Yacht races are won by good sailors racing fast boats. A good skipper takes decisions at key moments of the race based on the anticipated wind behaviour and on his position on the racing area and with respect to the competitors. His aim is generally to complete the race before all his opponents, or, when this is not possible, to perform better than some of them. In the past two decades some methods have been proposed to compute optimal strategies for a yacht race. Those strategies are aimed at minimizing the expected time needed to complete the race and are based on the assumption that the faster a yacht, the higher the number of races that it will win (and opponents that it will defeat). In a match race, however, only two yachts are competing. A skipper’s aim is therefore to complete the race before his opponent rather than completing the race in the shortest possible time. This means that being on average faster may not necessarily mean winning the majority of races. This thesis sets out to investigate the possibility of computing a sailing strategy for a match race that can defeat an opponent who is following a fixed strategy that minimises the expected time of completion of the race. The proposed method includes two novel aspects in the strategy computation: A short-term wind forecast, based on an Artificial Neural Network (ANN) model, is performed in real time during the race using the wind measurements collected on board. Depending on the relative position with respect to the opponent, decisions with different levels of risk aversion are computed. The risk attitude is modeled using Coherent Risk Measures. The proposed algorithm is implemented in a computer program and is tested by simulating match races between identical boats following progressively refined strategies. Results presented in this thesis show how the intuitive idea of taking more risk when losing and having a conservative attitude when winning is confirmed in the risk model used. The performance of ANN for short-term wind forecasting is tested both on wind speed and wind direction. It is shown that for time steps of the order of seconds and adequate computational power ANN perform better than linear models (persistence models, ARMA) and other nonlinear models (Support Vector Machines). The outcome of the simulated races confirms that maximising the probability of winning a match race does not necessarily correspond to minimising the expected time needed to complete the race.

The aim of thesis is the recognition of the symbols in musical notation. Functions are implemented searching for a template in the image.

>Magister Scientiae - MSc / There are five fundamental parameters that characterize any sign language gesture. They are hand shape, orientation, motion and location, and facial expressions. The SASL group at the University of the Western Cape has created systems to recognize each of these parameters in an input video stream. Most of these systems make use of the Support Vector Machine technique for the classification of data due to its high accuracy. It is, however, unknown how other machine learning techniques compare to Support Vector Machines in the recognition of each of these parameters. This research lays the foundation for the process of determining optimum machine learning techniques for each parameter by comparing Support Vector Machines to Artificial Neural Networks and Random Forests in the context of South African Sign Language hand shape recognition. Li, a previous researcher at the SASL group, created a state-of-the-art hand shape recognition system that uses Support Vector Machines to classify hand shapes. This research re-implements Li’s feature extraction procedure but investigates the use of Artificial Neural Networks and Random Forests in the place of Support Vector Machines as a comparison. The machine learning techniques are optimized and trained to recognize ten SASL hand shapes and compared in terms of classification accuracy, training time, optimization time and classification time.

An automated detection process for Eukaryotic ribonucleic acid (RNA) Polymerase II Promoter is presented in this dissertation. We employ an artificial neural network (ANN) in conjunction with features that were selected using an information-theoretic approach. Firstly an introduction is given where the problem is described briefly. Some background is given about the biological and genetic principles involved in DNA, RNA and Promoter detection. The automation process is described with each step given in detail. This includes the data information gathering, feature generation, and the full ANN process. The ANN section of the project is split up in a generation process, a training section as well as a testing section. Lastly the final detection program was tested and compared to other promoter detection systems. An improvement of at least 10% in positive prediction value (PPV) in comparison with current state-of-the-art solutions was obtained. Note: A Companion CD should accompany this report that contains all the program code and some of the source data that was used in this project. All the references to “Companion CD”, reference number [18] are references to these programs.acquisition process, how the different samples were split into different sets and statistical. / Dissertation (MEng (Computer Engineering))--University of Pretoria, 2007. / Electrical, Electronic and Computer Engineering / unrestricted

A polymorphic substance is capable of forming a number of different crystalline phases that are referred to as its polymorphs. The critical process that determines the outcome of a crystallization process in a polymorphic system is thought to be the nucleation state, which is the self-assembled stage just prior to the formation of crystals with long-range order. While nucleation is well known to be influenced by macroscopically measurable parameters such as temperature, supersaturation and solvent choice our understanding of the underlying molecular self-assembly processes is very limited. The research described in this thesis explores a new approach to extending our knowledge in this area by the use of a combination of medium throughput crystallisation experiments together with the computation of a range of molecular and solute/solvent descriptors of the system under study.The main objective of the work was to develop a protocol for relating experimental and computational data via artificial neural network (ANN) analysis, to identify significant links between experimental polymorphic outcomes and molecular properties. By creating a model that can predict the polymorphic form in a given experiment it is anticipated that our understanding of links between nucleation and crystallisation will be enhanced through the determining the pivotal properties of a molecule that cause it to form one polymorph over another. The ANN method was developed in the context of the carbamazepine system, applying several statistical techniques to the results of 88 crystallisation experiments, featuring 13 solvents, 3 evaporation rates and 4 temperatures. The results show that this approach allows the formulation of further research hypotheses through examination of the physical meaning of the set of descriptors identified by the ANN approach. Crucially, principal component analysis (PCA) was found to be able to efficiently narrow down large sets of computationally derived descriptors to a manageable set by removing redundancy through strongly cross-correlated parameters. The best ANN model generated in this research was capable of predicting the major polymorphic form in 89 % of cross-validation experiments.The optimised set of descriptors included both solute and solvent properties, which predominantly described the intermolecular interactions in solution. The physical meanings of the descriptors and their impact on the molecular processes during nucleation has been considered and their cross correlation has been examined. Initial results from further experimentation with the tolbutamide and ROY systems indicate that the methodology is also transferable to other polymorphic systems.

GIS has been used in Veterinary Science for a couple of year and the application thereof has been growing rapidly. A number of GIS models have been developed to predict the occurrences of certain types of insect species including the Culicoides species (spp), the insect vectors responsible for the transmission of the African horse sickness (AHS) virus. AHS is endemic to sub-Saharan Africa and is carried by two midges called Culicoides Imicola and Culicoides Bolitinos. The disease causes severe illness in horses and has significant economic impact if not dealt with timeously. Although these models had some success in the prediction of possible abundance of the Culicoides spp. the complicated nature and high number of variables influencing the abundance of Culicoides spp. posed some challenges to these GIS models. This informs the need for models that can accurately predict potential abundance of Culicoides spp to prevent unnecessary horse deaths. This lead the study to the use of a combination of a GIS and an artificial neural networks (ANN) to develop a model that can predict the abundance of C. Imicola and C. Bolitinos. ANNs are models designed to imitate the human brain and have the ability to learn through examples. ANNs can therefore model extremely complex features. In addition, using GIS maps to visualise the predictions will make the models more accessible to a wider range of practitioners. / Dissertation (MSc)--University of Pretoria, 2010. / Geography, Geoinformatics and Meteorology / MSc / Unrestricted

The aim of this work is to quantify the concentration of chromic acid (CA) in a saturated solution of chromium trioxide and sodium dichromate using Artificial Neural Networks (ANNs). A set of titration curves was obtained by automated acid-base titration according to a factorial experimental design that was developed for this purpose. These titration curves were divided into three subsets, a learning, training and test set for use by ANNs. Once trained, ANNs have the ability to recognize, generalize and relate the input to a particular output. Concentration of chromic acid (CA), total chromium(VI) and/or dichromate was used as the outputs and titration curves as the inputs to ANNs. Our aim here was to establish whether ANNs would be able to predict the concentration of chromic acid with an absolute error below 1%. For real world problem, the neural networks are only given the inputs and are expected to produce reasonable outputs corresponding to that inputs without any prior ‘knowledge’ about theory involved – here, no interpretation of titration curves was performed by ANNs. The test set of data that was not used for learning process, was used to validate the performance of the neural networks, to verify whether the ANNs learned the input-output patterns properly and how well trained ANNs were able to predict the concentrations of chromic acid, dichromate and total chromium. A number of ANNs models have been considered by varying the number of neurons in the hidden layer and parameters related to the learning process. It has been shown that ANNs can predict the concentration of chromic acid with required accuracy. A number of factors that affect the performance of the neural networks, such as the number of points in a titration curve, number of test points and their distribution within the training set, has been investigated. This work demonstrates that ANNs can be used for online monitoring of an electrolytic industrial process to manufacture chromic acid. / Dissertation (MSc)--University of Pretoria, 2009. / Chemistry / unrestricted

High blood pressure is a major public health issue. However, there are many physical and non-physical factors that affect the measurement of blood pressure (BP) over very short time spans. Therefore, it is very difficult to write a mathematical equation which includes all relevant factors needed to estimate accurate BP values. As a result, a possible solution to overcome these limitations is the use of an artificial neural network (ANN). The aim of this research is to design and implement a new ANN approach, which correlates the arterial pulse waveform shape to BP values, for estimation of BP in a single heartbeat. To test the feasibility of this approach, a pilot study was performed on an arterial pulse waveform dataset obtained from 11 patients with normal BP and 11 patients with hypertension. It was found that the proposed method can accurately estimate BP in single heartbeats and satisfy the requirements of the ANSI/AAMI standard for non-invasive measurement of BP.

Some of the largest United States bankruptcies of publicly-traded non-financial firms have occurred within the last decade. The continuing need to improve bankruptcy prediction has generated numerous research studies utilizing various prediction models. The purpose of this study is to test the usefulness of the multiple discriminant, probit, and artificial neural network (ANN) models in predicting bankruptcy in the United States textile-related industry. Financial data is examined for 47 bankrupt and 104 non-bankrupt publicly-traded firms in the textile-related industry during the time period 1998-2004, which includes the events of the Asian currency crisis and increased competition from China. Models developed by Altman (1968), Altman (1983), Zmijewski (1984) are compared to ANNs based upon each of these models. A comparison to an ANN including all of the ratios of the previous models and variables for firm size and domestic sales is also made. The Altman (1968) model and ANN 68 model are found to have the higher predictive power for one and two years prior to bankruptcy, respectively, for bankrupt firms. The ANN 84 model and the ANN 83 model have the highest correct classification results for nonbankrupt firms for the entire time period. Solvency and leverage variables appear to have the most impact on the bankruptcy prediction of textile-related firms. The additional variables of firm size and domestic sales are not found to improve the predictive accuracy. This study supports the continued use of the original Altman (1968) model for predicting bankruptcy in a manufacturing industry. Simultaneous utilization of the ANN 83 model to predict nonbankrupt firms is also suggested since the majority of the Altman (1968) variables can be used and the higher potential for improved predictability. This study may be extended to years after 2004 with consideration given to quarterly information, NAICs codes, and leverage variable alternatives.

This dissertation established a state-of-the-art programming tool for designing and training artificial neural networks (ANNs) and showed its applicability to brain research. The developed tool, called NeuralStudio, allows users without programming skills to conduct studies based on ANNs in a powerful and very user friendly interface. A series of unique features has been implemented in NeuralStudio, such as ROC analysis, cross-validation, network averaging, topology optimization, and optimization of the activation function’s slopes. It also included a Support Vector Machines module for comparison purposes. Once the tool was fully developed, it was applied to two studies in brain research. In the first study, the goal was to create and train an ANN to detect epileptic seizures from subdural EEG. This analysis involved extracting features from the spectral power in the gamma frequencies. In the second application, a unique method was devised to link EEG recordings to epileptic and non-epileptic subjects. The contribution of this method consisted of developing a descriptor matrix that can be used to represent any EEG file regarding its duration and the number of electrodes. The first study showed that the inter-electrode mean of the spectral power in the gamma frequencies and its duration above a specific threshold performs better than the other frequencies in seizure detection, exhibiting an accuracy of 95.90%, a sensitivity of 92.59%, and a specificity of 96.84%. The second study yielded that Hjorth’s parameter activity is sufficient to accurately relate EEG to epileptic and non-epileptic subjects. After testing, accuracy, sensitivity and specificity of the classifier were all above 0.9667. Statistical tests measured the superiority of activity at over 99.99 % certainty. It was demonstrated that 1) the spectral power in the gamma frequencies is highly effective in locating seizures from EEG and 2) activity can be used to link EEG recordings to epileptic and non-epileptic subjects. These two studies required high computational load and could be addressed thanks to NeuralStudio. From a medical perspective, both methods proved the merits of NeuralStudio in brain research applications. For its outstanding features, NeuralStudio has been recently awarded a patent (US patent No. 7502763).