The statistical properties of training, validation and test data play an important role in assuring optimal performance in artificial neural networks (ANN). Re-searchers have proposed randomized data partitioning (RDP) and stratified data partitioning (SDP) methods for partition of input data into training, vali-dation and test datasets. RDP methods based on genetic algorithm (GA) are computationally expensive as the random search space can be in the power of twenty or more for an average sized dataset. For SDP methods, clustering al-gorithms such as self organizing map (SOM) and fuzzy clustering (FC) are used to form strata. It is assumed that data points in any individual stratum are in close statistical agreement. Reported clustering algorithms are designed to form natural clusters. In the case of large multivariate datasets, some of these natural clusters can be big enough such that the furthest data vectors are statis-tically far away from the mean. Further, these algorithms are computationally expensive as well. Here a custom design clustering algorithm (CDCA) has been proposed to overcome these shortcomings. Comparisons have been made using three benchmark case studies, one each from classification, function ap-proximation and prediction domain respectively. The proposed CDCA data partitioning method was evaluated in comparison with SOM, FC and GA based data partitioning methods. It was found that the CDCA data partitioning method not only performed well but also reduced the average CPU time. / Engineering Management

Footings are often founded on multi-layered soil profiles. Real soil profiles are often multi-layered with material constantly varying with depth, which affects the footing response significantly. Furthermore, the properties of the soil are known to vary with location. The spatial variability of soil can be described by random field theory and geostatistics. The research presented in this thesis focuses on quantifying the effect of soil variability on the bearing capacity of rough strip footings on single and two layered, purely-cohesive, spatially variable soil profiles. This has been achieved by using Monte Carlo analysis, where the rough strip footings are founded on simulated soil profiles are analysed using finite element limit analysis. The simulations of virtual soil profiles are carried out using Local Average Subdivision (LAS), a numerical model based on the random field theory. An extensive parametric study has been carried out and the results of the analyses are presented as normalized means and coefficients of variation of bearing capacity factor, and comparisons between different cases are presented. The results indicate that, in general, the mean of the bearing capacity reduces as soil variability increases and the worst case scenario occurs when the correlation length is in the range of 0.5 to 1.0 times the footing width. The problem of estimating the bearing capacity of shallow strip footings founded on multi-layered soil profiles is very complex, due to the incomplete knowledge of interactions and relationships between parameters. Much research has been carried out on single- and two-layered homogeneous soil profiles. At present, the inaccurate weighted average method is the only technique available for estimating the bearing capacity of footing on soils with three or more layers. In this research, artificial neural networks (ANNs) are used to develop meta-models for bearing capacity estimation. ANNs are numerical modelling techniques that imitate the human brain capability to learn from experience. This research is limited to shallow strip footing founded on soil mass consisting of ten layers, which are weightless, purely cohesive and cohesive-frictional. A large number of data has been obtained by using finite element limit analysis. These data are used to train and verify the ANN models. The shear strength (cohesion and friction angle), soil thickness, and footing width are used as model inputs, as they are influencing factors of bearing capacity of footings. The model outputs are the bearing capacities of the footings. The developed ANN-based models are then compared with the weighted average method. Hand-calculation design formulae for estimation of bearing capacity of footings on ten-layered soil profiles, based on the ANN models, are presented. It is shown that the ANN-based models have the ability to predict the bearing capacity of footings on ten-layered soil profiles with a high degree of accuracy, and outperform traditional methods. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1368281 / Thesis (Ph.D.) - University of Adelaide, School of Civil, Environmental and Mining Engineering, 2009

Footings are often founded on multi-layered soil profiles. Real soil profiles are often multi-layered with material constantly varying with depth, which affects the footing response significantly. Furthermore, the properties of the soil are known to vary with location. The spatial variability of soil can be described by random field theory and geostatistics. The research presented in this thesis focuses on quantifying the effect of soil variability on the bearing capacity of rough strip footings on single and two layered, purely-cohesive, spatially variable soil profiles. This has been achieved by using Monte Carlo analysis, where the rough strip footings are founded on simulated soil profiles are analysed using finite element limit analysis. The simulations of virtual soil profiles are carried out using Local Average Subdivision (LAS), a numerical model based on the random field theory. An extensive parametric study has been carried out and the results of the analyses are presented as normalized means and coefficients of variation of bearing capacity factor, and comparisons between different cases are presented. The results indicate that, in general, the mean of the bearing capacity reduces as soil variability increases and the worst case scenario occurs when the correlation length is in the range of 0.5 to 1.0 times the footing width. The problem of estimating the bearing capacity of shallow strip footings founded on multi-layered soil profiles is very complex, due to the incomplete knowledge of interactions and relationships between parameters. Much research has been carried out on single- and two-layered homogeneous soil profiles. At present, the inaccurate weighted average method is the only technique available for estimating the bearing capacity of footing on soils with three or more layers. In this research, artificial neural networks (ANNs) are used to develop meta-models for bearing capacity estimation. ANNs are numerical modelling techniques that imitate the human brain capability to learn from experience. This research is limited to shallow strip footing founded on soil mass consisting of ten layers, which are weightless, purely cohesive and cohesive-frictional. A large number of data has been obtained by using finite element limit analysis. These data are used to train and verify the ANN models. The shear strength (cohesion and friction angle), soil thickness, and footing width are used as model inputs, as they are influencing factors of bearing capacity of footings. The model outputs are the bearing capacities of the footings. The developed ANN-based models are then compared with the weighted average method. Hand-calculation design formulae for estimation of bearing capacity of footings on ten-layered soil profiles, based on the ANN models, are presented. It is shown that the ANN-based models have the ability to predict the bearing capacity of footings on ten-layered soil profiles with a high degree of accuracy, and outperform traditional methods. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1368281 / Thesis (Ph.D.) - University of Adelaide, School of Civil, Environmental and Mining Engineering, 2009

Made available in DSpace on 2014-06-11T19:31:37Z (GMT). No. of bitstreams: 0 Previous issue date: 2003Bitstream added on 2014-06-13T20:22:35Z : No. of bitstreams: 1 nagaoka_me_dr_botfca.pdf: 501587 bytes, checksum: a5d05cfa41f21298548d31b5d95dc6b1 (MD5) / Este trabalho apresenta a aplicação de um sistema inteligente utilizando redes neurais artificiais para estimar valores de densidade do solo, a partir de parâmetros referentes à resistência do solo à penetração. Foram considerados solos preparados e não preparados, os não preparados foram os seguintes : teor de argila menor que 30 % (solo tipo 1), de 30 a 50 % (solo tipo 2) e maior que 50 % (solo tipo 3). Os preparados foram os seguintes: um com teor de argila menor que 30 % (solo tipo 1) e o outro com teor de argila maior que 50 % (solo tipo 3). O objetivo principal deste trabalho foi implementar diversas redes neurais do tipo perceptron multicamadas, alimentando-as com resistência do solo à penetração, teor de água e teor de argila, tendo como variável de saída a densidade do solo. Cada rede foi treinada variando o número de camadas escondidas e também variando o número de neurônios, de 10 a 40, em cada camada. Para cada arquitetura, a rede foi treinada 10 vezes, escolhendo-se no final do treinamento a arquitetura com menor erro relativo médio e menor variância em relação aos dados de validação. As análises realizadas mostraram que as arquiteturas de rede com apenas uma camada escondida forneceram melhores resultados. Todas as redes tiveram melhor desempenho em solo não preparado do que em solo preparado. A rede de arquitetura de 3 entradas, uma camada escondida com 30 neurônios e 1 saída forneceu excelente resultado para solo não preparado (com teor de argila entre 30 e 50 %). Constatou-se que a rede quando treinada com dados do solo preparado, juntamente com dados do solo não preparado, melhorou os resultados de estimação para o solo preparado, mas piorou para os solos não preparados. Constatou também que a rede quando treinada junto com dados que contém solo solto fornece resultados imprecisos. O mesmo ocorreu para dados com teor de água elevado. / This work presents the development of an intelligent system using artificial neural networks to estimate values of soil density. Prepared and non-prepared soils were considered in this work. The non-prepared soils were the following ones: clay content lesser than 30 % (soil type 1), 30 to 50 % (soil type 2) and larger than 50 % (soil type 3). The prepared soils were the following ones: soil with clay content lesser than 30 % (soil type 1) and soil with clay content larger than 50 % (soil type 3). The main objective of this work was to implement several neural networks of type multilayer perceptron, feeding them with data concerning to the soil compaction characteristics. The output computed by the neural network was the respective density of these soils. Each neural network was trained varying both number of hidden layers and number of neurons, which was changed from 10 to 40 neurons in each layer. In each architecture the network was trained 10 times and selected architecture was always that having either the least mean relative error or the least variance in relation to validation data. The carried out analyses showed that the neural architectures having only a hidden layer were those that provided the best results. All neural networks have presented more efficient results for non-prepared soils than prepared soils. The neural network constituted by three inputs and one output, having 30 neurons at hidden layer, has provided excellent results for non-prepared soils (clay content between 30 and 50 %). It was also verified that the neural network when trained with data referent to non-prepared and soils, which were put in the same data set, it became the results referent to prepared soils more efficient, but the results for non-prepared soils become worse. Another observed point was when the network had been trained with data constituted by soft soil... (Complete abstract, click electronic address below).

Made available in DSpace on 2014-06-11T19:34:58Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-10-11Bitstream added on 2014-06-13T18:45:51Z : No. of bitstreams: 1 almeida_lf_dr_guara.pdf: 1479231 bytes, checksum: d1b34509ec45ba1ae48a6450780e381d (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Um importante aspecto no processo produtivo é proporcionar o funcionamento das máquinas o maior tempo possível sem o comprometimento na qualidade final do produto. Nesse sentido, a utilização de uma política de manutenção adequada se torna necessária para o monitoramento do desgaste dos componentes das máquinas a fim de aumentar o tempo de sua utilização sem comprometer a qualidade do produto. A manutenção baseada em condição se apresenta como a abordagem mais apropriada para esse controle. Dentre as diversas abordagens utilizadas para o desenvolvimento de programas para esse tipo de manutenção, as técnicas baseadas em Inteligência Artificial vêm se destacando no que diz respeito ao seu desempenho. Diante desse contexto, essa tese propõe uma Rede Neural Artificial, a qual, devidamente parametrizada, possibilita sua aplicação tanto para análise de vibrações quanto análise de partículas de desgaste. Para tanto, foi implementado um protótipo denominado NEURALNET-CBM, subdividido em dois módulos, Vibrações e Partículas. Os resultados dos testes mostram a efetividade da rede proposta, com um índice de acerto acima de 90% na classificação e identificação de defeitos e partículas de desgaste. / An important aspect in the production process is to ensure the operability of a machine as long as possible without interfering on the final quality product. In this way, the use of a suitable maintenance policy is critical for monitoring the wear of the machine components in order to increase your useful life without any compromise of the product quality. The Condition-Based Maintenance is presented as the most appropriate approach for this control. Among several methods used to develop systems for this type of maintenance, techniques Artificial Intelligence has been standing out in relation their performance. Therefore, this thesis proposes a Artificial Neural Network, which, properly parameterized, it makes possible its application for both vibration and wear particle analysis. For this, we implemented a prototype named NEURALNET-CBM, divided into two modules: Vibration and Particle. The test results show the effectiveness of the proposed network, with accuracy rate greater than 90% in classifying and identification of defects and wear particles.

Made available in DSpace on 2014-06-11T19:22:26Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-07-29Bitstream added on 2014-06-13T19:07:53Z : No. of bitstreams: 1 ferreira_ms_me_prud.pdf: 1897420 bytes, checksum: 1aede604709494154b2f75d18806c9fc (MD5) / O conhecimento da distribuição espacial da concentração de componentes da água é de fundamental importância para inferir a respeito dos processos ecológicos que ocorrem num sistema hídrico sendo, entretanto, de difícil obtenção. Dentre as variáveis que merecem atenção no monitoramento de ambientes aquáticos, destaca-se a clorofila a, a qual é uma substância presente em algas responsáveis pela fotossíntese, organismos que constituem a base da cadeia alimentar nesses ambientes. Por se tratar de um pigmento fotossintetizante, a clorofila a apresenta a propriedade de interagir com a radiação eletromagnética, e dessa interação resultam diferentes processos, identificáveis por meio de sensores remotos. Assim sendo, a presente pesquisa se propôs a desenvolver um método de inferência da concentração de clorofila a utilizando Redes Neurais Artificiais (RNA). Utilizou-se como dados de entrada para a inferência combinações de bandas espectrais de uma imagem World View-2 e valores de concentração de clorofila a obtidos com um fluorômetro de campo, o qual possibilitou uma amostragem densa na área de estudos. A imagem multiespectral foi corrigida radiometricamente, eliminando efeitos de instrumentação e atmosféricos. Ainda, efetuou-se uma suavização espectral em cada uma das bandas e foi avaliado se esse tratamento na imagem possibilitaria... / The knowledge of the spatial distribution of water components concentrations is of fundamental importance to infer about the ecological processes that occur in an aquatic system, however, is difficult to obtain it. Among the variables that deserve attention in the monitoring of aquatic environments, cite the chlorophyll a, which is a substance of photosynthetic algae, organisms that are the basis of the food chain in these environments. Because it is a photosynthetic pigment, chlorophyll a has the property to interact with electromagnetic radiation, and it results in different processes, identifiable through remote sensing. Thus, this research intended to develop a chlorophyll a concentration inference method using Artificial Neural Networks (ANN). As input for the inference, it was used combinations of World View-2 spectral bands and chlorophyll a concentration values obtained with a field fluorometer, which allowed a dense sampling in the study area. The multispectral imagery was radiometrically corrected, eliminating the instrumentation and atmospheric effects. Still, it was performed a spectral smoothing in each of the spectral bands and evaluated whether this treatment would give... (Complete abstract click electronic access below)

Orientador: Mauro Hugo Mathias / Banca: Alvaro Manoel Souza Soares / Banca: José Elias Tomazini / Banca: Francisco Carlos Parquet Bizarria / Banca: Carlos Henrique Netto Lahoz / Resumo: Um importante aspecto no processo produtivo é proporcionar o funcionamento das máquinas o maior tempo possível sem o comprometimento na qualidade final do produto. Nesse sentido, a utilização de uma política de manutenção adequada se torna necessária para o monitoramento do desgaste dos componentes das máquinas a fim de aumentar o tempo de sua utilização sem comprometer a qualidade do produto. A manutenção baseada em condição se apresenta como a abordagem mais apropriada para esse controle. Dentre as diversas abordagens utilizadas para o desenvolvimento de programas para esse tipo de manutenção, as técnicas baseadas em Inteligência Artificial vêm se destacando no que diz respeito ao seu desempenho. Diante desse contexto, essa tese propõe uma Rede Neural Artificial, a qual, devidamente parametrizada, possibilita sua aplicação tanto para análise de vibrações quanto análise de partículas de desgaste. Para tanto, foi implementado um protótipo denominado NEURALNET-CBM, subdividido em dois módulos, Vibrações e Partículas. Os resultados dos testes mostram a efetividade da rede proposta, com um índice de acerto acima de 90% na classificação e identificação de defeitos e partículas de desgaste. / Abstract: An important aspect in the production process is to ensure the operability of a machine as long as possible without interfering on the final quality product. In this way, the use of a suitable maintenance policy is critical for monitoring the wear of the machine components in order to increase your useful life without any compromise of the product quality. The Condition-Based Maintenance is presented as the most appropriate approach for this control. Among several methods used to develop systems for this type of maintenance, techniques Artificial Intelligence has been standing out in relation their performance. Therefore, this thesis proposes a Artificial Neural Network, which, properly parameterized, it makes possible its application for both vibration and wear particle analysis. For this, we implemented a prototype named NEURALNET-CBM, divided into two modules: Vibration and Particle. The test results show the effectiveness of the proposed network, with accuracy rate greater than 90% in classifying and identification of defects and wear particles. / Doutor

In this thesis I look into how one can train and run an artificial neural network using Compute Shader and what kind of performance can be expected. An artificial neural network is a computational model that is inspired by biological neural networks, e.g. a brain. Finding what kind of performance can be expected was done by creating an implementation that uses Compute Shader and then compare it to the FANN library, i.e. a fast artificial neural network library written in C. The conclusion is that you can improve performance by training an artificial neural network on the compute shader as long as you are using non-trivial datasets and neural network configurations.

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.