Este trabalho realiza uma análise dos modelos pertencentes à Computação Neural e à Computação Evolutiva visando identificar semelhanças entre as áreas e sustentar mapeamentos entre as semelhanças identificadas. Neste contexto, a identificação de similaridades visando à resolução de problemas de otimização combinatorial resulta em uma comparação entre a Máquina de Boltzmann e os Algoritmos Evolutivos binários com população composta por um único indivíduo pai e um único indivíduo descendente. Como forma de auxiliar nas análises, o trabalho utiliza o Problema do Caixeiro Viajante como plataforma de ensaios, propondo mapeamentos entre as equações da Máquina de Boltzmann e os operadores evolutivos da Estratégia Evolutiva (1+1)-ES. / An analysis between the Evolutionary Computation and the Neural Computation fields was presented in order to identify similarities and mappings between the theories. In the analysis, the identification of similarities between the models designed for combinatorial optimization problems results in a comparison between the Boltzmann Machine and the Two-Membered Evolutionary Algorithms. In order to analyze the class of combinatorial optimization problems, this work used the Traveling Salesman Problem as a study subject, where the Boltzmann Machine equations were used to implement the evolutionary operators of an Evolution Strategy (1+1)-ES.

Neste trabalho foi realizada uma proposta de utilização de Sistemas de Inferência Nebulosos para controlar, em tempo de execução, parâmetros de Algoritmos Genéticos. Esta utilização busca melhorar o desempenho de Algoritmos Genéticos diminuindo, ao mesmo tempo: a média de iterações necessárias para que um Algoritmo Genético encontre o valor ótimo global procurado; bem como diminuindo o número de execuções do mesmo que não são capazes de encontrar o valor ótimo global procurado, nem mesmo para quantidades elevadas de iterações. Para isso, foram analisados os resultados de diversos experimentos com Algoritmos Genéticos, resolvendo instâncias dos problemas de Minimização de Funções e do Caixeiro Viajante, sob diferentes configurações de parâmetros. Com base nos resultados obtidos a partir destes experimentos, foi proposto um modelo com a troca de valores de parâmetros de Algoritmos Genéticos, em tempo de execução, pela utilização de Sistemas de Inferência Nebulosos, de forma a melhorar o desempenho do sistema, minimizando ambas as medidas citadas anteriormente. / This work addressed a proposal of the application of Fuzzy Systems to adjust parameters of Genetic Algorithms, during execution time. This application attempts to improve the performance of Genetic Algorithms by diminishing, at the same time: the average number of necessary generations for a Genetic Algorithm to find the desired global optimum value, as well as diminishing the number of executions of a Genetic Algorithm that are not capable of finding the desired global optimum value even for high numbers of generations. For that purpose, the results of many experiments with Genetic Algorithms were analyzed; addressing instances of the Function Minimization and the Travelling Salesman problems, under different parameter configurations. With the results obtained from these experiments, a model was proposed, for the exchange of parameter values of Genetic Algorithms, in execution time, by using Fuzzy Systems, in order to improve the performance of the system, minimizing both of the measures previously cited.

This thesis describes the addition of novel Geometric Guides to a generative Computer-Aided Design (CAD) application that supports early-stage concept generation. The application generates and evolves abstract 3D shapes, used to inspire the form of new product concepts. It was previously a conventional Interactive Evolutionary system where users selected shapes from evolving populations. However, design industry users wanted more control over the shapes, for example by allowing the system to influence the proportions of evolving forms. The solution researched, developed, integrated and tested is a more cooperative human-machine system combining classic user interaction with innovative geometric analysis. In the literature review, different types of Interactive Evolutionary Computation (IEC), Pose Normalisation (PN), Shape Comparison, and Minimum-Volume Bounding Box approaches are compared, with some of these technologies identified as applicable for this research. Using its Application Programming Interface, add-ins for the Siemens NX CAD system have been developed and integrated with an existing Interactive Evolutionary CAD system. These add-ins allow users to create a Geometric Guide (GG) at the start of a shape exploration session. Before evolving shapes can be compared with the GG, they must be aligned and scaled (known as Pose Normalisation in the literature). Computationally-efficient PN has been achieved using geometric functions such as Bounding Box for translation and scaling, and Principle Axes for the orientation. A shape comparison algorithm has been developed that is based on the principle of non-intersecting volumes. This algorithm is also implemented with standard, readily available geometric functions, is conceptually simple, accessible to other researchers and also offers appropriate efficacy. Objective geometric testing showed that the PN and Shape Comparison methods developed are suitable for this guiding application and can be efficiently adapted to enhance an Interactive Evolutionary Design system. System performance with different population sizes was examined to indicate how best to use the new guiding capabilities to assist users in evolutionary shape searching. This was backed up by participant testing research into two user interaction strategies. A Large Background Population (LBP) approach where the GG is used to select a sub-set of shapes to show to the user was shown to be the most effective. The inclusion of Geometric Guides has taken the research from the existing aesthetic focused tool to a system capable of application to a wider range of engineering design problems. This system supports earlier design processes and ideation in conceptual design and allows a designer to experiment with ideas freely to interactively explore populations of evolving solutions. The design approach has been further improved, and expanded beyond the previous quite limited scope of form exploration.

[pt] O crescente consumo de energia é preocupante, principalmente pelo uso de sistemas de condicionamento de ar e de iluminação artificial. Nas edificações modernas, os projetos arquitetônicos vêm negligenciando os fatores que proporcionam o conforto ambiental. Baseando-se nos conceitos da arquitetura sustentável, esta dissertação propõe e modela um sistema que otimiza os parâmetros da edificação que influenciarão no consumo de energia elétrica, nos custos com a construção e na emissão de poluentes pela edificação. Propõe-se um modelo de algoritmos genéticos que, juntamente com um programa de simulação de energia, EnergyPlus, constitui o modelo evolucionário desenvolvido neste trabalho. Este modelo otimiza parâmetros como: dimensionamento de aberturas e de pédireito; orientação da edificação; condicionamento do ar; disposição de árvores no entorno da edificação; etc . O modelo evolucionário tem sua ação e eficácia testados em estudo de casos - edificações desenhadas por projetista -, em que se alteram: espessura das paredes, altura de pé direito, largura de janelas, orientação quanto ao Norte geográfico, localização de elementos sombreantes (árvores), uso ou não de bloqueadores solares. Estes fatores influenciarão no conforto térmico da edificação e, consequentemente, no consumo elétrico dos sistemas de condicionamento de ar e de iluminação artificial, que por sua vez, influenciam os parâmetros que se pretende otimizar. Os resultados obtidos mostram que as otimizações feitas pelo modelo evolucionário foram efetivas, minimizando o consumo de energia pelos sistemas de condicionamento de ar e de iluminação artificial em comparação com os resultados obtidos com as edificações originais fornecidas pelo projetista. / [en] The continuous rising on energy consumption is a concerning issue, especially regarding the use of air conditioning systems and artificial lighting. In modern buildings, architectural designs are neglecting the factors that provide environmental comfort in a natural way. Based on concepts of sustainable architecture, this work proposes and models a system that optimizes the parameters of a building that influence the consumption of electricity, the costs with the building itself, and the emission of pollutants by these buildings. For this purpose a genetic algorithm model is proposed, which works together with an energy simulation program called EnergyPlus, both comprising the evolutionary model developed in this work. This model is able to optimize parameters like: dimensions of windows and ceiling height; orientation of a building; air conditioning; location of trees around a building; etc. The evolutionary model has its efficiency tested in case studies - buildings originally designed by a designer -, and the following specifications provided by the designer have been changed by the evolutionary model: wall thickness, ceiling height, windows width, building orientation, location of elements that perform shading function (trees), the use (or not) of sun blockers. These factors influence the building s heat comfort and therefore the energy consumption of air conditioning systems and artificial lighting which, in turn, influence the parameters that are meant to be optimized. The results show that the optimizations made by the evolutionary model were effective, minimizing the energy consumption for air conditioning systems and artificial light in comparison with the results obtained with the original buildings provided by the designer.

[pt] Este trabalho investiga métodos de detecção de falhas baseados em sistemas imunológicos artificiais, especificamente aqueles baseados no algoritmo de seleção negativa (NSA) e em outras técnicas de reconhecimento próprio/nãopróprio. Inicialmente, foi proposto um esquema de representação baseado em hiperesferas com centros e raios variáveis e três modelos capazes de gerar detectores, com esta representação, de forma eficiente. O primeiro modelo utiliza algoritmos genéticos onde cada gene do cromossomo contém um índice para um ponto de uma distribuição quasi-aleatória que servirá como centro do detector e uma função decodificadora responsável por determinar os raios apropriados. A aptidão do cromossomo é dada por uma estimativa do volume coberto através uma integral de Monte Carlo. O segundo modelo utiliza o particionamento Quadtree para gerar o posicionamento dos detectores e o valor dos raios. Este modelo pode realizar o particionamento a partir de uma função de detecção ou através de divisões recursivas de um detector inicial que ocupa todo o espaço. O terceiro modelo é inspirado nas redes imunológicas. Neste modelo, as células B representam os detectores e a rede formada por eles dá a posição e o raio de cada detector. Experimentos com dados sintéticos e reais demonstram a capacidade dos algoritmos propostos e que eles apresentam melhorias nos aspectos de escalabilidade e desempenho na detecção de falhas. / [en] This work investigates fault detection methods based on Artificial Immune Systems, specifically the negative selection algorithm (NSA) and other self/nonself recognition techniques. First, there was proposed a representation scheme based on hyperspheres with variable center and radius, and three models, which are very capable to generate detectors, based on that representation scheme, in an effective way. The first model employs Genetic Algorithms where each chromosome gene represents an index to a point in a quasi- random distribution, that will serve as a detector center, a decoder function will be responsible to determine the appropriate radius. The chromosome fitness is given by a valuation of the covered volume, which is calculated through a Monte Carlo integral. The second model uses the Quadtree space partition technique to generate the detectors positions and their radius. The space partition could be done by using a detection function or by recursive divisions of an initial detector that occupies the whole space. In third model, inspired on immune networks, the B cells represent the detectors and the network that is established by them gives the location and radius of each detector. Experiments with syntetic and real data show that the proposed algorithms improve scalability and perform better in fault detection.

This research presents engineering tools that address an important area impacting many persons worldwide: epilepsy. Over 60 million people are affected by epilepsy, a neurological disorder characterized by recurrent seizures that occur suddenly. Surgery and anti-epileptic drugs (AED s) are common therapies for epilepsy patients. However, only persons with seizures that originate in an unambiguous, focal portion of the brain are candidates for surgery, while AED s can lead to very adverse side-effects. Although medical devices based upon focal cooling, drug infusion or electrical stimulation are viable alternatives for therapy, a reliable method to automatically pinpoint dysfunctional brain and direct these devices is needed. This research introduces a method to effectively localize epileptic networks, or connectivity between dysfunctional brain, to guide where to insert electrodes in the brain for therapeutic devices, surgery, or further investigation. The method uses an evolutionary algorithm (EA) and frequent itemset mining (FIM) to detect and cluster frequent concentrations of epileptic neuronal action potentials within human intracranial electroencephalogram (EEG) recordings. In an experiment applying the method to seven patients with neocortical epilepsy (a total of 35 seizures), the approach reliably identifies the seizure onset zone, in six of the subjects (a total of 31 seizures). Hopefully, this research will lead to a better control of seizures and an improved quality of life for the millions of persons affected by epilepsy.

The object of this research is to optimize the overall power system performance using FACTS devices. Particularly, it is intended to improve the reliability, and the performance of the power system considering steady state operating condition as well as the system subjected to small and large disturbances. The methodology proposed to achieve this goal corresponds to an enhanced particle swarm optimizer (Enhanced-PSO) that is proven in this work to have several advantages, in terms of accuracy and computational effort, as compared with other existing methods. Once the performance of the Enhanced PSO is verified, a multi-stage PSO-based optimization framework is proposed for optimizing the power system reliability (N-1 contingency criterion). The algorithm finds optimal settings for present infrastructure (generator outputs, transformers tap ratios and capacitor banks settings) as well as optimal control references for distributed static series compensators (DSSC) and optimal locations, sizes and control settings for static compensator (STATCOM) units. Finally, a two-stage optimization algorithm is proposed to improve the power system performance in steady state conditions and when small and large perturbations are applied to the system. In this case, the algorithm provides optimal control references for DSSC modules, optimal location and sizes for capacitor banks, and optimal location, sizes and control parameters for STATCOM units (internal and external controllers), so that the loadability and the damping of the system are maximized at minimum cost. Simulation results throughout this research show a significant improvement of the power system reliability and performance after the system is optimized.

This thesis was inspired by the difficulties of artificial evolutionary systems in finding elegant and well structured, regular solutions. That is that the solutions found are usually highly disorganized, poorly structured and exhibit limited re-use, resulting in bloat and other problems. This is also true of previous developmental evolutionary systems, where structural regularity emerges only by chance. We hypothesise that these problems might be ameliorated by incorporating repeated evaluations on increasingly difficult problems in the course of a developmental process. This thesis introduces a new technique for learning complex problems from a family of structured increasingly difficult problems, Evolutionary Developmental Evaluation (EDE). This approach appears to give more structured, scalable and regular solutions to such families of problems than previous methods. In addition, the thesis proposes some bio-inspired components that are required by developmental evolutionary systems to take full advantage of this approach. The key part of this is the developmental process, in combination with a varying fitness function evaluated at multiple stages of development, generates selective pressure toward generalisation. This also means that parsimony in structure is selected for without any direct parsimony pressure. As a result, the system encourages the emergence of modularity and structural regularity in solutions. In this thesis, a new genetic developmental system called Developmental Tree Adjoining Grammar Guided Genetic Programming (DTAG3P), is implemented, embodying the requirements above. It is tested on a range of benchmark problems. The results indicate that the method generates more regularly-structured solutions than the competing methods. As a result, the system is able to scale, at least on the problem classes tested, to very complex instances the system encourages the emergence of modularity and structural regularity in solutions. In this thesis, a new genetic developmental system called Developmental Tree Adjoining Grammar Guided Genetic Programming (DTAG3P), is implemented, embodying the requirements above. It is tested on a range of benchmark problems. The results indicate that the method generates more regularly-structured solutions than competing methods. As a result, the system is able to scale, at least on the problem classes tested, to very complex problem instances.

Este trabalho apresenta um método para o planejamento da implantação de uma estrutura de rede de monitoramento meteorológico que utiliza o resultado do modelo numérico de previsão do tempo MBAR como parâmetro para a avaliação da qualidade das soluções candidatas. O problema é abordado com o uso dos modelos de problemas de localização discretos: o Problema de Cobertura de Conjuntos (Set Covering Location Problem) e o Problema de Cobertura Máxima (Maximal Covering Location Problem). O problema de planejamento da estrutura é dividido em duas fases. Na primeira fase, busca-se determinar a quantidade mínima necessária e localização de estações meteorológicas necessárias para atender todos os locais de demanda. Na segunda etapa, busca-se determinar a ordem de instalação das estações meteorológicas, maximizando a área de cobertura a cada nova instalação. O método proposto é detalhado, apresentando as etapas envolvidas, as informações de entrada e saída de cada etapa e as alterações necessárias para avaliação de outros algoritmos. Um conjunto de três cenários foi planejado utilizando os limites políticos do Estado do Paraná e as informações de rodovias, áreas urbanas e altimetria. O primeiro cenário avaliou-se a capacidade das abordagens utilizadas em resolver o Problema de Cobertura de Conjuntos. No segundo cenário avaliou-se a capacidade das abordagens utilizadas em resolver o Problema de Cobertura Máxima e no terceiro cenário propõe-se o Problema de Cobertura Máxima considerando a existência de uma rede de monitoramento anterior no local do experimento. Na resolução do problema foram comparadas abordagens exata e heurística. Na abordagem exata utilizou-se o método Branch & Bound para resolução do problema via Programação Linear Inteira Mista. Na abordagem heurística utilizou-se a estratégia evolutiva Evolução Diferencial. O método proposto destaca-se pela flexibilidade na substituição dos métodos utilizados na abordagem, permitindo a avaliação de outras técnicas. / This paper presents a method for planning the deployment of a network’s infrastructure for meteorological monitoring that uses the results of the numerical weather prediction MBAR as parameter for evaluating the quality of candidate solutions. The problem is addressed with the use of models for discrete location problems: the Set Covering Location Problem and Maximal Covering Location Problem. The problem of planning the structure is divided into two phases. In the first phase, we seek to determine the minimum required a mountand location of weather stations required to meet all demand. In the second step, we seek to determine the order of installation of weather stations, maximizing the coverage area for each new installation. The proposed method is detailed, presenting the steps involved, the input and output information of each step and the changes necessary to evaluate other algorithms. A set of three scenarios was designed using the political boundaries of the State of Paraná, highways, urban areas and altimetry information. The first scenario evaluates the ability of the approaches used to solve the Set Covering Location Problem. In the second scenario, we evaluated the ability of the approaches used to solve the Maximal Covering Location Problem and the third scenario proposes the Maximum Coverage Location Problem considering the previous existence of a monitoring network. For the solution of these problems it was compared exact and heuristic approaches. The exact approach used the method Branch & Bound for solving the problem via Mixed Integer Linear Programming. The heuristic approach used the evolutionary strategy Differential Evolution. The proposed method is distinguished by flexibility for substitution of the methods used, allowing the evaluation of other techniques.

A inferência gramatical lida com o problema de aprender um classificador capaz de reconhecer determinada construção ou característica em um conjunto qualquer de exemplos. Neste trabalho, um modelo de inferência gramatical baseado em uma variante de Programação Genética é proposto. A representação de cada indivíduo é baseada em uma lista ligada de árvores representando o conjunto de produções da gramática. A atuação dos operadores genéticos é feita de forma heurística. Além disto, dois novos operadores genéticos são apresentados. O primeiro, denominado Aprendizagem Incremental, é capaz de reconhecer, com base em exemplos, quais regras de produção estão faltando. O segundo, denominado Expansão, é capaz de prover a diversidade necessária. Em experimentos efetuados, o modelo proposto inferiu com sucesso seis gramáticas regulares e duas gramáticas livres de contexto: parênteses e palíndromos de quatro letras, tanto o comum quanto o disjunto, sendo superior a abordagens recentes. Atualmente, modelos de inferência gramatical têm sido aplicados a problemas de reconhecimento de sequências biológicas de DNA. Neste trabalho, dois problemas de identificação de padrão foram abordados: reconhecimento de promotores e splice-junction. Para o primeiro, o modelo proposto obteve resultado superior a outras abordagens. Para o segundo, o modelo proposto apresentou bons resultados. O modelo foi estendido para o uso de gramáticas fuzzy, mais especificamente, as gramáticas fuzzy fracionárias. Para tal, um método de estimação adequado dos valores da função de pertinência das produções da gramática é proposto. Os resultados obtidos na identificação de splice-junctions comprovam a utilidade do modelo de inferência gramatical fuzzy proposto. / Grammatical inference deals with the task of learning a classifier that can recognize a particular pattern in a set of examples. In this work, a new grammatical inference model based on a variant of Genetic Programming is proposed. In this approach, an individual is a list of structured trees representing their productions. Ordinary genetic operators are modified so as to bias the search and two new operators are proposed. The first one, called Incremental Learning, is able to recognize, based on examples, which productions are missing. The second, called Expansion is able to provide the diversity necessary to achieve convergence. In a suite of experiments performed, the proposed model successfully inferred six regular grammars and two context-free grammars: parentheses and palindromes with four letters, including the disjunct one. Results achieved were better than those obtained by recently published algorithms. Nowadays, grammatical inference has been applied to problems of recognition of biological sequences of DNA. In this work, two problems of this class were addressed: recognition of promoters and splice junction detection. In the former, the proposed model obtained results better than other published approaches. In the latter, the proposed model showed promising results. The model was extended to support fuzzy grammars, namely the fuzzy fractional grammars. Furthermore, an appropriate method of estimation of the values of the production's membership function is also proposed. Results obtained in the identification of splice junctions shows the utility of the fuzzy inference model proposed.