Knowledge extraction from a mixed transfer function artificial neural network

One of the big problems with Artificial Neural Networks (ANN) is that their results are not intuitively clear. For example, if we use the traditional neurons, with a sigmoid activation function, we can approximate any function, including linear functions, but the coefficients (weights) in this approximation will be rather meaningless. To resolve this problem, this paper presents a novel kind of ANN with different transfer functions mixed together. The aim of such a network is to i) obtain a better generalization than current networks ii) to obtain knowledge from the networks without a sophisticated knowledge extraction algorithm iii) to increase the understanding and acceptance of ANNs. Transfer Complexity Ratio is defined to make a sense of the weights associated with the network. The paper begins with a review of the knowledge extraction from ANNs and then presents a Mixed Transfer Function Artificial Neural Network (MTFANN). A MTFANN contains different transfer functions mixed together rather than mono-transfer functions. This mixed presence has helped to obtain high level knowledge and similar generalization comparatively to monotransfer function nets in a global optimization context.

Review of applications of the Cutting Angle methods

The theory of abstract convexity provides us with the necessary tools for building accurate one-sided approximations of functions. Cutting angle methods have recently emerged as a tool for global optimization of families of abstract convex functions. Their applicability have been subsequently extended to other problems, such as scattered data interpolation. This paper reviews three different applications of cutting angle methods, namely global optimization, generation of nonuniform random variates and multivatiate interpolation.

Introduction to evolutionary computing in system design

In this chapter, an introduction on the use of evolutionary computing techniques, which are considered as global optimization and search techniques inspired from biological evolutions, in the domain of system design is presented. A variety of evolutionary computing techniques are first explained, and the motivations of using evolutionary computing techniques in tackling system design tasks are then discussed. In addition, a number of successful applications of evolutionary computing to system design tasks are described.

Characterizing compactness of geometrical clusters using fuzzy measures

Certain tasks in image processing require the preservation of fine image details, while applying a broad operation to the image, such as image reduction, filtering, or smoothing. In such cases, the objects of interest are typically represented by small, spatially cohesive clusters of pixels which are to be preserved or removed, depending on the requirements. When images are corrupted by the noise or contain intensity variations generated by imaging sensors, identification of these clusters within the intensity space is problematic as they are corrupted by outliers. This paper presents a novel approach to accounting for spatial organization of the pixels and to measuring the compactness of pixel clusters based on the construction of fuzzy measures with specific properties: monotonicity with respect to the cluster size; invariance with respect to translation, reflection, and rotation; and discrimination between pixel sets of fixed cardinality with different spatial arrangements. We present construction methods based on Sugeno-type fuzzy measures, minimum spanning trees, and fuzzy measure decomposition. We demonstrate their application to generating fuzzy measures on real and artificial images.

Algoritmos para o problema de localização de uma facilidade com distâncias limitadas e restrições de atendimento

The objective in the facility location problem with limited distances is to minimize the sum of distance functions from the facility to the customers, but with a limit on each distance, after which the corresponding function becomes constant. The problem has applications in situations where the service provided by the facility is insensitive after a given threshold distance (eg. fire station location). In this work, we propose a global optimization algorithm for the case in which there are lower and upper limits on the numbers of customers that can be served

Algoritmos genéticos: uso de lógica nebulosa e análise de convergência por cadeia de Markov

In this work, the Markov chain will be the tool used in the modeling and analysis of convergence of the genetic algorithm, both the standard version as for the other versions that allows the genetic algorithm. In addition, we intend to compare the performance of the standard version with the fuzzy version, believing that this version gives the genetic algorithm a great ability to find a global optimum, own the global optimization algorithms. The choice of this algorithm is due to the fact that it has become, over the past thirty yares, one of the more importan tool used to find a solution of de optimization problem. This choice is due to its effectiveness in finding a good quality solution to the problem, considering that the knowledge of a good quality solution becomes acceptable given that there may not be another algorithm able to get the optimal solution for many of these problems. However, this algorithm can be set, taking into account, that it is not only dependent on how the problem is represented as but also some of the operators are defined, to the standard version of this, when the parameters are kept fixed, to their versions with variables parameters. Therefore to achieve good performance with the aforementioned algorithm is necessary that it has an adequate criterion in the choice of its parameters, especially the rate of mutation and crossover rate or even the size of the population. It is important to remember that those implementations in which parameters are kept fixed throughout the execution, the modeling algorithm by Markov chain results in a homogeneous chain and when it allows the variation of parameters during the execution, the Markov chain that models becomes be non - homogeneous. Therefore, in an attempt to improve the algorithm performance, few studies have tried to make the setting of the parameters through strategies that capture the intrinsic characteristics of the problem. These characteristics are extracted from the present state of execution, in order to identify and preserve a pattern related to a solution of good quality and at the same time that standard discarding of low quality. Strategies for feature extraction can either use precise techniques as fuzzy techniques, in the latter case being made through a fuzzy controller. A Markov chain is used for modeling and convergence analysis of the algorithm, both in its standard version as for the other. In order to evaluate the performance of a non-homogeneous algorithm tests will be applied to compare the standard fuzzy algorithm with the genetic algorithm, and the rate of change adjusted by a fuzzy controller. To do so, pick up optimization problems whose number of solutions varies exponentially with the number of variables

The cross-entropy method and its application to minimize the ripple of magnetic levitation forces of a maglev system

This paper explores firstly the potential of a new evolutionary method - the Cross-Entropy (CE) method in solving continuous inverse electromagnetic problems. For this purpose, an adaptive updating formula for the smoothing parameter, some mutation operation, and a new termination criterion are proposed. The proposed CE based metaheuristics is applied to reduce the ripple of the magnetic levitation forces of a prototype Maglev system. The numerical results have shown that the ripple of the magnetic levitation forces of the prototype system is reduced significantly after the design optimization using the proposed algorithm.

Predicting structural models for silicon clusters

This article introduces an efficient method to generate structural models for medium-sized silicon clusters. Geometrical information obtained from previous investigations of small clusters is initially sorted and then introduced into our predictor algorithm in order to generate structural models for large clusters. The method predicts geometries whose binding energies are close (95%) to the corresponding value for the ground-state with very low computational cost. These predictions can be used as a very good initial guess for any global optimization algorithm. As a test case, information from clusters up to 14 atoms was used to predict good models for silicon clusters up to 20 atoms. We believe that the new algorithm may enhance the performance of most optimization methods whenever some previous information is available. (C) 2003 Wiley Periodicals, Inc.

A new implementation of population based incremental learning method for optimizations in electromagnetics

To enhance the global search ability of population based incremental learning (PBIL) methods, it is proposed that multiple probability vectors are to be included on available PBIL algorithms. The strategy for updating those probability vectors and the negative learning and mutation operators are thus re-defined correspondingly. Moreover, to strike the best tradeoff between exploration and exploitation searches, an adaptive updating strategy for the learning rate is designed. Numerical examples are reported to demonstrate the pros and cons of the newly implemented algorithm.

New optimality conditions for nonsmooth control problems

This work considers nonsmooth optimal control problems and provides two new sufficient conditions of optimality. The first condition involves the Lagrange multipliers while the second does not. We show that under the first new condition all processes satisfying the Pontryagin Maximum Principle (called MP-processes) are optimal. Conversely, we prove that optimal control problems in which every MP-process is optimal necessarily obey our first optimality condition. The second condition is more natural, but it is only applicable to normal problems and the converse holds just for smooth problems. Nevertheless, it is proved that for the class of normal smooth optimal control problems the two conditions are equivalent. Some examples illustrating the features of these sufficient concepts are presented. © 2012 Springer Science+Business Media New York.

Extração semi-automática do eixo de rodovia em imagens de média e alta resolução usando programação dinâmica

Pós-graduação em Ciências Cartográficas - FCT

Manobras ótimas de atitude de satélites artificiais utilizando algoritmos genéticos

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Empilhamento pelo método superfície de reflexão comum 2-D com topografia e introdução ao caso 3-D

O método de empilhamento sísmico CRS simula seções sísmicas ZO a partir de dados de cobertura múltipla, independente do macro-modelo de velocidades. Para meios 2-D, a função tempo de trânsito de empilhamento depende de três parâmetros, a saber: do ângulo de emergência do raio de reflexão normal (em relação à normal da superfície) e das curvaturas das frentes de onda relacionadas às ondas hipotéticas, denominadas NIP e Normal. O empilhamento CRS consiste na soma das amplitudes dos traços sísmicos em dados de múltipla cobertura, ao longo da superfície definida pela função tempo de trânsito do empilhamento CRS, que melhor se ajusta aos dados. O resultado do empilhamento CRS é assinalado a pontos de uma malha pré-definida na seção ZO. Como resultado tem-se a simulação de uma seção sísmica ZO. Isto significa que para cada ponto da seção ZO deve-se estimar o trio de parâmetros ótimos que produz a máxima coerência entre os eventos de reflexão sísmica. Nesta Tese apresenta-se fórmulas para o método CRS 2-D e para a velocidade NMO, que consideram a topografia da superfície de medição. O algoritmo é baseado na estratégia de otimização dos parâmetros de fórmula CRS através de um processo em três etapas: 1) Busca dos parâmetros, o ângulo de emergência e a curvatura da onda NIP, aplicando uma otimização global, 2) busca de um parâmetro, a curvatura da onda N, aplicando uma otimização global, e 3) busca de três parâmetros aplicando uma otimização local para refinar os parâmetros estimados nas etapas anteriores. Na primeira e segunda etapas é usado o algoritmo Simulated Annealing (SA) e na terceira etapa é usado o algoritmo Variable Metric (VM). Para o caso de uma superfície de medição com variações topográficas suaves, foi considerada a curvatura desta superfície no algoritmo do método de empilhamento CRS 2-D, com aplicação a dados sintéticos. O resultado foi uma seção ZO simulada, de alta qualidade ao ser comparada com a seção ZO obtida por modelamento direto, com uma alta razão sinal-ruído, além da estimativa do trio de parâmetros da função tempo de trânsito. Foi realizada uma nálise de sensibilidade para a nova função de tempo de trânsito CRS em relação à curvatura da superfície de medição. Os resultados demonstraram que a função tempo de trânsito CRS é mais sensível nos pontos-médios afastados do ponto central e para grandes afastamentos. As expressões da velocidade NMO apresentadas foram aplicadas para estimar as velocidades e as profundidades dos refletores para um modelo 2-D com topografia suave. Para a inversão destas velocidades e profundidades dos refletores, foi considerado o algoritmo de inversão tipo Dix. A velocidade NMO para uma superfície de medição curva, permite estimar muito melhor estas velocidades e profundidades dos refletores, que as velocidades NMO referidas as superfícies planas. Também apresenta-se uma abordagem do empilhamento CRS no caso 3-D. neste caso a função tempo de trânsito depende de oito parâmetros. São abordadas cinco estratégias de busca destes parâmetros. A combinação de duas destas estratégias (estratégias das três aproximações dos tempos de trânsito e a estratégia das configurações e curvaturas arbitrárias) foi aplicada exitosamente no empilhamento CRS 3-D de dados sintéticos e reais.

Sintonia de controlador fuzzy por algoritmo genético em sistema de nível de líquidos

O presente trabalho demonstra a aplicação de um Algoritmo Genético com o intuito de projetar um controlador Fuzzy MISO, através da sintonia de seus parâmetros, em um processo experimental de nivelamento de líquido em um tanque, cuja dinâmica apresenta características não-lineares. Para o projeto e sintonia do controlador, foi utilizado o suporte do software Matlab, e seus pacotes Simulink e Global Optimization Toolbox. O Controlador Fuzzy ora projetado teve seu desempenho avaliado através de ensaios em tempo real em um Sistema de Nível de Liquido.

Investigating Smart Sampling as a population initialization method for Differential Evolution in continuous problems

Recently, researches have shown that the performance of metaheuristics can be affected by population initialization. Opposition-based Differential Evolution (ODE), Quasi-Oppositional Differential Evolution (QODE), and Uniform-Quasi-Opposition Differential Evolution (UQODE) are three state-of-the-art methods that improve the performance of the Differential Evolution algorithm based on population initialization and different search strategies. In a different approach to achieve similar results, this paper presents a technique to discover promising regions in a continuous search-space of an optimization problem. Using machine-learning techniques, the algorithm named Smart Sampling (SS) finds regions with high possibility of containing a global optimum. Next, a metaheuristic can be initialized inside each region to find that optimum. SS and DE were combined (originating the SSDE algorithm) to evaluate our approach, and experiments were conducted in the same set of benchmark functions used by ODE, QODE and UQODE authors. Results have shown that the total number of function evaluations required by DE to reach the global optimum can be significantly reduced and that the success rate improves if SS is employed first. Such results are also in consonance with results from the literature, stating the importance of an adequate starting population. Moreover, SS presents better efficacy to find initial populations of superior quality when compared to the other three algorithms that employ oppositional learning. Finally and most important, the SS performance in finding promising regions is independent of the employed metaheuristic with which SS is combined, making SS suitable to improve the performance of a large variety of optimization techniques. (C) 2012 Elsevier Inc. All rights reserved.