Thesis (M. Phil.)--University of Hong Kong, 2005. / Title proper from title frame. Also available in printed format.

Thesis (doctoral)--Ludwig Maximilians-Universität zu München, 1979.

Thesis (M.S.)--University of Wisconsin--Madison, 1965. / eContent provider-neutral record in process. Description based on print version record. Bibliography: l. 80.

Thesis--University of Wisconsin--Madison. / Typescript. Vita. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 159-164).

Thesis (MScIng)--University of Stellenbosch, 2007. / Bibliography. Also available via the Internet.

<p> When reviewing the performance of Intelligent Virtual Assistants (IVAs), it is desirable to prioritize conversations involving misunderstood human inputs. These conversations uncover error in natural language understanding and help prioritize and expedite improvements to the IVA. As human reviewer time is valuable and manual analysis is time consuming, prioritizing the conversations where misunderstanding has likely occurred reduces costs and speeds improvement. A system for measuring the posthoc <i>risk of missed intent </i> associated with a single human input is presented. Numerous indicators of risk are explored and implemented. These indicators are combined using various means and evaluated on real world data. In addition, the ability for the system to adapt to different domains of language is explored. Finally, the system performance in identifying errors in IVA understanding is compared to that of human reviewers and multiple aspects of system deployment for commercial use are discussed.</p><p>

<p> As the complexity of mankind&rsquo;s day-to-day challenges increase, so does a need for the optimization of know solutions to accommodate for this increase in complexity. Today&rsquo;s computer systems use the Input, Processing, and Output (IPO) model as a way to deliver efficiency and optimization in human activities. Since the relative quality of an output utility derived from an IPO based computer system is closely coupled to the quality of its input media, the measure of the Optimal Quotient (OQ) is the ratio of the input to output which is 1:1. This relationship ensures that all IPO based computers are not just linearly predictable, but also characterized by the Garbage In Garbage Out (GIGO) design concept. While current IPO based computer systems have been relatively successful at delivering some measure of optimization, there is a need to examine (Li &amp; Malik, 2016) alternative methods of achieving optimization. The purpose of this quantitative research study, through an experimental research design, is to determine the effects of the application of a Swarm Intelligence algorithm on the efficiency of the Cosmos Open Source Managed Operating System. </p><p> By incorporating swarm intelligence into an improved IPO design, this research addresses the need for optimization in computer systems through the creation of an improved operating system Scheduler. The design of a Swarm Intelligence Operating System (SIOS) is an attempt to solve some inherent vulnerabilities and problems of complexity and optimization otherwise unresolved in the design of conventional operating systems. This research will use the Cosmos open source operating system as a test harness to ensure improved internal validity while the subsequent measurement between the conventional and improved IPO designs will demonstrate external validity to real world applications. </p><p>

Orientador: Heloisa Vieira da Rocha / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-11T23:33:28Z (GMT). No. of bitstreams: 1 CerqueiraFilho_ReginaldodeCastro_M.pdf: 6466959 bytes, checksum: 3f7e88d58d46232b960dc6532eb285ef (MD5) Previous issue date: 1999 / Resumo: Não informado. / Abstract: Not informed. / Mestrado / Mestre em Ciência da Computação

Submitted by Marcele Costal de Castro (costalcastro@gmail.com) on 2017-09-12T16:04:59Z No. of bitstreams: 1 MAURO VITOR DE OLIVEIRA M.PDF: 4253155 bytes, checksum: c6bda172032d3e3627f6cf0a0aff3ec0 (MD5) / Made available in DSpace on 2017-09-12T16:04:59Z (GMT). No. of bitstreams: 1 MAURO VITOR DE OLIVEIRA M.PDF: 4253155 bytes, checksum: c6bda172032d3e3627f6cf0a0aff3ec0 (MD5) Previous issue date: 1999-06 / Este trabalho desenvolve dois modelos de validação de sinal que utilizam fedes neuronais para realizar a redundância analítica de sinais monitorados em uma instalação industrial. Um dos modelos utiliza uma única rede neuronal para efetuar a redundância analítica dos sinais e o outro modelo utiliza várias redes neuronais para efetuar esta redundância, cada uma trabalhando em uma região de operação específica da instalação. Para efetuar a divisão de todos os pontos de operação da instalação em várias regiões específicas foram estudados quatro métodos de separação em classes. Adicionalmente os modelos fornecem uma indicação da confiança das suas respostas através de conceitos baseados em lógica nebulosa. Os modelos foram implementados em linguagem C e foram ensaiados com sinais monitorados do reator nuclear Angra I, durante uma de suas partidas até atingir 100% de potência. / This work develops two models of signal validation in which the analytical redundancy of the monitored signals from an industrial plant is made by neural networks. In one model the analytical redundancy is made by only one neural network while in the other it is done by several neural networks, each one working in a specific part of the entire operation region of the plant. Four cluster techniques were tested to separate the entire region of operation in several specific regions. An additional information of systems' reliability is supplied by a fuzzy inference system. The models were implemented in C language and tested with signals acquired from Angra I nuclear power plant, from its start to 100% of power.

Submitted by Marcele Costal de Castro (costalcastro@gmail.com) on 2017-09-12T16:25:20Z No. of bitstreams: 1 MAURO VITOR DE OLIVEIRA D.PDF: 3909160 bytes, checksum: 3b128386b6cec06b4642bf16c9686322 (MD5) / Made available in DSpace on 2017-09-12T16:25:20Z (GMT). No. of bitstreams: 1 MAURO VITOR DE OLIVEIRA D.PDF: 3909160 bytes, checksum: 3b128386b6cec06b4642bf16c9686322 (MD5) Previous issue date: 2005-10 / Este trabalho desenvolve uma metodologia para validação de sinal através de modelos empíricos de, com técnicas de inteligência artificial, para realizar a redundância analítica de sinais monitorados em uma planta industrial. A metodologia proposta reúne de forma sistêmica todas as etapas que devem compor um modelo de validação de sinal. Além disso, baseado na metodologia proposta é implementado um modelo de validação de sinal. Para validar a metodologia, são apresentados, segundo a metodologia, diversos trabalhos realizados em validação de sinal. Além disso, baseado na metodologia proposta é implementado um modelo de validação de sinal. A seleção das etapas de composição deste modelo é feita com base nos resultados obtidos nos ensaios obtidos dos ensaios realizados em diferentes técnicas de implementação de cada etapa da metodologia. Ao final do trabalho, o modelo de validação de sinal implementado é ensaiado com sinais monitorados da planta de potência nuclear Angra I, durante uma de suas partidas até atingir 100% de potência. Foram selecionados para validação os sinais da potência nuclear e da temperatura da água na perna quente do circuito 1 do primário do reator. Os resultados obtidos mostraram que o modelo proposto apresenta desempenho eficiente na validação de sinal.