• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 17
  • 8
  • 4
  • 2
  • 2
  • Tagged with
  • 39
  • 39
  • 17
  • 16
  • 15
  • 11
  • 8
  • 7
  • 7
  • 7
  • 6
  • 6
  • 6
  • 5
  • 5
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

Búsqueda de recursos en redes peer-to-peer totalmente descentralizadas basada en redes neuronales artificiales

Corbalán, Leonardo César 11 November 2014 (has links)
Las redes Peer-to-Peer (P2P) puras no estructuradas como Gnutella, dónde los nodos se conectan entre sí como pares o iguales, sin roles diferenciados ni jerarquías de ninguna clase, son sistemas distribuidos, dinámicos, sin punto alguno de centralización, que favorecen la robustez y tolerancia a fallos. Sin embargo, la búsqueda de recursos en estos sistemas constituye un problema esencial. El algoritmo de búsqueda BFS de Gnutella genera gran cantidad de tráfico dificultando su escalabilidad. Esta tesis propone un nuevo algoritmo de búsqueda denominado Búsqueda Inteligente Incremental P2P (BII-P2P) en el que los nodos, asistidos por sus redes neuronales locales, propagan selectivamente las solicitudes de búsquedas sólo al subconjunto más apropiado de vecinos. Así se mejora significativamente el algoritmo de Gnutella consiguiendo mayor porcentaje de hallazgos con menor cantidad de tráfico generado sobre la red P2P. El rendimiento de este algoritmo de búsqueda inteligente se ve potenciado por una conveniente estrategia de exploración incremental.
2

Forecasting Global Temperature Variations by Neural Networks

Miyano, Takaya, Girosi, Federico 01 August 1994 (has links)
Global temperature variations between 1861 and 1984 are forecast usingsregularization networks, multilayer perceptrons and linearsautoregression. The regularization network, optimized by stochasticsgradient descent associated with colored noise, gives the bestsforecasts. For all the models, prediction errors noticeably increasesafter 1965. These results are consistent with the hypothesis that thesclimate dynamics is characterized by low-dimensional chaos and thatsthe it may have changed at some point after 1965, which is alsosconsistent with the recent idea of climate change.s
3

Smarter NEAT Nets

Dehaven, Ryan Swords 01 August 2013 (has links) (PDF)
This paper discusses a modification to improve usability and functionality of a ge- netic neural net algorithm called NEAT (NeuroEvolution of Augmenting Topolo- gies). The modification aims to accomplish its goal by automatically changing parameters used by the algorithm with little input from a user. The advan- tage of the modification is to reduce the guesswork needed to setup a successful experiment with NEAT that produces a usable Artificial Intelligence (AI). The modified algorithm is tested against the unmodified NEAT with several different setups and the results are discussed. The algorithm shows strengths in some areas but can increase the runtime of NEAT due to the addition of parameters into the solution search space.
4

Modelagem estocástica de uma população de neurônios / Stochastic modelling of a population of neurons

Yaginuma, Karina Yuriko 08 May 2014 (has links)
Nesta tese consideramos uma nova classe de sistemas markovianos de partículas com infinitas componentes interagentes. O sistema representa a evolução temporal dos potenciais de membrana de um conjunto infinito de neurônios interagentes. Provamos a existência e unicidade do processo construindo um pseudo-algoritmo de simulação perfeita e mostrando que este algoritmo roda em um número finito de passos quase certamente. Estudamos também o comportamento do sistema quando consideramos apenas um conjunto finito de neurônios. Neste caso, construímos um procedimento de simulação perfeita para o acoplamento entre o processo limitado a um conjunto finito de neurônios e o processo que considera todos os neurônios do sistema. Como consequência encontramos um limitante superior para a probabilidade de discrepância entre os processos. / We consider a new class of interacting particle systems with a countable number of interacting components. The system represents the time evolution of the membrane potentials of an infinite set of interacting neurons. We prove the existence and uniqueness of the process, by the construction of a perfect simulation procedure. We show that this algorithm is successful, that is, we show that the number of steps of the algorithm is finite almost surely. We also study the behaviour of the system when we consider only a finite number of neurons. In this case, we construct a perfect simulation procedure for the coupling of the process with a finite number of neurons and the process with a infinite number of neurons. As a consequence we obtain an upper bound for the error we make when sampling from a finite set of neurons instead of the infinite set of neurons.
5

Modelagem matemática de um processo industrial de produção de cloro e soda por eletrólise de salmoura visando sua otimização. / Mathematical modeling of an industrial process for chlorine and caustic manufacturing using brine electrolysis aiming at its optimization.

De Jardin Júnior, Roberto Nicolas 14 September 2006 (has links)
O presente trabalho envolve a elaboração de um modelo matemático para um processo industrial de produção de cloro e soda a partir de salmoura, visando sua otimização em termos de eficiência de produção e dos custos dos consumos de energia elétrica e vapor. O estudo contemplou duas etapas do processo: eletrólise e concentração de licor de NaOH por evaporação. Para a unidade de eletrólise não foram encontrados na literatura modelos fenomenológicos adequados à simulação do processo. Por essa razão, foram desenvolvidos modelos empíricos baseados em redes neurais tipo ?feedforward? constituídas por três camadas, a partir de dados da operação industrial. Para a unidade de evaporação foi elaborado um balanço de energia adequado à estimativa do consumo de vapor. Porém, devido à falta de modelos para previsão das relações de equilíbrio para o sistema, o modelo fenomenológico foi substituído por um modelo de redes neurais tipo ?feedforward? de três camadas também para essa unidade. Para ajuste dos modelos, uma base de dados foi montada a partir de dados de operação do processo da Carbocloro S.A. Indústrias Químicas, localizada em Cubatão-SP, analisados por meio de técnicas estatísticas multivariadas, visando detectar e eliminar erros grosseiros e dados anômalos, além de identificar correlações entre variáveis e diferentes regimes operacionais da planta de produção de cloro e soda. Os modelos ajustados para os diferentes circuitos de células de eletrólise, bem como para a etapa de evaporação, apresentaram boa concordância com os dados operacionais. Isto possibilitou sua utilização para simular a operação das unidades de células eletrolíticas e evaporação no processo industrial de produção de cloro-soda, com células tipo diafragma. O modelo matemático baseado em redes neurais foi utilizado em estudos de otimização do processo, de modo a maximizar o ganho financeiro na unidade industrial, para uma dada condição de operação. / The present work consists on the development of a mathematical model on an industrial chlorine and sodium hydroxide production plant, aiming at the optimization of production efficiency and costs saving concerning electrical energy and vapor consumption. Two process steps were considered in the study: electrolysis and NaOH-liquor concentration by evaporation. Since there are no adequate models reported in the literature for simulating electrolysis-based processes like the one considered, empirical models for the different types of electrolysis cells were developed based on the fitting of neural networks to operational data from industrial operation. In this case, feedforward neural networks containing three neuron layers were fitted to the data. The raw data obtained from industrial operation at Carbocloro plant, in Cubatão ? SP, were first treated by means of multivariate statistical techniques, with the purpose of detecting and eliminating data containing gross errors and outliers, as well as to identify correlations among variables and different operational regimes of the industrial plant. Although material and energy balances for the evaporation step have been initially adopted, this approach could not be used in simulations due to the lack of valid models to predict liquid ? vapor equilibria for the specific system. Thus, a neural network model was also fitted to data from operation of the evaporation step. Fitting of the neural network models resulted in good agreement between model predictions and measured values of the model output variables, and this enabled their use in simulation studies for the electrolysis and evaporation process steps. The neural network-based mathematical model was utilized in process optimization studies aiming at the best financial gain under given operational conditions.
6

Hybrid neural net and physics based model of a lithium ion battery

Refai, Rehan 12 July 2011 (has links)
Lithium ion batteries have become one of the most popular types of battery in consumer electronics as well as aerospace and automotive applications. The efficient use of Li-ion batteries in automotive applications requires well designed battery management systems. Low order Li-ion battery models that are fast and accurate are key to well- designed BMS. The control oriented low order physics based model developed previously cannot predict the temperature and predicts inaccurate voltage dynamics. This thesis focuses on two things: (1) the development of a thermal component to the isothermal model and (2) the development of a hybrid neural net and physics based battery model that corrects the output of the physics based model. A simple first law based thermal component to predict the temperature model is implemented. The thermal model offers a reasonable approximation of the temperature dynamics of the battery discharge over a wide operating range, for both a well-ventilated battery as well as an insulated battery. The model gives an accurate prediction of temperature at higher SOC, but the accuracy drops sharply at lower SOCs. This possibly is due to a local heat generation term that dominates heat generation at lower SOCs. A neural net based modeling approach is used to compensate for the lack of knowledge of material parameters of the battery cell in the existing physics based model. This model implements a neural net that corrects the voltage output of the model and adds a temperature prediction sub-network. Given the knowledge of the physics of the battery, sparse neural nets are used. Multiple types of standalone neural nets as well as hybrid neural net and physics based battery models are developed and tested to determine the appropriate configuration for optimal performance. The prediction of the neural nets in ventilated, insulated and stressed conditions was compared to the actual outputs of the batteries. The modeling approach presented here is able to accurately predict voltage output of the battery for multiple current profiles. The temperature prediction of the neural nets in the case of the ventilated batteries was harder to predict since the environment of the battery was not controlled. The temperature predictions in the insulated cases were quite accurate. The neural nets are trained, tested and validated using test data from a 4.4Ah Boston Power lithium ion battery cell. / text
7

Reconocimiento de gestos dinámicos

Quiroga, Facundo January 2014 (has links)
El objetivo de esta tesina es estudiar, desarrollar, analizar y comparar distintas técnicas de aprendizaje automático aplicables al reconocimiento automático de gestos dinámicos. Para ello, se definió un modelo de gestos a reconocer, se generó una base de datos de prueba con gestos llamadas LNHG, y se estudiaron e implementaron clasificadores basados en máquinas de vectores de soporte (SVM), redes neuronales feedfoward (FF) y redes neuronales competitivas (CPN), utilizando representaciones locales y globales para caracterizar los gestos. Además, se propone un nuevo modelo de reconocimiento de gestos, el clasificador neuronal competitivo (CNC). Los gestos a reconocer son movimientos de la mano, con invariancia a la velocidad, la rotación, la escala y la traslación. La captura de la información referida a los gestos para generar la base de datos se realizó mediante el dispositivo Kinect y su SDK correspondiente, que reconoce las partes del cuerpo y determina sus posiciones en tiempo real. Los clasificadores se entrenaron con dichos datos para poder determinar si una secuencia de posiciones de la mano es un gesto. Se implementó una librería de clasificadores con los métodos mencionados anteriormente, junto con las transformaciones para llevar una secuencia de posiciones a una representación adecuada para el reconocimiento. Se realizaron experimentos con la base de datos LNHG, compuesta de gestos que representan dígitos y letras, y con un base de datos de otro autor con gestos típicos de interacción, obteniendo resultados satisfactorios.
8

Coding strategies for genetic algorithms and neural nets

Hancock, Peter J. B. January 1993 (has links)
The interaction between coding and learning rules in neural nets (NNs), and between coding and genetic operators in genetic algorithms (GAs) is discussed. The underlying principle advocated is that similar things in "the world" should have similar codes. Similarity metrics are suggested for the coding of images and numerical quantities in neural nets, and for the coding of neural network structures in genetic algorithms. A principal component analysis of natural images yields receptive fields resembling horizontal and vertical edge and bar detectors. The orientation sensitivity of the "bar detector" components is found to match a psychophysical model, suggesting that the brain may make some use of principal components in its visual processing. Experiments are reported on the effects of different input and output codings on the accuracy of neural nets handling numeric data. It is found that simple analogue and interpolation codes are most successful. Experiments on the coding of image data demonstrate the sensitivity of final performance to the internal structure of the net. The interaction between the coding of the target problem and reproduction operators of mutation and recombination in GAs are discussed and illustrated. The possibilities for using GAs to adapt aspects of NNs are considered. The permutation problem, which affects attempts to use GAs both to train net weights and adapt net structures, is illustrated and methods to reduce it suggested. Empirical tests using a simulated net design problem to reduce evaluation times indicate that the permutation problem may not be as severe as has been thought, but suggest the utility of a sorting recombination operator, that matches hidden units according to the number of connections they have in common. A number of experiments using GAs to design network structures are reported, both to specify a net to be trained from random weights, and to prune a pre-trained net. Three different coding methods are tried, and various sorting recombination operators evaluated. The results indicate that appropriate sorting can be beneficial, but the effects are problem-dependent. It is shown that the GA tends to overfit the net to the particular set of test criteria, to the possible detriment of wider generalisation ability. A method of testing the ability of a GA to make progress in the presence of noise, by adding a penalty flag, is described.
9

Um modelo de sistema nervoso para o problema do controle de animação por dinâmica direta / A model of nervous system for the problem of the control of animation for direct dynamics

Nogueira, Yuri Lenon Barbosa January 2007 (has links)
NOGUEIRA, Yuri Lenon Barbosa. Um modelo de sistema nervoso para o problema do controle de animação por dinâmica direta. 2007. 70 f. Dissertação (Mestrado em ciência da computação)- Universidade Federal do Ceará, Fortaleza-CE, 2007. / Submitted by Elineudson Ribeiro (elineudsonr@gmail.com) on 2016-07-12T19:02:24Z No. of bitstreams: 1 2007_dis_ylbnogueira.pdf: 1821730 bytes, checksum: 0c95ce921f248632982cb423d7e67c58 (MD5) / Approved for entry into archive by Rocilda Sales (rocilda@ufc.br) on 2016-07-22T12:47:46Z (GMT) No. of bitstreams: 1 2007_dis_ylbnogueira.pdf: 1821730 bytes, checksum: 0c95ce921f248632982cb423d7e67c58 (MD5) / Made available in DSpace on 2016-07-22T12:47:46Z (GMT). No. of bitstreams: 1 2007_dis_ylbnogueira.pdf: 1821730 bytes, checksum: 0c95ce921f248632982cb423d7e67c58 (MD5) Previous issue date: 2007 / Direct dynamics animation consists of synthesizing the movements of a model from the specification of its physical properties (mass and moment of inertia), the conditions of bond between its contracting parties, the conditions of contact with other bodies and the forces that acts on it. This approaching has the advantage to generate animations with physical realism. The problem, that continues relevant as inquiry object, is the control of the model: “What forces must be applied to the model to generate the desired movement?”. The solution of the problem, presented in this work, assumes that the studied model consists of a structure of rigid link bodies whose movements are generated by internal actuators, with its forces defined by a nervous system. With use of artificial neural networks and evolutionary computation, the proposed controller is capable of adapting itself to control different articulated models, and to generate varied types of movements while it keeps the stability even with small variations of the terrain. The presented model possesses, in its core, a Central Pattern Generator (CPG) based on neural oscillators, that has their activities regulated by the sensorial module, to allow the balance of the structure and stability of the movement, responding to environment variations. For the adaptation to the articulated structure and learning of movements, the controller has a cognitive module, responsible for the search of neural parameters, through genetic algorithms, and the feedback networks (sensorial answers to environment variations), with genetic programming. Results are presented related to the control of models humanoid, cheetah, frog, luxo and luxo-2, having these last two ones equal topologies, but with variations in the sizes of the bodies and freedom of the joints. All the models are tested in plain land and with slope. / A animaçãao por dinâmica direta consiste em sintetizar os movimentos de um modelo a partir da especificação de suas propriedades físicas (massa e momento de inércia), das condições de vínculo entre suas partes componentes, das condições de contato com outros corpos e das forças que nele atuam. Essa abordagem tem a vantagem de gerar animações com realismo físico. O problema, que continua relevante como objeto de investigação, é o de controle do modelo: “Que forças devem ser aplicadas ao modelo para gerar o movimento desejado?”. A solução do problema proposto apresentada neste trabalho assume que o modelo estudado constitui-se de uma estrutura de corpos rígidos articulados cujos movimentos são gerados por atuadores internos, com suas forças definidas por um sistema nervoso. Com o uso de redes neurais artificiais e computação evolucionária, o controlador proposto é capaz de adaptar-se para controlar diferentes modelos articulados, e para gerar variados tipos de movimentos enquanto mantém a estabilidade mesmo quando há pequenas variações do terreno. O modelo proposto possui, em seu núcleo, um gerador central de padrões (CPG - Central Pattern Generator) baseado em osciladores neurais, e o mesmo tem sua atividade regulada por módulos sensoriais, para permitir o equilíbrio da estrutura e estabilidade do movimento, respondendo àss variações do ambiente. Para a adaptação à estrutura articulada e aprendizagem de movimentos, o controlador possui ainda um módulo cognitivo, responsável pela busca dos parâmetros neurais, através de algoritmos genéticos, e das redes de retroalimentação (sensoriamento), com programação genética. Resultados são apresentados em associação ao controle dos modelos humanóide, cheetah, sapo, luxo e luxo-2, sendo esses dois últimos iguais topologicamente, mas com variações nos tamanhos dos corpos e liberdade das juntas. Todos os modelos são testados em terreno plano e com rampa.
10

Aplicação de redes neurais no controle de teores de cobre e ouro do depósito de Chapada (GO)

Cintra, Evandro Cardoso [UNESP] 28 November 2003 (has links) (PDF)
Made available in DSpace on 2014-06-11T19:32:21Z (GMT). No. of bitstreams: 0 Previous issue date: 2003-11-28Bitstream added on 2014-06-13T20:03:46Z : No. of bitstreams: 1 cintra_ec_dr_rcla.pdf: 4074998 bytes, checksum: 46f75c3ee3bbcbcc6fcce47c41570c71 (MD5) / Este estudo desenvolve a aplicação da técnica de redes neurais artificiais no controle de teor de minério em frentes de lavra a partir de observações geológicas e geotécnicas. A área de estudo da aplicação é o depósito de cobre e ouro de Chapada (Goiás), hospedado por rochas da seqüência vulcano-sedimentar neoproterozóica de Chapada-Mara Rosa. Trata-se de um depósito mineral tipo epigenético, ligado a processos de alteração hidrotermal, associado a zonas estruturalmente favoráveis. As observações geológicas e geotécnicas constituem um banco de dados com 21.212 registros e 21 variáveis, provenientes de amostras de 237 furos de sondagem rotativa diamantada. As variáveis de entrada incluem litologia, porcentagem de sulfetos, razão calcopirita/pirita, freqüência de fraturas, RQD, e alterações hidrotermais tais como: cloritização, sericitização, silicificação, epidotização, carbonatização e piritização. As variáveis de saída são: teores de cobre e ouro. O modelo de rede neural utilizado foi o de múltiplas camadas (MLP) alimentada adiante ( feedforward ) totalmente interconectada, com 30 neurônios na camada oculta e 2 neurônios na camada de saída. A rede foi treinada com o algoritmo de retropropagação de Levenberg-Marquardt acoplado com regularização bayesiana. Obteve-se um índice de acertos de 80% na predição de teores de cobre em bancadas simuladas. / This study deals with application of artificial neural networks (ANNs) on grade control at mine sites inputting both geological and geotechnical variables. Case study is Chapada copper-gold deposit (Goiás, Brazil), located in the neoproterozoic Chapada-Mara Rosa volcano-sedimentary sequence. Ore is closely related to hydrothermal alteration, structurally controlled. The geological and geotechnical database contain 21,212 records on 21 variables taken from 237 diamond drill holes. Input variables include lithology, sulfide percentage, chalcopyrite/pyrite ratio, fracture frequency, RQD, and hydrothermal alterations such as chloritization, sericitization, silicification, epidotization, carbonatization and pyritization. Output variables are gold and copper grades. Neural network model is feedforward multi-layer perceptron (MLP), fully connected with 30 hidden and 2 output neurons. Network was trained with Levenberg-Marquardt backpropagation algorithm associated with bayesian regularization. Success rate on predicting copper grades on simulated mine benches was over 80%.

Page generated in 0.0662 seconds