Cereal grains are an important part of human diet; hence, there is a need to maintain high quality and these grains must be free of physical and biological contaminants. A procedure was developed to differentiate physical contaminants from wheat using NIR (1000-1600 nm) hyperspectral imaging. Three experiments were conducted to select the best combinations of spectral pre-processing technique and statistical classifier to classify physical contaminants: seven foreign material types (barley, canola, maize, flaxseed, oats, rye, and soybean); six dockage types (broken wheat kernels, buckwheat, chaff, wheat spikelets, stones, and wild oats); and two animal excreta types (deer and rabbit droppings) from Canada Western Red Spring (CWRS) wheat. These spectra were processed using five spectral pre-processing techniques (first derivative, second derivative, Savitzky-Golay (SG) smoothing and differentiation, multiplicative scatter correction (MSC), and standard normal variate (SNV)). The raw and pre-processed data were classified using Support Vector Machines (SVM), Naïve Bayes (NB), and k-nearest neighbors (k-NN) classifiers. In each experiment, two-way and multi-way classifications were conducted. Among all the contaminant types, stones, chaff, deer droppings and rabbit droppings were classified with 100% accuracy using the raw reflectance spectra and different statistical classifiers. The SNV technique with k-NN classifier gave the highest accuracy for the classification of foreign material types from wheat (98.3±0.2%) and dockage types from wheat (98.9±0.2%). The MSC and SNV techniques with SVM or k-NN classifier gave perfect classification (100.0±0.0%) for the classification of animal excreta types from wheat. Hence, the SNV technique with k-NN classifier was selected as the best model. Two separate model performance evaluation experiments were conducted to identify and quantify (by number) the amount of contaminant type present in wheat. The overall identification accuracy of the first degree of contamination (one contaminant type with wheat) and the highest degree of contamination (all the contaminant type with wheat) was 97.6±1.6% and 92.5±6.5%, for foreign material types; 98.0±1.8% and 94.3±6.2%r for dockage types; and 100.0±0.0% and 100.0±0.0%, respectively for animal excreta types. The canola, stones, deer, and rabbit droppings were perfectly quantified (100.0±0.0%) at all the levels of contaminations. / February 2016

Kernel methods are a popular choice in solving a number of problems in statistical machine learning. In this thesis, we propose new methods for two important kernel based classification problems: 1) learning from highly unbalanced large-scale datasets and 2) selecting a relevant subset of input features for a given kernel specification. The first problem is known as the rare class problem, which is characterized by a highly skewed or unbalanced class distribution. Unbalanced datasets can introduce significant bias in standard classification methods. In addition, due to the increase of data in recent years, large datasets with millions of observations have become commonplace. We propose an approach to address both the problem of bias and computational complexity in rare class problems by optimizing area under the receiver operating characteristic curve and by using a rare class only kernel representation, respectively. We justify the proposed approach theoretically and computationally. Theoretically, we establish an upper bound on the difference between selecting a hypothesis from a reproducing kernel Hilbert space and a hypothesis space which can be represented using a subset of kernel functions. This bound shows that for a fixed number of kernel functions, it is optimal to first include functions corresponding to rare class samples. We also discuss the connection of a subset kernel representation with the Nystrom method for a general class of regularized loss minimization methods. Computationally, we illustrate that the rare class representation produces statistically equivalent test error results on highly unbalanced datasets compared to using the full kernel representation, but with significantly better time and space complexity. Finally, we extend the method to rare class ordinal ranking, and apply it to a recent public competition problem in health informatics. The second problem studied in the thesis is known as the feature selection problem in literature. Embedding feature selection in kernel classification leads to a non-convex optimization problem. We specify a primal formulation and solve the problem using a second-order trust region algorithm. To improve efficiency, we use the two-block Gauss-Seidel method, breaking the problem into a convex support vector machine subproblem and a non-convex feature selection subproblem. We reduce possibility of saddle point convergence and improve solution quality by sharing an explicit functional margin variable between block iterates. We illustrate how our algorithm improves upon state-of-the-art methods.

Neural networks are seldom used as a modelling tool by statisticians. This is often due to the lack of knowledge in the eld of neural networks as neural networks are frequently perceived as mysterious methods that evolved from the eld of computer science. In this dissertation an attempt will be made to show that neural network methods are closely related to statistical methods. In particular we will show how a backpropagation neural network can be used for statistical applications like regression and classi cation which will include the setting up a of neural network for di erent objectives and also using a neural network for predictive inference. Through simulations we will show an e cient method to t a neural network in practical applications. A neural network will then be employed in a practical application to illustrate how to use a neural network in a regression or classi cation context. This application will also show the necessity of statistical knowledge when using a neural network as a modelling tool. / Dissertation (MSc)--University of Pretoria, 2010. / Statistics / unrestricted

Submitted by (ana.araujo@ufrpe.br) on 2016-08-16T14:12:24Z No. of bitstreams: 1 Simone Castelo Branco Simoes.pdf: 1283329 bytes, checksum: ab664570df5d0a685483c6dfc554deb4 (MD5) / Made available in DSpace on 2016-08-16T14:12:24Z (GMT). No. of bitstreams: 1 Simone Castelo Branco Simoes.pdf: 1283329 bytes, checksum: ab664570df5d0a685483c6dfc554deb4 (MD5) Previous issue date: 2007-02-27 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / In pattern recognition, the medical diagnosis has received great attention. In gene-ral, the emphasis has been to identify one best model for diagnostic forecast, measured according to generalization ability. In this context, ensembles methods have been eficients, can be considered on the improvement of performance in diagnostic tasks that demand greater precision. The bagging method, purposed from Breiman (1996), uses bootstrap to generate different samples of the training set, building classifiers with the generated samples and combining different forecasts for majority vote. In general, empirical estudies are done for evaluate the bagging performance. In this thesis, we investigate the bagging generalization ability for statistical usual classifiers and the multilayer perceptron net through sthocastic simulation. Different structures of separation of populations are build from especific distributions. Additionally, we make an application on diagnostic suport of brest cancer. The results were obtained using R. In general, we observed that bagging performance depends on the population separation behavior. In the application, bagging showed to be e±cient on sensibility improvement. / Em reconhecimento de padrões, o diagnóstico médico tem recebido grande atenção. Em geral, a ênfase tem sido a identificação de um melhor modelo de previsão diagnóstica, avaliado de acordo com a habilidade de generalização. Nesse contexto, métodos que combinam classificadores têm se mostrado muito eficazes, podendo ser considerados no melhoramento de desempenho em tarefas diagnósticas que exigem maior precisão. O método bagging, proposto por Breiman (1996), utiliza bootstrap para gerar diferentes amostras do conjunto de treinamento, construindo classificadores com as amostras geradas e combinando as diferentes previsões por voto majoritário. Em geral, estudos empíricos são realizados para avaliar o desempenho do bagging. Nesta dissertação , investigamos a habilidade de generalização do bagging para classificadores estatísticos usuais e a rede perceptron de múltiplas camadas através de simulações estocásticas. Diferentes estruturas de separação das populações são construídas a partir de distribuições específicas consideradas. Adicionalmente, realizamos uma alicação no suporte ao diagnóstico de câncer de mama. Os resultados foram obtidos utilizando o ambiente de programação análise de dados e gráficos R. Em geral, as simulações realizadas indicam que o desempenho do bagging depende do comportamento de separação das populações. Na aplicação, o bagging mostrou ser eficiente no melhoramento da sensibilidade.

Depuis les années 80, Les matériaux composites stratifié verre résine (SVR) ont été utilisés pour la construction des tuyaux et des réservoirs dans l'industrie chimique, y compris pour le stockage d’acides. Ce matériau composite présente une résistance supérieure à la corrosion. Cependant, des auteurs ont observé des ruptures accidentelles de réservoirs (horizontaux et verticaux) contenant des acides (chlorhydrique et sulfurique). Ces ruptures sont attribuées au mécanisme de corrosion sous contrainte (CSC). La corrosion des fibres de verre dans une solution acide est moins connue mais reste très importante. Ce mécanisme de corrosion, appelée désalcalinisation de la fibre peut provoquer la fissuration de la fibre de verre.Des essais de corrosion avec de l’acide chlorhydrique (37%) ont été effectués sur éprouvette SVR. Ces essais de corrosion ont été suivis par émission acoustique. Les observations au microscope électroniques à balayage (MEB) et les analyses physico-chimiques confirment la corrosion de fibres de verre dans une solution de HCl. L’utilisation de la micro-tomographie nous montre que cette technique permet d’avoir une information sur la profondeur d’attaque du matériau.Une approche statistique est utilisée pour caractériser les paramètres de la salve d’émission acoustique afin de les séparer. Le Clustering est fait en utilisant la méthode des k-moyennes. Trois classes d’émission acoustique distinctes ont ainsi été identifiées. L’analyse croisée de l’émission acoustique et des observations ont permis de relier les classes observées aux conséquences de la corrosion du SVR. / Since the 1980, Glass Reinforced Plastic (GRP) has been used for construction of pipes and tanks in the chemical industry, including the storage of mineral acids. This composite material offers superior and cost effective corrosion resistance. However, authors found accidental breakage of tanks (horizontal and vertical) containing mineral acids (hydrochloric and sulphuric). These failures are attributed to environmental stress-corrosion cracking (ESCC) mechanism. The corrosion of glass fibers in mineral acid solution is less known but very important. The mechanism of the corrosion, called leaching, is thought to induce tensile stresses in the surface of the glass. These stresses could be large enough to cause cracking of the fiber glass.Corrosion tests have been performed on GRP specimen. Aggressive environments used are hydrochloric acid (37%) This environment is known to react with E-glass. Corrosion tests have been monitored by acoustic emission.SEM observations and physicochemical analysis confirm the corrosion of glass fibers in HCl solution. The use of micro - tomography allows to have information on the depth of degradation of the material.Statistical approaches are used to characterize hit’s parameters. Clustering is made by using k-mean’s method. Three distinct acoustic emission classes are identified. Thanks to SEM observations and acoustic emission results, clusters can be assigned to the appearance of minor defects in the material.

This master‘s thesis deals with using surface texture parameters for conformity assessment and process control. The aim of the thesis was to create an overview of surface texture parameters, an overview of procedures for conformity assessment and process control using surface texture parameters, practical application on an industrial product and practical recommendations. The thesis includes an overview of profile and areal surface texture parameters, including an original translation of terms of the areal method. Methods of conformity assessment and process control in connection with the surface texture parameters are also described. Statistical analysis was performed based on 7200 values of surface parameters and 1843200 values of profile parameters which were measured on parts from serial production. The last chapter includes practical recommendations.

There has been an increase of attention placed on the diagnosis of Attention Deficit Hyperactivity Disorder (ADHD), within South Africa. This has led to a number of controversies surrounding the legitimacy of ADHD diagnoses. And how effective the systems of categorising and diagnosing disorders are in aiding a number of practitioners in formulating a disorder. There is a substantial agreement within the literature that the understanding of ADHD is limited, the focus is mainly on the symptoms of disorders. This study explores the perceptions practitioners in the field, in identifying the effectiveness of the Diagnostic and Statistical Manual for Mental disorders (DSM) is for diagnosing ADHD. It became evident throughout this study that there is no consensus around the efficacy of the DSM. Furthermore, ADHD is not completely understood and therefore creates serious implications for the treatment and diagnosis of the disorder.

Disertacijoje nagrinėjamas mokslo publikacijų automatinio klasifikavimo uždavinys. Šis uždavinys sprendžiamas taikant tikimybinius diskriminantinės analizės metodus. Pagrindinis darbo tikslas - sukurti konstruktyvius klasifikavimo metodus, kurie leistų atsižvelgti į mokslo publikacijų tekstų specifiką. Disertaciją sudaro įvadas, trys pagrindiniai skyriai, rezultatų apibendrinimas, naudotos literatūros ir autoriaus publikacijų disertacijos tema sąrašai ir vienas priedas. Įvadiniame skyriuje aptariama tiriamoji problema, darbo aktualumas, aprašomas tyrimų objektas, formuluojamas pagrindinis darbo tikslas bei uždaviniai, aprašoma tyrimų metodika, darbo mokslinis naujumas, pasiektų rezultatų praktinė reikšmė, ginamieji teiginiai. Įvado pabaigoje pristatomos disertacijos tema autoriaus paskelbtos publikacijos ir pranešimai konferencijose bei disertacijos struktūra. Pirmajame skyriuje matematiškai apibrėžtas ir detalizuotas sprendžiamas uždavinys, pateikta analitinė kitų autorių darbų apžvalga. Pasirinkti ir išanalizuoti keli populiarūs klasifikavimo algoritmai, kurie eksperimentinėje darbo dalyje lyginti su autoriaus pasiūlytaisiais. Antrajame skyriuje sudarytas mokslo terminijos pasiskirstymo tekstuose tikimybinis modelis, išskirti atskiri atvejai, galiojant įvestoms prielaidoms apie terminų tarpusavio sąryšių formas, pasiūlytos modelio identifikavimo procedūros bei suformuluoti konstruktyvūs mokslo publikacijų klasifikavimo algoritmai. Trečiajame skyriuje pateikti pagrindiniai... [toliau žr. visą tekstą] / The dissertation considers the problem of automatic classification of scientific publications. The problem is addressed by using probabilistic methods of the discriminant analysis. The main goal of the dissertation is to create constructive classification methods that would allow to take into consideration specificity of scientific publication text. The dissertation consists of Introduction, 3 chapters, Conclusions, References, list of author's publications, and one Appendix. The introduction reveals the investigated problem, importance of the thesis and the object of research and describes the purpose and tasks of the paper, research methodology, scientific novelty, the practical significance of results examined in the paper and defended statements. The introduction ends in presenting the author’s publications on the subject of the defended dissertation, offering the material of made presentations in conferences and defining the structure of the dissertation. Chapter 1 presents a detailed mathematical formulation of the considered problem, reviews scientific papers on the subject, and analyses a few popular classification algorithms that in Chapter 3 are compared to the ones proposed in this paper. Chapter 2 develops the probabilistic model for scientific terminology distribution over texts, discusses special cases of the model under specific assumptions on forms of terminology relations, suggests the model identification procedures, and formulates constructive scientific... [to full text]

The dissertation considers the problem of automatic classification of scientific publications. The problem is addressed by using probabilistic methods of the discriminant analysis. The main goal of the dissertation is to create constructive classification methods that would allow to take into consideration specificity of scientific publication text. The dissertation consists of Introduction, 3 chapters, Conclusions, References, list of author's publications, and one Appendix. The introduction reveals the investigated problem, importance of the thesis and the object of research and describes the purpose and tasks of the paper, research methodology, scientific novelty, the practical significance of results examined in the paper and defended statements. The introduction ends in presenting the author’s publications on the subject of the defended dissertation, offering the material of made presentations in conferences and defining the structure of the dissertation. Chapter 1 presents a detailed mathematical formulation of the considered problem, reviews scientific papers on the subject, and analyses a few popular classification algorithms that in Chapter 3 are compared to the ones proposed in this paper. Chapter 2 develops the probabilistic model for scientific terminology distribution over texts, discusses special cases of the model under specific assumptions on forms of terminology relations, suggests the model identification procedures, and formulates constructive scientific... [to full text] / Disertacijoje nagrinėjamas mokslo publikacijų automatinio klasifikavimo uždavinys. Šis uždavinys sprendžiamas taikant tikimybinius diskriminantinės analizės metodus. Pagrindinis darbo tikslas - sukurti konstruktyvius klasifikavimo metodus, kurie leistų atsižvelgti į mokslo publikacijų tekstų specifiką. Disertaciją sudaro įvadas, trys pagrindiniai skyriai, rezultatų apibendrinimas, naudotos literatūros ir autoriaus publikacijų disertacijos tema sąrašai ir vienas priedas. Įvadiniame skyriuje aptariama tiriamoji problema, darbo aktualumas, aprašomas tyrimų objektas, formuluojamas pagrindinis darbo tikslas bei uždaviniai, aprašoma tyrimų metodika, darbo mokslinis naujumas, pasiektų rezultatų praktinė reikšmė, ginamieji teiginiai. Įvado pabaigoje pristatomos disertacijos tema autoriaus paskelbtos publikacijos ir pranešimai konferencijose bei disertacijos struktūra. Pirmajame skyriuje matematiškai apibrėžtas ir detalizuotas sprendžiamas uždavinys, pateikta analitinė kitų autorių darbų apžvalga. Pasirinkti ir išanalizuoti keli populiarūs klasifikavimo algoritmai, kurie eksperimentinėje darbo dalyje lyginti su autoriaus pasiūlytaisiais. Antrajame skyriuje sudarytas mokslo terminijos pasiskirstymo tekstuose tikimybinis modelis, išskirti atskiri atvejai, galiojant įvestoms prielaidoms apie terminų tarpusavio sąryšių formas, pasiūlytos modelio identifikavimo procedūros bei suformuluoti konstruktyvūs mokslo publikacijų klasifikavimo algoritmai. Trečiajame skyriuje pateikti pagrindiniai... [toliau žr. visą tekstą]

The objectives of our study were to determine the potential of 1H MRS in detecting (1) early colonic inflammation, (2) effects of various fatty acids on normal colon and (3) their effects on IBD. Sprague dawley rat fed with 2% carrageenan was used as a model of IBD. Flaxseed oil served as ω-3, corn oil as ω-6 and beef tallow as saturated fatty acid sources. Control group animals were fed 5% corn oil, whereas, those in high-fat diet groups received an additional 7% of the respective fatty acids. After 2 weeks, 1H MRS and histology were conducted on excised colonic mucosa. Statistical classification strategy (SCS) used for analyzing 1H MRS data achieved an accuracy of 82 % in stage 1, 90-100% in stage 2 and 96-100% in stage 3. This implies that 1H MRS is a sensitive tool to diagnose early IBD and the effects of dietary fat on IBD.