Small carnivores and middle-sized mammals (mesomammals) are ubiquitous in fossil sites in South Africa, but their taphonomy is poorly understood. This study presents an actualistic investigation of bone modification by two captive small carnivores; the caracal (Caracal caracal) and honey badger (Mellivora capensis), housed at the Johannesburg Zoo. The carnivores were fed domestic rabbit (Oryctolagus cuniculus) carcasses as proxies for mesomammals and the bone modification of the resulting refuse and scatological assemblages were assessed in terms of their skeletal part representation, breakage patterns, digestive modifications and tooth marks. The investigation revealed that skeletal part representation and breakage patterns in the caracal and honey badger assemblages resembled those reported from other small carnivores. The caracal and honey badger assemblages were distinct from other carnivores in having overall light digestive modifications and a high frequency of tooth marks. Digestion was greater and tooth marks less frequent in the caracal than in the honey badger. Results were applied to the fossil assemblage of Cooper’s D which has a large assemblage of mesomammals and small carnivores. While a taphonomic analysis of Cooper’s D has not been published, initial results suggest that small carnivores had a great potential as contributors in the formation of the assemblage. The findings of this study emphasise the need for employing a variety of bone modifications in the identification of a small carnivore as an accumulator since there is rarely a single characteristic that is diagnostic for a particular carnivore.

A research report submitted in partial fulfilment of the requirements for the degree of Master of Science in Engineering (Mining) to the faculty of engineering and the built environment, University of the Witwatersrand, Johannesburg Date of final submission 25 May 2017 / The Australasian code for Reporting of Exploration Results, Mineral Resources and Ore Reserves, of the Joint Ore Reserves Committee (JORC) sets out minimum standards, recommendations and guidelines for Public Reporting in Australasia. (JORC, (2012)). The Committee for Mineral Reserve International Reporting Standards (CRIRSCO) created a set of standard international definitions for reporting Mineral Resources and Mineral Reserves based on the evolving JORC code’s definitions (CRIRSCO, (2013)). CRIRSCO’s members are National Reporting Organisations (NRO’s) which are responsible for developing mineral reporting codes for Australia (JORC), Canada (CIM Standing Committee on Reserve Definitions), Chile (National Committee), Europe (PERC), Russia (NAEN), South Africa (SAMCODES) and USA (SME) (JORC, (2012)). The NRO’s for; South Africa (SAMREC), Australia (JORC) and Canada (CIM Standing Committee on Reserve Definitions) published supporting standards for Coal Resource and Reserve Classification and Reporting namely, South African National Standard: South African Guide to the Systematic Evaluation of Coal Resources and Coal Reserves (SANS10320:2004), the Australian Guidelines for the Estimation and Classification of Coal Resources (Australian Guidelines (2014)) and the GSC Paper 88-21: A Standardized Coal Resource/Reserve Reporting System for Canada (Hughes, et al., (1989)). With the objective to identify the most appropriate Coal Resource Classification approach for the Witbank Coalfields in South Africa, Coal Resource Classification methods applied elsewhere in the world were investigated, these countries include Canada and Australia. SANS10320:2004 relies on a minimum drillhole spacing dependant on two different coal seam deposit types, whereas the Australian Guideline for the Estimation and Classification of Coal Resources (2014) provide a guide as to which geological aspects need to be considered when classifying a coal deposit into the appropriate confidence category, and no fixed drillhole spacing is recommended. The Canadian Standardized Coal Resource/Reserve Reporting System (1989) differs from the afore mentioned standards in that it is a prescriptive method based on specific levels of geological complexity, governed by specific fixed parameters. None of the other Coal Reporting codes/standards use a broad sweeping fixed drillhole spacing to classify Coal Resources as in South Africa. It is noted from experience as well as by Coal Resource Classification methods used elsewhere in the world that the use of proposed fixed drillhole spacing, such as currently in use in SANS10320:2004, is an unsatisfactory method for assessing the uncertainty and variability associated with coal deposits. The Coal Resource Classification methodologies utilised on a local scale in South Africa, were investigated to establish how mining houses manage and assess the variability in their Coal Resources. Fourteen mines operating throughout the Witbank coalfield were compared, it was found that although the Coal Resource Classification of the governing code requires a 350m drillhole spacing for highest level of confidence, the mines drill to a much smaller grid for increased confidence. Despite this, the mines still report on the SANS10320:2004 minimum standard in the public domain. A map was created based on the average drillhole spacing drilled per mine. From this it was deduced that there are zones of higher coal seam variability which required a closer spaced drilling grid to derive sufficient geological confidence in the estimates. Based on these deductions four zones of comparable continuity/variability, were identified. The zones identified by means of geological investigation and those identified by differences in variability as perceived by the Competent Person (CP) correlate. The highest variability and smallest drillhole spacing is located toward the western portion of the coalfield whereas the lowest variability with the largest drillhole spacing is located toward the east. The geologically complex Witbank coalfield was divided into four geo-zones/domains based on the depositional environment, basement rocks and post depositional influences. It is evident that a suitable Coal Resource Classification approach; which considers the characteristics of the geozones are followed. The question of which other classification methods are appropriate if not a predetermined drillhole spacing is addressed by this research. Statistics on relevant variables can provide a measure of uncertainty and therefore reliability in the estimates, for this reason three methods of uncertainty and probability characterisation were investigated. Of the three, namely; Non-linear estimation approach, conditional simulation (CS) and global estimation variance (GEV), the latter was deemed the most appropriate. GEV forms the basis of Drillhole Spacing Analysis (DHSA) and was applied to a mid-sized coal mine within the western portion of the Witbank coalfield. The analysis did not result in robust Coal Resource classification of estimates but rather provided more insight into the variability of the deposit. The results of DHSA are easily manipulated and are open for interpretation, it is therefore suggested as a valuable exercise/tool for understanding and assessing coal seam variability and to be used as a guide in Coal Resource classification. Onsite practical geological information should not be underestimated and geostatistics should always confirm the geology. A purely mathematical approach to Coal Resource classification would be a gross oversight, a combination of geological factors in association with statistical inferences is suggested. A scorecard method with associated weights is proposed to improve the confidence in the Coal Resource classification. / MT 2017

Nesta dissertação, são apresentados os processos de caracterização, classificação e de compartimentação geomecânica dos maciços rochosos, com a utilização dos dados pertinentes às fases de estudos nos projetos de engenharia. Como área de pesquisa analisou-se os dados referentes aos estudos da fundação da barragem, de uma Usina Hidroelétrica, localizada no Rio Pelotas, entre os estados do Rio Grande do Sul (Margem esquerda) e Santa Catarina (Margem Direita) em um sítio constituído por uma seqüência de derrames basálticos da Formação Serra Geral na Bacia do Paraná. Este maciço revela-se verticalmente heterogêneo, constituído por uma seqüência de 15 (quinze) derrames (relativamente pouco espessos) que possuem distintas características tanto sobre os aspectos geomecânicos, estruturais e tectônicos; quanto da diferenciação litológica de cada um - que compreende a seqüência de basalto denso, basalto vesículo amigdaloidal e brecha basáltica. O intuito da pesquisa foi mostrar a caracterização do maciço abrangendo aspectos estratigráficos, litológicos, estruturais, tectônicos e geomecânicos, para que fosse possível a formulação da compartimentação e dos modelos geológicos e geomecânicos do maciço. Na composição da base de dados foi realizado um levantamento de campo por meio de investigações de superfície e subsuperfície, para que posteriormente fossem realizados os trabalhos de gabinete e as análises laboratoriais. Para o estabelecimento do modelo geológico orientou-se pelos dados litoestratigráficos, estruturais e tectônicos obtidos através dos mapeamentos de superfície e das investigações de subsuperfície. A compartimentação geomecânica foi composta por informações oriundas dos itens de caracterização do maciço (alteração, fraturamento, Rock qualitiy designation - RQD, e permeabilidade); e pela classificação geomecânica, quanto a esta, empregaram-se as metodologia Rock Mass Rating - RMR, Quality - Q e Geological Strength Index - GSI. O propósito dessa classificação visava estabelecer um paralelo entre os parâmetros metodológicos utilizados e os resultados obtidos - em que se notou uma grande dependência da compartimentação geomecânica ao modelo geológico. / In this dissertation, the processes of characterization, classification and geomechanical compartmentation of the rocky mass range are investigated with the utilization of data resulting from the study phases in engineering projects. As a research area, the data referring to the studies of the foundation of a hydroelectric power plant dam, located on Rio Pelotas, between the states of Rio Grande do Sul (left bank) and Santa Catarina (right bank) on a site formed by a sequence of basaltic flood of Serra Geral formation at the Paraná Basin have been analyzed. This rock mass range, vertically heterogeneous, is formed by a sequence of 15 (fifteen) floods (relatively little solid) which possess distinct characteristics on the geomechanical, structural and tectonic aspects, as well as on the lithological differentiation of each one - which includes the sequence of dense basalt, vesicles-amygdaloidal basalt and basaltic breach. The intention of the research was to show the characterization of the mass range, embracing stratigraphic, lithological, structural, tectonic and geomechanical aspects, so that it would be possible the formulation of the compartmentation and geological and geomechanical models of the mass range. In the database composítion, a field survey was carried out through surface and subsurface investigations, so that afterwards the laboratorial analysis could be accomplished. For the establishment of the geological model, the lithostratigraphic, structural and tectonic data have been examined, which have been obtained through the mapping of the surface and subsurface investigations. The geomechanical compartmentation was put together by the information resulting from the characterization items of the mountain range (alteration, fracture, Rock quality designation - RQD and permeability); and by the geomechanical classification; for that matter, the Rock Mass Rating - RMR, Quality - Q and Geological Strength Index - GSI methodologies have been employed. The purpose of this classification was to establish a parallel between the methodological parameters used and the results obtained - where a large dependence of the geomechanical compartmentation on the geological model had been observed.

There is an increasing need of automation for routine tasks like sorting agricultural produce in large scale post-harvest processing. Among different kinds of sensors used for such automation tasks, near-infrared (NIR) technology provides a rapid and effective solution for quantitative analysis of quality indices in food products. As industries and farms are adopting modern data-driven technologies, there is a need for evaluation of the modelling tools to find the optimal solutions for problem solving. This study aims to understand the process of evaluation of the modelling tools, in view of near-infrared data obtained from green leafy vegetables. The first part of this study deals with prediction of the type of leafy green vegetable from the near-infrared reflectance spectra non-destructively taken from the leaf surface. Supervised classification methods used for the classification task were k-nearest neighbors (KNN), support vector machines (SVM), linear discriminant analysis (LDA) classifier, regularized discriminant analysis (RDA) classifier, naïve Bayes classifier, bagged trees, random forests, and ensemble discriminant subspace classifier. The second part of this study deals with prediction of total glucosinolate and total polyphenol contents in leaves using Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR). Optimal combination of predictors were chosen by using recursive feature elimination. NIR spectra taken from 283 different samples were used for classification task. Accuracy rates of tuned classifiers were compared for a standard test set. The ensemble discriminant subspace classifier was found to yield the highest accuracy rates (89.41%) for the standard test set. Classifiers were also compared in terms of accuracy rates and F1 scores. Learning rates of classifiers were compared with cross-validation accuracy rates for different proportions of dataset. Ensemble subspace discriminants, SVM, LDA and KNN were found to be similar in their cross-validation accuracy rates for different proportions of data. NIR spectra as well as reference values for total polyphenol content and total glucosinolate contents were taken from 40 samples for each analyses. PLSR model for total glucosinolate prediction built with spectra treated with Savitzky-Golay second derivative yielded a RMSECV of 0.67 μmol/g of fresh weight and cross-validation R2 value of 0.63. Similarly, PLSR model built with spectra treated with Savitzky-Golay first derivative yielded a RMSECV of 6.56 Gallic Acid Equivalent (GAE) mg/100g of fresh weight and cross-validation R-squared value of 0.74. Feature selection for total polyphenol prediction suggested that the region of NIR between 1300 - 1600 nm might contain important information about total polyphenol content in the green leaves.

Economic theory does not always specify the functional relationship between dependent and explanatory variables, or even isolate a particular set of covariates. This means that model uncertainty is pervasive in empirical economics. In this paper, we indicate how Bayesian semi-parametric regression methods in combination with stochastic search variable selection can be used to address two model uncertainties simultaneously: (i) the uncertainty with respect to the variables which should be included in the model and (ii) the uncertainty with respect to the functional form of their effects. The presented approach enables the simultaneous identification of robust linear and nonlinear effects. The additional insights gained are illustrated on applications in empirical economics, namely willingness to pay for housing, and cross-country growth regression.

"Precise and accurate localization and motion classification is an emerging fundamental areas for scientific research and engineering developments. Such science and technology began from the broad out door area applications, and gradually grew into smaller and more complicated in-door area and more recently it is proceeding into in-body area networking for medical applications. Localization and motion classification technologies have their own specific challenges depending on the application and environment, which are left for scientists and engineers to overcome. One major challenge is that location estimation and motion classification often use hand-held devices or wearable sensors. Such devices and sensors usually work in indoor, near body environments and the human object has certain effects on the measurements. In that situation, existing mathematical models for general environments are no longer accurate and new models and analytical approaches are required to deal with the human body effects. This has opened opportunities for researchers to tackle a number of demanding problems. This dissertation focuses on three novel problems in localization and motion classification using radio propagation (RF) modeling, in and around the human body. (1) We develop an empirical Time-of-Arrival (TOA) ranging error model for radio propagation from body-mounted sensors to external access points, for human body tracking in indoor environment. This model reflects the effects of human angular motion on TOA ranging estimation, which enables accurate analysis for conventional TOA-based human tracking systems. (2) We use empirical data collected from a RF connection between a pair of body-mounted sensors to classify seven frequently appeared human body motions. This RF based classification approach has enabled health monitoring applications for first responders, hospital patient, and elderly care centers and in most of the situations it can replace the costly video base monitoring systems. (3) We use radio propagation models from body-mounted sensor to medical implants and the moving pattern of micro-robots inside the body to analyze the accuracy of hybrid localization inside the human body. This analysis demonstrates the feasibility of millimeter level of accurate localization inside the human body, which opens up possibilities for 3D reconstruction of the interior of human GI tract."

Bone cement surgery is a new technique widely used in medical field nowadays. In this thesis I analyze 48 bone cement types using their content of 20 elements. My goal is to ?find a method to classify new found bone cement sample into these 48 categories. Here I will use multinomial logistic regression method to see whether it works or not. Due to the lack of observations, I generate enough data by adding white noise in proper scales to the original data again and again, and then I get a data set of over 100 times as many points as the original one. Then I use purposeful variable selection method to pick the covariates I need, rather than stepwise selection. There are 15 covariates left after the selection, and then I use my new data set to fit such a multinomial logistic regression model. The model doesn't perform that good in goodness of ?fit test, but the result is still acceptable, and the diagnostic statistics also indicate a good performance. Combined with clinical experience and prior conditions, this model is helpful in this classification case.

Early time series classification is the task of predicting the class label of a time series before it is observed in its entirety. In time-sensitive domains where information is collected over time it is worth sacrificing some classification accuracy in favor of earlier predictions, ideally early enough for actions to be taken. However, since accuracy and earliness are contradictory objectives, a solution to this problem must find a task-dependent trade-off. There are two common state-of-the-art methods. The first involves an analyst selecting a timestep at which all predictions must be made. This does not capture earliness on a case-by-case basis, so if the selecting timestep is too early, all later signals are missed, and if a signal happens early, the classifier still waits to generate a prediction. The second method is the exhaustive search for signals, which encodes no timing information and is not scalable to high dimensions or long time series. We design the first early classification model called EARLIEST to tackle this multi-objective optimization problem, jointly learning (1) to decide at which time step to halt and generate predictions and (2) how to classify the time series. Each of these is learned based on the task and data features. We achieve an analyst-controlled balance between the goals of earliness and accuracy by pairing a recurrent neural network that learns to classify time series as a supervised learning task with a stochastic controller network that learns a halting-policy as a reinforcement learning task. The halting-policy dictates sequential decisions, one per timestep, of whether or not to halt the recurrent neural network and classify the time series early. This pairing of networks optimizes a global objective function that incorporates both earliness and accuracy. We validate our method via critical clinical prediction tasks in the MIMIC III database from the Beth Israel Deaconess Medical Center along with another publicly available time series classification dataset. We show that EARLIEST out-performs two state-of-the-art LSTM-based early classification methods. Additionally, we dig deeper into our model's performance using a synthetic dataset which shows that EARLIEST learns to halt when it observes signals without having explicit access to signal locations. The contributions of this work are three-fold. First, our method is the first neural network-based solution to early classification of time series, bringing the recent successes of deep learning to this problem. Second, we present the first reinforcement-learning based solution to the unsupervised nature of early classification, learning the underlying distributions of signals without access to this information through trial and error. Third, we propose the first joint-optimization of earliness and accuracy, allowing learning of complex relationships between these contradictory goals.

Training a large dataset has become a serious issue for researchers because it requires large memories and can take a long time for computing. People are trying to process large scale dataset not only by changing programming model, such as using MapReduce and Hadoop, but also by designing new algorithms that can retain performance with less complexity and runtime. In this thesis, we present implementations of incremental learning and online learning methods to classify a large traffic light dataset for traffic light recognition. The introduction part includes the concepts and related works of incremental learning and online learning. The main algorithm is a modification of IMORL incremental learning model to enhance its performance over the learning process of our application. Then we briefly discuss how the traffic light recognition algorithm works and the problem we encounter during training. Rather than focusing on incremental learning, which uses batch to batch data during training procedure, we introduce Pegasos, an online style primal gradient-based support vector machine method. The performance of Pegasos for classification is extraordinary and the number of instances it uses for training is relatively small. Therefore, Pegasos is the recommended solution to the large dataset training problem.

Orientador: Ricardo Cardoso Benine / Banca: Flávio C. Theodoro de Lima / Banca: Fernando Jereep / Banca: Fernando Carvalho / Banca: William Ricardo Amancio Santana / Resumo: As espécies atualmente alocadas em Moenkhausia não formam um grupo monofilético. Estudos atuais buscam estabelecer novas relações e readequações destas espécies. No entanto, são verificados equívocos nas identificações que levam à relações não verdadeiras. O presente estudo reúne o conhecimento taxonômico sobre Moenkhausia, corrige, complementa e padroniza descrições das espécies. As espécies foram reunidas em grupos, com base em caracteres da morfologia externa e no conhecimento filogenético disponível para o gênero: Moenkhausia oligolepis - padrão de colorido reticulado e presença de banda escura no pedúnculo caudal; M. lepidura - mácula restrita ao lobo superior da nadadeira caudal; M. dichroura - mácula em ambos os lobos da nadadeira caudal; M. doceana - mácula umeral estendida horizontalmente ou duas máculas umerais, mácula no pedúnculo caudal ou nos raios medianos da nadadeira caudal; seis, ou mais, séries de escamas acima da linha lateral; M. xinguensis - maior altura do corpo na origem da nadadeira dorsal, mácula umeral ausente ou inconspícua e cinco series de escamas acima da linha lateral; M. eigenmanni - concentração de cromatóforos nas escamas, especialmente na região mediana dos flancos; corpo, geralmente, longo e caracteristicamente convexo na região pré-dorsal; M. ceros - baixa concentração de cromatóforos nas escamas, pequeno porte, faixa escura ou iridescente desde a região mediana do corpo ao pedúnculo caudal ou raios medianos da nadadeira caudal. Esses agrupamentos não são exclusivamente naturais e não devem ser tomados como relações filogenéticas. No entanto, são eficazes para a identificação das espécies e elaboração de futuras hipóteses de relações em Moenkhausia / Abstract: The species currently allocated in Moenkhausia are not a monophyletic group. Current studies try to establish new relationships and adequation these species. However, identifications mistakes are usualy verified. What show wrong relationships. This study gathers the taxonomic knowledge of Moenkhausia, fixe, standardizes and complements species descriptions. Most species were grouped into groups, based on characters from the external morphology and phylogenetic knowledge available to the genre: Moenkhausia oligolepis - reticulate patters, presence of dark colored band on the caudal peduncle; M. lepidura - dark spot restricted to the upper lobe of the caudal fin; M. dichroura - spot in both lobes of caudal fin; M. doceana - one horizontally extended humeral spot or two humeral spot, spot presence in the caudal peduncle or in the middle rays of the caudal fin, six or more series of scales above lateral line; M. xinguensis - deep body at origin of dorsal fin, absent or inconspicuous humeral spot, five series of scales above lateral line; M. eigenmanni - concentration of chromatophores on the scales, especially in the middle part of the flanks, usually long body and characteristically convex pre-dorsal region; M. ceros - low concentration of chromatophores on the scales, small, dark or iridescent line from the middle region of the body to the caudal peduncle or middle rays of the caudal fin. These groupings are not natural and should not be taken as phylogenetic relationships. However, they are effective for species identification and development of future hypotheses of relationships in Moenkhausia / Doutor