Thesis (Ph.D.) - University of Glasgow, 2007. / Ph.D. thesis submitted to the Department of Computing Science, Faculty of Information and Mathematical Sciences, University of Glasgow, 2007. Includes bibliographical references. Print version also available.

Thesis (Ph. D.)--State University of New York at Binghamton, Thomas J. Watson School of Engineering and Applied Science, Department of Computer Science, 2009. / Includes bibliographical references.

Thesis (MSc)--University of Stellenbosch, 2002. / ENGLISH ABSTRACT: In this thesis, we present a technique for the automatic detection of image orientation using Support Vector Machines (SVMs). SVMs are able to handle feature spaces of high dimension and automatically choose the most discriminative features for classification. We investigate the use of various kernels, including heavy tailed RBF kernels. We compare the classification performance of SVMs with the performance of multilayer perceptrons and a Bayesian classifier. Our results show that SVMs out perform both of these methods in the classification of individual images. We also implement an application for the classification of film rolls in a photographic workflow environment with 100% classification accuracy. / AFRIKAANSE OPSOMMING: In hierdie tesis, gebruik ons 'n tegniek vir die automatiese klassifisering van beeldoriëntasie deur middel van Support Vector Machines (SVM's). SVM's kan kenmerkruimtes van 'n hoë dimensie hanteer en kan automaties die mees belangrike kenmerke vir klassifikasie kies. Ons vors die gebruik van verskeie kerne, insluitende RBF-kerne, na. Ons vergelyk die klassifiseringsresultate van SVM's met die van multilaagperseptrone en 'n Bayes-klassifiseerder. Ons bewys dat SVM's beter resultate gee as beide van hierdie metodes vir die klassifikasie van individuele beelde. Ons implementeer ook a toepassing vir die klassifisering van rolle film in a fotografiese werkvloei-omgewing met 100% klassifikasie akuraatheid.

Vector processors conventionally have been used as an attached processor to a host computer. Within this limited scope, the application programmer must use the host computer in order to benefit from the vector processor resources. By using a local area network, programmers are no longer constrained to a specific host computer. The vector processor may be shared as a network resource. The vector processor service is developed within a distributed environment and, therefore, must address concerns pertinent to distributed system architecture. These issues include implementation methodologies, interprocess communication performance, protocol processing and network throughput, and the level of transparency of the implementations. This thesis presents models that facilitate implementation of the vector processor service over a local area network (LAN). The research investigates the performance of different interprocess communication techniques and alternative transport protocols, specifically TCP and UDP. Additionally, two LAN technologies are examined for the vector processor service, namely Ethernet and Fiber Data Distributed Interface (FDDI). Experiments are performed for a batch application using an Ethernet local area network. Simulations are done for a real-time application utilizing an Ethernet LAN as well as for the same batch application. Additionally, a FDDI local area network is simulated for a real-time application. Results indicate that the model based on remote program execution has better performance because of lower network communication overhead compared to the model based on remote procedure calls. However, the remote procedure call model provides for a more transparent implementation of the vector processor service. This thesis also discusses methods to improve the performance of the vector processor service, including better implementations and transport protocols, alternative remote procedure call protocols, and new multiprocessor architectures. / Master of Science

A dissertation submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in partial fulfillment of the requirements for the degree of Master of Science. Pretoria 1991. / Remote sensing can be used to produce maps of land-cover, but to be of use to the GIS community these maps must first be vectorized in an intelligent manner. Existing algorithms suffer from the defects of being slow, memory intensive and producing vast quantities of very short vectors. Furthermore if these vectors are thinned via standard algorithms, errors are introduced. The process of vectorizing raster maps is subject to major ambiguities. Thus an infinite family of vector maps ccrresponds to each raster map. This dissertation presents an algorithm for converting raster maps in a rapid manner to accurate vector maps with a minimum of vectors. The algorithm converts raster maps to vector maps using local information only, (a two by two neighbourhood). the method is "exact" in the sense that rasterizing the resulting polygons would produce exactly the same raster map, pixel for pixel. The method is "near optimal" in that it produces, in a local sense, that "exacb" vector map having the least number of vectors. The program is built around a home-grown object oriented Programming System (OOPS) for the C programming language. The main features of the OOPS system, (called OopCdaisy), are virtual and static methods, polymorphism, generalized containers, container indices and thorough error checking, The following general purpose objects are implemented with a large number of sophistiated methods :- Stacks, LIFO lists, scannable containers with indices, trees and 2D objects like points, lines etc. / AC2017

Abstract not available

Thesis (M.Eng.(Microelectronic Engineering))--University of Pretoria, 2009. / Includes summaries in Afrikaans and English. Includes bibliographical references (leaves [74]-78). Mode of access: World Wide Web.

The parallelization and vectorization of Monte Carlo algorithms for modelling charge transport in semiconductor devices are considered. The standard ensemble Monte Carlo simulation of a three parabolic band model for GaAs is first presented as partial verification of the simulation. The model includes scattering due to acoustic, polar-optical and intervalley phonons. This ensemble simulation is extended to a full device simulation by the addition of real-space positions, and solution for the electrostatic potential from the charge density distribution using Poisson's equation. Poisson's equation was solved using the cloud-in-cell scheme for charge assignment, finite differences for spatial discretization, and simultaneous over-relaxation for solution. The particle movement (acceleration and scattering) and the solution of Poisson's are both separately parallelized. The parallelization techniques used in both parts are based on the use of semaphores for the protection of shared resources and processor synchronization. The speed increase results for parallelization with and without vectorization on the Ardent Titan II are presented. The results show saturation due to memory access limitations at a speed increase of approximately 3.3 times the serial case when four processors are used. Vectorization alone provides a speed increase of approximately 1.6 times when compared with the nonvectorized serial case. It is concluded that the speed increase achieved with the Titan II is limited by memory access considerations and that this limitation is likely to plague shared memory machines for the forseeable future. For the program presented here, vectorization is concluded to provide a better speed increase per day of development time than parallelization. However, when vectorization is used in conjunction with parallelization, the speed increase due to vectorization is negligible. / Graduation date: 1993

Orientador: Maria Cristina Dias Tavares / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e Computação / Made available in DSpace on 2018-08-07T13:37:43Z (GMT). No. of bitstreams: 1 Paz_MarcosdeAraujo_D.pdf: 6127820 bytes, checksum: 21f6dc485bcf097846769f086c2362ad (MD5) Previous issue date: 2005 / Resumo: O objetivo deste trabalho é contribuir para o estudo de linhas de transmissão e o desenvolvimento de modelos de linhas voltados a programas do tipo EMTP. O trabalho desenvolvido consiste em aproveitar as propriedades e semelhanças dos modos não homopolares de propagação das linhas de transmissão, sejam elas idealmente transpostas ou que apresentem plano de simetria vertical. Por meio de manipulações da matriz de transformação fase-modo-fase através do uso de vetores ortogonais e elementos complexos no domínio modal, pode-se reduzir a dimensão dos vetores de estado. O procedimento proposto é direcionado às modelagens que utilizam transformação fase-modo-fase na determinação das variáveis de estado (corrente e tensão) nas linhas de transmissão. Os recursos usados levam a uma redução tanto em número de operações por iteração quanto em alocação de estados, otimizando o procedimento de cá1culo e aumentando a velocidade de simulação no processamento. A economia computacional que o procedimento proposto proporciona pode ser usada em ferramentas de simulação em tempo real / Abstract: The research objective is to contribute with the study of the transmission line and the development of line models in EMTP-type programs. The proposed model takes advantage of some similarities between non homopolar modes, applying orthogonal vectors and complex number theory to manipulate the state vectors in modal domain reducing the state vectors dimension. The proposed procedure is aimed to the modelling that use phase-mode-phase transformation in the determination of state variables (current and voltage) in transmission lines. The resources used lead to a reduction both in the number of operations per iteration and state allocation, optimizing the procedure and increasing the processing simulation speed. The computational economy, which the proposed procedure provide, can be applied to real-time simulation tools / Doutorado / Energia Eletrica / Doutor em Engenharia Elétrica