Thesis (MComm (Statistics and Actuarial Science))--University of Stellenbosch, 2009. / Incidence can be defined as the rate of occurence of new infections of a disease like HIV and is an useful estimate of trends in the epidemic. Annualised incidence can be expressed as a proportion, namely the number of recent infections per year divided by the number of people at risk of infection. This number of recent infections is dependent on the window period, which is basically the period of time from seroconversion to being classified as a long-term infection for the first time. The BED capture enzyme immunoassay was developed to provide a way to distinguish between recent and long-term infections. An optical density (OD) measurement is obtained from this assay. Window period is defined as the number of days since seroconversion, with a baseline OD value of 0, 0476 to the number of days to reach an optical density of 0, 8.The aim of this study is to describe different techniques to estimate the window period which may subsequently lead to alternative estimates of annualised incidence of HIV infection. These various techniques are applied to different subsets of the Zimbabwe Vitamin A for Mothers and Babies (ZVITAMBO) dataset. Three different approaches are described to analyse window periods: a non-parametric survival analysis approach, the fitting of a general linear mixed model in a longitudinal data setting and a Bayesian approach of assigning probability distributions to the parameters of interest. These techniques are applied to different subsets and transformations of the data and the estimated mean and median window periods are obtained and utilised in the calculation of incidence.

In naturally ventilated buildings, occupants play a key role in the performance and energy efficiency of the building operation, mainly through the opening and closing of windows. To include the effects of building occupants within building performance simulation, several useful models describing building occupants and their window opening/closing behaviour have been generated in the past 20 years. However, in these models, the occupants are classified based on the whole population or on sub-groups within a building, whilst the behavioural difference between individuals is commonly ignored. This research project addresses this latter issue by evaluating the importance of the modelling and prediction of occupants window behaviour individually, rather than putting them into a larger population group. The analysis is based on field-measured data collected from a case study building containing a number of single-occupied cellular offices. The study focuses on the final position of windows at the end of the working day. In the survey, 36 offices and their occupants were monitored, with respect to the occupants presence and window use behaviour, in three main periods of a year: summer, winter and transitional. From the behaviour analysis, several non-environmental factors, namely, season, floor level, gender and personal preference, are identified to have a statistically significant effect on the end-of-day window position in the building examined. Using these factors, occupants window behaviour is modelled by three different classification methods of building occupants, namely, whole population, sub-groups and personal preference. The preference-based model is found to perform much better predictive ability on window state when compared with those developed based on whole population and sub-groups. When used in a realistic building simulation problem, the preference-based prediction of window behaviour can reflect well the different energy performance among individual rooms, caused by different window use patterns. This cannot be demonstrated by the other two models. The findings from this research project will help both building designers and building managers to obtain a more accurate prediction of building performance and a better understanding of what is happening in actual buildings. Additionally, if the habits and behavioural preferences of occupants are well understood, this knowledge can be potentially used to increase the efficiency of building operation, by either relocating occupants within the building or by educating them to be more energy efficient.

With the continued progression of Moore’s law into the sub-14nm technology nodes, interconnect RC and power dissipation scaling play an increasingly important role in overall product performance. As critical dimensions in the mainstream Cu/ULK interconnect system shrink below 30nm, corresponding increases in relative process variation and decreases in overall process window mandate increasingly complex integrated solutions. Traditional metallization processes, e.g. PVD barrier and seed layers, no longer scale for all layout configurations as they reach physical and geometric limitations. Interactions between design, OPC, and patterning also play more and more critical roles with respect to reliability and yield in volume manufacturing; stated simply, scaling is no longer “business as usual”. Restricted design layouts, prescriptive design rules, novel materials, and holistic integration solutions each therefore become necessary to maximize available process windows, thus enabling new generations of cost-competitive products in the marketplace.

published_or_final_version / Computer Science and Information Systems / Master / Master of Philosophy

Transmission Control Protocol (TCP) designed to deliver seamless and reliable end-to-end data transfer across unreliable networks works impeccably well in wired environment. In fact, TCP carries the around 90% of Internet traffic, so performance of Internet is largely based on the performance of TCP. However, end-to-end throughput in TCP degrades notably when operated in wireless networks. In wireless networks, due to high bit error rate and changing level of congestion, retransmission timeouts for packets lost in transmission is unavoidable. TCP misinterprets these random packet losses, due to the unpredictable nature of wireless environment, and the subsequent packet reordering as congestion and invokes congestion control by triggering fast retransmission and fast recovery, leading to underutilization of the network resources and affecting TCP performance critically. This thesis reviews existing approaches, details two proposed systems for better handling in networks with random loss and delay. Evaluation of the proposed systems is conducted using OPNET simulator by comparing against standard TCP variants and with varying number of hops.

In this thesis we are concerned with the existence of infrared fixed points and the conformal window for gauge theories with fermions. We are particularly interested in those theories that are candidates for walking technicolor. We discuss the background of technicolor and the techniques relevant to a theoretical understanding of the conformal window. Following this we extend the ideas of metric confinement and causal analyticity to theories with fermions in non-fundamental representations. We use these techniques to, respectively, provide a lower bound on the lower end of the conformal window and to provide a measure of perturbativity. As well as analytic calculations we use lattice techniques to investigate two particular candidate theories for walking technicolor - SU(2) with two adjoint fermions and with six fundamental fermions. We use Schrodinger Functional techniques to investigate the running of the theory across a wide range of scales. We measure both the running of the coupling and an estimator for the fermion mass anomalous dimension, $gamma$. We find that both theories are consistent with an infrared fixed-point. However, paying particular attention to our error estimates, we are unable to absolutely confirm their existence. This is a not unexpected result for SU(2) with two adjoint fermions but is rather surprising for SU(2) with only six fundamental fermions. In the region where we are consistent with a fixed point we find $0.05<gamma<0.56$ for $SU(2)$ with two adjoint fermions and $0.135<gamma<1.03$ for $SU(2)$ with six fundamental fermions. The measurement of $gamma$ for $SU(2)$ with two adjoint fermions is the first determination of $gamma$ for any candidate theory of walking technicolor.

Ad hoc networks represent the current de-facto alternative for infrastructure-less environments, due to their self-configuring and resilience characteristics. Ad hoc networks flexibility benefits, such as unrestrained computing, lack of centralisation, and ease of deployment at low costs, are tightly bound with relevant deficiencies such as limited resources and management difficulty. Ad hoc networks witnessed high attention from the research community due to the numerous challenges faced when deploying such a technology in real scenarios. Starting with the nature of the wireless environment, which raises significant transmission issues when compared with the wired counterpart, ad hoc networks require a different approach when addressing the data link problems. Further, the high packet loss due to wireless contention, independent of network congestion, requires a different approach when considering quality of service degradation and unfair channel resources distribution among competing flows. Although these issues have already been considered to some extent by researchers, there is still room to improve quality of service by reducing the effect of packet loss and fairly distributing the medium access among competing nodes. The aim of this thesis is to propose a set of mechanisms to alleviate the effect of packet loss and to improve fairness in ad hoc networks. A transport layer algorithm has been proposed to overcome the effects of hidden node collisions and to reduce the impact of wireless link contention by estimating the four hop delay and pacing packet transmissions accordingly. Furthermore, certain topologies have been identified, in which the standard IEEE 802.11 faces degradation in channel utilisation and unfair bandwidth allocation. Three link layer mechanisms have been proposed to tackle the challenges the IEEE 802.11 faces in the identified scenarios to impose fairness in ad hoc networks through fairly distributing channel resources between competing nodes. These mechanisms are based on monitoring the collision rate and penalising the greedy nodes where no competing nodes can be detected but interference exists, monitoring traffic at source nodes to police access to the channel where only source nodes are within transmission range of each other, and using MAC layer acknowledgements to flag unfair bandwidth allocation in topologies where only the receivers are within transmission range of each other. The proposed mechanisms have been integrated into a framework designed to adapt and to dynamically select which mechanism to adopt, depending on the network topology. It is important to note that the proposed mechanisms and framework are not alternatives to the standard MAC protocol but are an enhancement and are triggered by the failure of the IEEE 802.11 protocol to distribute the channel resources fairly. All the proposed mechanisms have been validated through simulations and the results obtained from the experiments show that the proposed schemes fairly distribute channel resources fairly and outperform the performance of the IEEE 802.11 protocol in terms of channel utilisation as well as fairness.

The objective of this study was to analyze the visual communication of fashion brand identity through the concept of window displays. Window displays are often categorized as an aspect of a much larger concept of visual merchandising that should be seen as a crucial part of the entire marketing communications. Although the concept of window displays is not new, their effect on the overall brand image creation remains relatively unexplored. The interpretation of communication in a visual form is complex in nature, and this research approaches the matter from a semiotic perspective. The approach was selected due to its capability to explore deeper meanings behind visual objects. In this study, window display is understood as a crucial part of the sign system of a fashion store. Window display is commonly regarded as the starting point of the entire shopping experience, and therefore in the lack of previous research, it is a logical starting point for this study. The analytical framework of the study follows the thee metafunctions of semiotic modes as discussed in the Theory of the Grammar of Visual Design by Kress and van Leeuwen (2006). The analytical framework is then applied to analyze six different window displays produced by the same fashion brand. Findings of the research are discussed in terms of the three metafunctions of ideational, interpersonal and textual, and the purpose is to increase the knowledge of how window displays can be used to depict different relations between brand, people, places and things. In terms of brand identity communication, the most important contribution is to understand the complex relationship that is formed between the window (the brand) and its viewer (the customer). As a result, this research finds that the window displays are loaded with multiple visual cues all of which intend to serve the same purpose of brand identity communication. Clothing and accessories are the most salient elements used and they can be understood as the embodiment of the brand identity. Images of models and mannequins are used to address the viewers and create a specific relationship with them. The less salient elements in the window displays are used to support the message the brand is conveying through the display.

Die ständig steigende Rechenleistung von PCs und Workstations ermöglicht es, die spektralen Varianzen direkt durch Fouriertransformation einer Zeitreihe zu schätzen. Dabei ist ein geeignetes Fenster auszuwählen. Die Nichtbeachtung dieser Tatsache führt zwangsläufig zum Rechteckfenster und zum Verschmieren der Information über den ganzen Spektralbereich. Von den hier untersuchten Fenstern hat das Hanning-Fenster die besten Eigenschaften mit dem stärksten Abfall im benachbarten Frequenzbereich. Analog der Fensterung von Auto- und Kreuzkorrelationsfunktion im Frequenzraum wird das Hanning-Fenster für die Fouriertransformation einer Meßreihe im Frequenzraum angegeben. / Due to the permanent grow in PC and workstation power it is possible to estimate the spectral variances directly by the Fourier transform of a time series. Thereby one has to choose an appropriate data window. Ignoring this leads to an unsuited square window function where the spectral power in one frequency bin contains leakage from frequency components that are bins away. From the window functions we tested here, the Hanning window has best properties with minimum leakage. Analogous to windowing of the auto- and crosscorrelation function in the frequency domain we give also the Hanning window for the Fourier transform of time series in the frequency domain.

Now a days, Closed Circuit Television (CCTV) cameras are installedeverywhere in public places to monitor illegal activities like armedrobberies. Mostly CCTV footages are used as post evidence after theoccurrence of crime. In many cases a person might be monitoringthe scene from CCTV but the attention can easily drift on prolongedobservation. Eciency of CCTV surveillance can be improved by in-corporation of image processing and object detection algorithms intomonitoring process.The object detection algorithms, previously implemented in CCTVvideo analysis detect pedestrians, animals and vehicles. These algo-rithms can be extended further to detect a person holding weaponslike rearms or sharp objects like knives in public or restricted places.In this work the detection of weapon from CCTV frame is acquiredby using Histogram of Oriented Gradients (HOG) as feature vector andarticial neural networks performing back-propagation algorithm forclassication.As a weapon in the hands of a human is considered to be greaterthreat as compared to a weapon alone, in this work the detection ofhuman in an image prior to a weapon detection has been found advan-tageous. Weapon detection has been performed using three methods.In the rst method, the weapon in the image is detected directly with-out human detection. Second and third methods use HOG and back-ground subtraction methods for detection of human prior to detectionof a weapon. A knife and a gun are considered as weapons of inter-est in this work. The performance of the proposed detection methodswas analysed on test image dataset containing knives, guns and im-ages without weapon. The accuracy rate 84:6% has been achievedby a single-class classier for knife detection. A gun and a knife havebeen detected by the three-class classier with an accuracy rate 83:0%.