns change from stable cyclonic motion over the north pole to flux pathways that feature more meridional fluxes through the North Atlantic and North Pacific into the Arctic. / Die Methode der Self-Organizing Maps (SOMs) wurde auf acht CMIP5-Modelle mit jeweils zwei verschiedenen Treibhausgasszenarien angwendet. Die SOMs wurden für jedes Modell und jede der beiden Modelläufe für den horizontalen Temperaturfluss in 500 hPa im Winter erstellt. Zusätzlich zu den aus der Analyse von Reanalyse-Daten erwarteten drei Transportwegen (pazifisch, atlantisch und sibirisch/kontinental) wurden Überlagerungen dieser gefunden. Es konnte gezeigt werden, dass die grundsätzliche Struktur der Transporte indirekt abhängig von der Treibhausgaskonzentration ist. Die Ergebnisse deuten darauf hin, dass sich die generelle Struktur des atmosphärischen Transports von einer stabilen zyklonalen Bewegung über dem Nordpol sich zu Transporten verschiebt, welche meridionale Transporte über den Nortdatlantik und den Nordpazifik in die Arktis führen.

Depuis une quinzaine d'années, l'Imagerie par Résonance Magnétique fonctionnelle (IRMf) permet d'extraire de l'information sur le fonctionnement cérébral et particulièrement sur la localisation des processus cognitifs. L'information contenue par les acquisitions en IRMf est extraite à l'aide du modèle linéaire général et du processus d'inférence statistique. Bien que cette méthode dite « classique » ait permis de valider la plupart des modèles lésionnels de manière non invasive, elle souffre de certaines limites. Pour résoudre ce problème, différentes techniques d'analyse ont émergé et proposent une nouvelle façon d'interpréter les données de la neuroimagerie. Nous présentons deux nouvelles méthodes multivariées basées sur les cartes de Kohonen. Nos méthodes analysent les données IRMf avec le moins d'a priori possibles. En parallèle, nous tentons d'extraire de l'information sur les réseaux neuronaux impliqués dans les émotions. La première de ces méthodes s'intéresse à l'information de spécialisation fonctionnelle et la seconde à l'information de connectivité fonctionnelle. Nous présentons les résultats qui en découlent, puis chacune des méthodes est comparée à l'analyse dite classique en termes d'informations extraites. De plus, notre attention s'est focalisée sur la notion de valence émotionnelle et nous tentons d'établir l'existence d'un éventuel réseau partagé entre valence positive et valence négative. La constance de ce réseau est évaluée à la fois entre modalités perceptives et entre catégories de stimuli. Chacune des méthodes proposées permet de corroborer l'information recueillie par la méthode classique, en apportant de nouvelles informations sur les processus étudiés. Du point de vue des émotions, notre travail met en lumière un partage du réseau cérébral pour les va-lences négative et positive ainsi qu'une constance de cette information dans certaines régions cérébrales entre modalités perceptives et entre catégories. / In the last fifteen years, functional magnetic resonance imaging (fMRI) have been used to extract information about cognitive processes location. The information contained in fMRI acquisitions is usually extracted using the general linear model coupled to the statistical inference process. Although this classical method has validated noninvasively most of the lesional models, it suffers from some limitations. To solve this problem, various analysis techniques have emerged and propose a new way of interpreting neuroimaging data. In this thesis, we present two multivariate methods to analyze fMRI data with the least possible a priori. In parallel, we are trying to extract information about brain emotion processing. The first method focuses on the brain functional specialization and the second method on the brain functional connectivity. After results presentation, each method is compared to the so-called classical analysis in terms of extracted information. In addition, emphasis was put on the concept of emotional valence. We try to establish the existence of a possible split between positive and negative valence networks. The consistency of the network is evaluated across both perceptual modalities and stimuli categories. Each of the proposed methods are as accurate as the conventional method and provide new highlights on the studied processes. From the perspective of emotions, our work highlights a shared brain network for positive and negative valences and a consistency of this information in some brain regions across both perceptual modalities and stimuli categories.