Thesis (Ph. D.)--University of Washington, 1990. / Vita. Includes bibliographical references (leaves [249]-290).

Thesis (Ph. D.)--University of Michigan, 1961. / eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 502-523).

Thesis (Ph.D.)--Lehigh University, 2008. / Adviser: R. A. Pearson.

Thesis (Ph. D.)--University of Michigan, 1961. / eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 502-523).

Thesis (Ph. D.)--University of Washington, 1990. / Vita. Includes bibliographical references (leaves [249]-290).

Pattern recognition has become an accessible tool in developing advanced adaptive products. The need for such products is not diminishing but on the contrary, requirements for systems that are more and more aware of their environmental circumstances are constantly growing. Feed-forward neural networks are used to learn patterns in their training data without the need to discover by hand the relationships present in the data. However, the problem of estimating the required size of the neural network is still not solved. If we choose a neural network that is too small for a particular given task, the network is unable to "comprehend" the intricacies of the data. On the other hand if we choose a network size that is too big for the given task, we will observe that there are too many parameters to be tuned for the network, or we can fall in the "Curse of dimensionality" or even worse, the training algorithm can easily be trapped in local minima of the error surface. Therefore, we choose to investigate possible ways to find the 'Goldilocks' size for a feed-forward neural network (which is just right in some sense), being given a training set. Furthermore, we used a common paradigm used by the Roman Empire and employed on a wide scale in computer programming, which is the "Divide-et-Impera" approach, to divide a given dataset in multiple sub-datasets, solve the problem for each of the sub-dataset and fuse the results of all the sub-problems to form the result for the initial problem as a whole. To this effect we investigated modular neural networks and their performance.

The very large dimensionality of real world datasets is a challenging problem for classification algorithms, since often many features are redundant or irrelevant for classification. In addition, a very large number of features leads to a high computational time for classification algorithms. Feature selection methods are used to deal with the large dimensionality of data by selecting a relevant feature subset according to an evaluation criterion. The vast majority of research on feature selection involves conventional single-label classification problems, where each instance is assigned a single class label; but there has been growing research on more complex multi-label classification problems, where each instance can be assigned multiple class labels. This thesis proposes three types of new Multi-Label Correlation-based Feature Selection (ML-CFS) methods, namely: (a) methods based on hill-climbing search, (b) methods that exploit biological knowledge (still using hill-climbing search), and (c) methods based on genetic algorithms as the search method. Firstly, we proposed three versions of ML-CFS methods based on hill climbing search. In essence, these ML-CFS versions extend the original CFS method by extending the merit function (which evaluates candidate feature subsets) to the multi-label classification scenario, as well as modifying the merit function in other ways. A conventional search strategy, hill-climbing, was used to explore the space of candidate solutions (candidate feature subsets) for those three versions of ML-CFS. These ML-CFS versions are described in detail in Chapter 4. Secondly, in order to try to improve the performance of ML-CFS in cancer-related microarray gene expression datasets, we proposed three versions of the ML-CFS method that exploit biological knowledge. These ML-CFS versions are also based on hill-climbing search, but the merit function was modified in a way that favours the selection of genes (features) involved in pre-defined cancer-related pathways, as discussed in detail in Chapter 5. Lastly, we proposed two more sophisticated versions of ML-CFS based on Genetic Algorithms (rather than hill-climbing) as the search method. The first version of GA-based ML-CFS is based on a conventional single-objective GA, where there is only one objective to be optimized; while the second version of GA-based ML-CFS performs lexicographic multi-objective optimization, where there are two objectives to be optimized, as discussed in detail in Chapter 6. In this thesis, all proposed ML-CFS methods for multi-label classification problems were evaluated by measuring the predictive accuracies obtained by two well-known multi-label classification algorithms when using the selected featuresม namely: the Multi-Label K-Nearest neighbours (ML-kNN) algorithm and the Multi-Label Back Propagation Multi-Label Learning Neural Network (BPMLL) algorithm. In general, the results obtained by the best version of the proposed ML-CFS methods, namely a GA-based ML-CFS method, were competitive with the results of other multi-label feature selection methods and baseline approaches. More precisely, one of our GA-based methods achieved the second best predictive accuracy out of all methods being compared (both with ML-kNN and BPMLL used as classifiers), but there was no statistically significant difference between that GA-based ML-CFS and the best method in terms of predictive accuracy. In addition, in the experiment with ML-kNN (the most accurate) method selects about twice as many features as our GA-based ML-CFS; whilst in the experiments with BPMLL the most accurate method was a baseline method that does not perform any feature selection, and runs the classifier once (with all original features) for each of the many class labels, which is a very computationally expensive baseline approach. In summary, one of the proposed GA-based ML-CFS methods managed to achieve substantial data reduction, (selecting a smaller subset of relevant features) without a significant decrease in predictive accuracy with respect to the most accurate method.

Made available in DSpace on 2014-06-11T19:32:50Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-03-02Bitstream added on 2014-06-13T20:44:06Z : No. of bitstreams: 1 chimendes_vcg_dr_guara.pdf: 3671160 bytes, checksum: 460d8025fe5913fb7e71a5cd2ccc933d (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Este trabalho tem por objetivo identificar possíveis causas da lacuna existente entre as pesquisas acadêmicas e a aplicação destas pelas empresas, traduzindo-se em um baixo grau de inovação nos produtos ou serviços prestados de que resulta o pequeno número de patentes concedidas em nosso País. Foi utilizada a técnica de análise de conteúdo para exame dos dados coletados. Dado o grau de subjetividade deste trabalho as hipóteses formuladas não são irrefutáveis, porém a abordagem procura possibilitar que a ciência prospere também através do reconhecimento de seus erros e da adoção de medidas destinadas a eliminá-los. As mudanças tecnológicas tornam a concorrência cada vez mais acirrada em todos os níveis, inclusive entre regiões e países. O conhecimento tornou-se essencial para a geração de riqueza e para a promoção do bem–estar social. A inovação é um dos principais fatores para que o emprego e a empresa não entrem em um estado estacionário rumo à extinção. Fortalecer a capacidade inovadora do País, com a crescente utilização da tecnologia gerada internamente, faz com que a tecnologia, no sentido econômico, seja um bem passível de comercialização. Sabe-se que a produção de conhecimento no Brasil é significativa, porém a transformação desse conhecimento em resultado econômico ainda tem muito que avançar. Além disso, o país investe pouco em P&D (pesquisa e desenvolvimento), se comparado às economias mais desenvolvidas. As principais causas apontadas para esse distanciamento estão relacionadas às diferentes missões da universidade, da empresa e do Estado, à formação dos recursos humanos, à baixa cultura empreendedora e às dificuldades oriundas dos marcos regulatórios, tudo servindo de obstáculo para alavancar o desenvolvimento científico e tecnológico de produtos, processos e serviços / This work aims to identify possible causes of the gap between academic research and its utilization by companies, causing a low degree of innovation in products or services which results in the small number of patents granted in our country. We used the technique of content analysis to examine the data collected. Given the degree of subjectivity of the working hypotheses, they are not irrefutable; however the approach of this work seeks to enable science to prosper also by recognizing its errors and by adopting the necessary steps to eliminate them. Technological changes contribute to the increasingly fierce competition on all levels, including among regions and countries. Knowledge has become essential for generation of wealth and promotion of social welfare. Innovation is a key factor for employment and for companies to avoid a stationary state which would certainly lead to its extinction. Strengthening the innovative capacity of the country, with the increasing use of domestically generated technology, makes the technology an asset which can be marketed in the economic sense. It is known that the knowledge production in Brazil is significant, however transferring that knowledge into economic output still has a long way to go. Moreover, the country invests little in R & D (research and development) compared to more developed economies. The main causes for this detachment are related to the different missions of universities, companies and State, the training of human resources, the low entrepreneurial culture and the difficulties arising from regulatory frameworks, all serving as a barrier to enhance the scientific and technological development of products, procedures and services

The work presented in this thesis is divided into four chapters. Chapter I describes the control of the crystallization behavior of syndiotactic polystyrene (sPS) by chemical modification of sPS using sulfonation. In Chapter II, preparation of wood-ceramic composites through sol-gel processes is described. Chapter III describes the end-functionalization of poly(ethylene oxide)s (PEOs), and the interfacial properties of the functionalized PEOs. Finally, the distribution of chain conformation of PEO in the liquid state is discussed in Chapter IV. Direct sulfonation of highly stereoregular syndiotactic polystyrene (sPS) has been accomplished in chloroform. The degree of sulfonation can be effectively controlled. The crystallization behavior of sulfonated sPS is considerably different than the unmodified polymer. The crystallinity and the crystallization kinetics of sPS decrease with increasing extent of modification. Wood-ceramic composites were prepared by introducing a ceramic component into pine, a softwood, through sol-gel processes. Several ceramic precursors, including SiCl$\sb4$, Si(OCH$\sb3)\sb4$, and CH$\sb3$SiCl$\sb3$, were used. The ceramic content introduced into the wood structure can be controlled by varying the reaction time and the moisture content of wood. The modification of wood by CH$\sb3$SiCl$\sb3$ in supercritical CO$\sb2$ penetrated the entire wood structure, generating a macroscopically uniform distribution of the ceramic component in wood. Samples with one or both ends of monodisperse PEO functionalized with perfluorodecanoyl groups (PEO$\sp{\rm F}$ or PEO$\sp{\rm F2}$) were synthesized and blends of these end-capped PEOs with PEOs of the same molecular weight (M$\sb{\rm n}$$\sim$2000 - $\sim$16000) were prepared as cast films. Due to the lower surface energy of the fluorocarbon end groups, the modified PEOs preferentially adsorb to the free polymer surface. The surface concentration of the perfluoroalkyl end groups was measured by XPS which indicates that perfluoroalkyl chain ends adsorb to the polymer surface in a reasonably close-packed fashion (at all molecular weights) and leave a zone depleted of fluorine immediately beneath the highly fluorinated surface region. There is only a slight effect of molecular weight on surface fluorine content indicating a "stretched brush" conformation for the higher molecular weight samples. The adsorption of these polymers at the air-water interface was studied as well. The packing density and the orientation angle of the fluorinated chain end segregated at the interface were assessed by external reflectance IR. Isotropic Raman spectra of PEO in aqueous solution and in the melt were simulated by superposing calculated spectra resulting from a series of normal coordinate calculations performed for an ensemble of conformers, and compared with the corresponding experimental results. The conformational distribution for poly(ethylene oxide) in the molten state favors the tgg' conformer, and the aqueous solution of poly(ethylene oxide) contains mostly tgt conformers. The results for PEO were supported by measurements and computations made using 1,2-dimethoxyethane as a model.

The Moravian town of Zlin has now dual look: It is not only historical important architectural monument, but it is also dynamic town with vital public and bussiness life and University of Tomas Baťa. The factory covers important architectural and spiritual heritage, which is even today everywhere visible. Designed science-technical park is situated in Zlin, in previous Baťa´s factory, in the place of late swimming pool for employee. Manufactory is typical of its urbanistic solution. Location of science-technical park is on the border line of the “new“ and “old“ centre of Zlin and shakeups inauguration centre-line of previous factory.