Papers by Arturo H E R N Á N D E Z Aguirre
We propose a new approach to solve shape optimization problems based on estimation of distributio... more We propose a new approach to solve shape optimization problems based on estimation of distribution algorithms (EDA’s) combined with information from the physical problem (finite elements connectivities). Our algorithm improves exploration capacity by the regularization of the probability vectors. Therefore, the number of small holes in the structure is decreased as well as unconnected elements (elements connected at a vertex whose sides are not shared). We use a multiobjective approach to find Pareto solutions for two design goals: minimum weight and minimum displacement at some specific nodes. The solutions must fulfill three design constraints: maximum permissible Von Misses Stress, requirement of connectedness by sides in elements, and designs with small holes are not allowed.
Advances in artificial intelligence
Nova Scientia, 2014
El problema de síntesis de mecanismos consiste en determinar las dimensiones de los eslabones a f... more El problema de síntesis de mecanismos consiste en determinar las dimensiones de los eslabones a fin de que éstos puedan completar una tarea, la cual es definida por el posicionamiento en un conjunto de puntos denominados puntos de precisión. En este trabajo se aborda este problema como un ejercicio de optimización, que minimiza la distancia entre la posición de un mecanismo propuesto y los puntos de precisión, tomando como variables de decisión las dimensiones del mecanismo, su sistema coordenado relativo y parámetros que determinan la velocidad del mismo y la posición inicial. La propuesta final es un algoritmo para el diseño automatizado de mecanismos que cumplan en lo mayor posible con la tarea deseada. Se utilizó un Algoritmo de Estimación de Distribución (EDA, por sus siglas en inglés), el cual aproxima el óptimo mediante muestreos subsecuentes de una distribución Normal multivariada, donde cada muestra representa una propuesta de un mecanismo. Después, se simula y se mide la d...
Nova Scientia, 2014
Una nueva manera de modelar dependencias probabilísticas en el algoritmo de Maximización de Infor... more Una nueva manera de modelar dependencias probabilísticas en el algoritmo de Maximización de Información Mutua mediante Clústeres de Entrada (MIMIC) es presentada. Mediante cópulas es posible separar la estructura de dependencia de las distribuciones marginales en una distribución conjunta. El uso de cópulas como un mecanismo para modelar distribuciones y su aplicación a MIMIC es ilustrado en la función de prueba Rosenbrock.
This paper introduces the COPSO algorithm (Constrained Optimization via Particle Swarm Optimizati... more This paper introduces the COPSO algorithm (Constrained Optimization via Particle Swarm Optimization) for the solution of single objective constrained optimization problems. The approach includes two new perturbation operators to prevent premature convergence, and a new ring neighborhood structure. A constraint handling technique based on feasibility and sum of constraints violation, is equipped with an external file to store particles we termed "tolerant". The goal of the file is to extend the life period of those particles that otherwise would be lost after the adjustment of the tolerance of equality constraints. COPSO is applied to various engineering design problems, and for the solution of state of the art benchmark problems. Experiments show that COPSO is robust, competitive and fast.
Particle evolutionary swarm optimization with linearly decreasing ∈-tolerance
Lecture Notes in Computer Science, 2005
We introduce the PESO (Particle Evolutionary Swarm Optimization) algorithm for solving single obj... more We introduce the PESO (Particle Evolutionary Swarm Optimization) algorithm for solving single objective constrained optimization problems. PESO algorithm proposes two perturbation operators: “c-perturbation” and “m-perturbation”. The goal of these operators is to prevent premature convergence and the poor diversity issues observed in Particle Swarm Optimization (PSO) implementations. Constraint handling is based on simple feasibility rules, enhanced with a dynamic ε-tolerance
Http Dx Doi Org 10 1080 0305215031000091569, Sep 17, 2010
This paper proposes a scheme in which case-based reasoning techniques are employed to extract des... more This paper proposes a scheme in which case-based reasoning techniques are employed to extract design patterns from a genetic algorithm used to optimize ¡ Corresponding author 1 combinational circuits at the gate level. The approach seems to be able to (implicitly) rediscover several of the traditional Boolean rules used for circuit simplification and it also (implicitly) finds new simplification rules. Also, we illustrate how the approach can be used to reduce convergence times of a genetic algorithm using previously previously found solutions as cases to solve similar problems.
Use of Evolutionary Techniques to Automate the Design of Combinational Circuits
In this paper we propose an approach based on a genetic algorithm (GA) to design combinationallog... more In this paper we propose an approach based on a genetic algorithm (GA) to design combinationallogic circuits in which the objective is to minimize their total number of gates.Our results compare favorably against those produced by human designers and even anotherGA-based approach. We also briefly analyze the solutions found by the GA trying tofind some clues on how it reduces
Un Sistema Evolutivo Robusto para la Síntesis de Circuitos Analógicos
Computacion Y Sistemas, Jun 1, 2010
Gecco, 2008
An open problem in multi-objective optimization using the Pareto optimality criteria, is how to e... more An open problem in multi-objective optimization using the Pareto optimality criteria, is how to evaluate the performance of different evolutionary algorithms that solve multi-objective problems. As the output of these algorithms is a non-dominated set (NS), this problem can be reduced to evaluate what NS is better than the others based on their projection on the objective space. In this work we propose a new performance measure for the evaluation of NSs, that does not need any information a priori of the multiobjective problem. Neither it needs any parameter tuning. Besides, its evaluations of the NSs agree with intuition. Also, we introduce a benchmark of test cases to evaluate performance measures, that considers several topologies of the Pareto Front. Convergence: it refers to how close is a NS to P*. It comes directly from the definition of the Pareto's optimality. Dispersion: we are interested in the different tradeoff for the objective functions in a multiobjective problem. As the output of a multiobjective algorithm is a finite set of solutions, it is desirable to maximize the information these solutions provide. For this reason we expect a good NS to be as uniformly distributed as possible, in order to avoid zones of the Pareto Front with too many solution and zones with too few. Extension: It refers to the range of values of the objective functions. It is desirable to have information in all ranges of the objective functions. This characteristic is closely related with dispersion. Actually, dispersion and extension can be treated as a single characteristic. From now we refer to dispersion-extension as DE Convergence is related with the information in the direction of improvement of all objective functions, we mean, the direction in which all objective functions decrease their value. DE is related to information orthogonal to the direction just mentioned. In other words, DE is related with the information in all those directions in which if we improve one or more objective functions, we degrade other(s). For these reasons we consider convergence and DE as orthogonal characteristics. Convergence is considered the most important property, because it is related with "how optimal" are the points in the NS. So, it is not a surprise that a lot of the research in performance measures for NSs is focused on convergence. A classical work is that of Hansen and Jaszkiewicz [8], where they define the three following relationships between two NSs.
Proceedings of the 12th Annual Conference on Genetic and Evolutionary Computation, Jul 7, 2010
A new Estimation of Distribution Algorithm is presented. The proposed algorithm, called D-vine ED... more A new Estimation of Distribution Algorithm is presented. The proposed algorithm, called D-vine EDA, uses a graphical model which is based on pair copula decomposition. By means of copula functions it is possible to model the dependence structure in a joint distribution with marginals of different type. Thus, this paper introduces the D-vine EDA and performs experiments and statistical tests to assess the best algorithm. The set of experiments shows the potential of the D-vine EDA.
GA-based learning of kDNFsn Boolean formulas
Lecture Notes in Computer Science, 2001
Using genetic programing and multiplexers for the synthesis of logic circuits
Eng Optimiz, 2004
This article introduces the circuit design problem as a synthesis procedure. An evolutionary tech... more This article introduces the circuit design problem as a synthesis procedure. An evolutionary technique denominated Genetic Programing (GP) is proposed as the main engine for the synthesis of logic circuits. This article argues that the synthesis of circuits using bottomup procedures ...
EDITORIAL Vol. 13 No. 4 Número Especial en Sistemas Híbridos Inteligentes
Proceedings of the 10th Annual Conference on Genetic and Evolutionary Computation, 2008
The intention of the present work is to apply data mining and PSO to propose the solution of a sp... more The intention of the present work is to apply data mining and PSO to propose the solution of a specific problem about society modelling. We analyze the voting behavior and ratings of judges in a popular song contest held every year in Europe. The dataset makes it possible to analyze the determinants of success, and gives a rare opportunity to run a direct test of vote trading from logrolling. We show that they are rather driven by linguistic and cultural proximities between singers and voting countries. With this information it is possible to predict the final rank of a new country in the contest.
Extraction and reuse of design patterns from genetic algorithms using case-based reasoning
Soco, 2005
Abstract. In this paper, we propose a case-based reason-ing scheme in which we extract domain kno... more Abstract. In this paper, we propose a case-based reason-ing scheme in which we extract domain knowledge (in the form of design patterns) from a genetic algorithm used to optimize combinational logic circuits at the gate level. Such information is used in two ways: first, we show how ...
Optimal Synthesis of Mechanisms by using an EDA based on the Normal distribution
Nova Scientia, 2014
Comparing Different Serial and Parallel
ABSTRACT
Proc.~Evolutionary Multi-Criterion Optimization: Third Int'l Conference (EMO 2005)
Designing the Boltzmann Estimation of Multivariate Normal Distribution: Issues, goals and solutions
2015 IEEE Congress on Evolutionary Computation (CEC), 2015
Uploads
Papers by Arturo H E R N Á N D E Z Aguirre