Catálogo de publicaciones - libros

This paper presents ParadisEO-MOEO, a white-box object-oriented generic framework dedicated to the flexible design of evolutionary multi-objective algorithms. This paradigm-free software embeds some features and techniques for Pareto-based resolution and aims to provide a set of classes allowing to ease and speed up the development of computationally efficient programs. It is based on a clear conceptual distinction between the solution methods and the multi-objective problems they are intended to solve. This separation confers a maximum design and code reuse. ParadisEO-MOEO provides a broad range of archive-related features (such as elitism or performance metrics) and the most common Pareto-based fitness assignment strategies (MOGA, NSGA, SPEA, IBEA and more). Furthermore, parallel and distributed models as well as hybridization mechanisms can be applied to an algorithm designed within ParadisEO-MOEO using the whole version of ParadisEO. In addition, GUIMOO, a platform-independant free software dedicated to results analysis for multi-objective problems, is briefly introduced.

The inadequacy of classical methods to handle (RAPs) draw the attention of evolutionary algorithms (EAs) to these problems. The potentialities of EAs are exploited in the present work for handling two such RAPs of quite different natures, namely (1) university class timetabling problem and (2) land-use management problem. In many cases, these problems are over-simplified by ignoring many important aspects, such as different types of constraints and multiple objective functions. In the present work, two EA-based multi-objective optimizers are developed for handling these two problems by considering various aspects that are common to most of their variants. Finally, the similarities between the problems, and also between their solution techniques, are analyzed through the application of the developed optimizers on two real problems.

Multi-Objective Evolutionary Algorithms (MOEA) have been succesfully applied to solve control problems. However, many improvements are still to be accomplished. In this paper a new approach is proposed: the Multi-Objective Pole Placement with Evolutionary Algorithms (MOPPEA). The design method is based upon using complex-valued chromosomes that contain information about closed-loop poles, which are then placed through an output feedback controller. Specific cross-over and mutation operators were implemented in simple but efficient ways. The performance is tested on a mixed multi-objective / control problem.

The phylogeny reconstruction problem consists of determining the most accurate tree that represents evolutionary relationships among species. Different criteria have been employed to evaluate possible solutions in order to guide a search algorithm towards the best tree. However, these criteria may lead to distinct phylogenies, which are often conflicting among them. In this context, a multi-objective approach can be useful since it could produce a spectrum of equally optimal trees (Pareto front) according to all criteria. We propose a multi-objective evolutionary algorithm, named PhyloMOEA, which employs the maximum parsimony and likelihood criteria to evaluate solutions. PhyloMOEA was tested using four datasets of nucleotide sequences. This algorithm found, for all datasets, a Pareto front representing a trade-off between the criteria. Moreover, SH-test showed that most of solutions have scores similar to those obtained by phylogenetic programs using one criterion.

A perturbation method is proposed to detect convergence of the Pareto front for multi-objective algorithms and to investigate its effect on the rate of convergence of the optimization. Conventionally, evolutionary algorithms are allowed to run for a fixed number of trial solutions which can result in a premature convergence or in an unnecessary number of calls to a computationally intensive real world problem. Combination of evolutionary multi-objective algorithms with perturbation method will improve the rate of convergence of the optimization. This is a very important characteristic in reducing number of generations and therefore reducing the computational time which is important in real world problems where cost and time constraint prohibit repeated runs of the algorithm and the simulation. The performance of the method will be examined by its application to two water distribution networks from literature. The results will be compared with previously published results from literature and those generated by evolutionary multi-objective algorithm. It will be shown that the method is able to find the Pareto optimal front with less computational effort.

The importance of multi-objective optimization is globably established nowadays. Furthermore, a great part of real-world problems are subject to uncertainties due to, , noisy or approximated fitness function(s), varying parameters or dynamic environments. Moreover, although evolutionary algorithms are commonly used to solve multi-objective problems on the one hand and to solve stochastic problems on the other hand, very few approaches combine simultaneously these two aspects. Thus, flow-shop scheduling problems are generally studied in a single-objective deterministic way whereas they are, by nature, multi-objective and are subjected to a wide range of uncertainties. However, these two features have never been investigated at the same time.

In this paper, we present and adopt a proactive stochastic approach where processing times are represented by random variables. Then, we propose several multi-objective methods that are able to handle any type of probability distribution. Finally, we experiment these methods on a stochastic bi-objective flow-shop problem.

This paper presents the corrective process control system for achieving a target quality level in glass melting processes. Since automated data collection devices would monitor and log process attributes that are assumed to correlate to a quality level in the glass melting process, appropriate process control logics utilizing the collected data are definitely needed. In this paper, an evolutionary algorithm based search logic is newly proposed. The objective of the proposed logic is to find the best process condition composed of the process attributes which can generate the target quality level. The proposed logic tries to find the best process condition that needs to satisfy the following two criteria: 1) a process condition should require minimal changes from the current setting of the process attributes; and 2) a process condition can generate the exact or closest value against the target quality level. A case study and a developed process control system are presented.

This paper presents a multicriterion algorithm for dealing with joint facility location and network design problems, formulated as bi-objective problems. The algorithm is composed of two modules: a multiobjective quasi-Newton algorithm, that is used to find the location of the facilities; and a multiobjective genetic algorithm, which is responsible for finding the efficient topologies. These modules are executed in an iterative way, to make the estimation of whole Pareto set possible. The algorithm has been applied to the expansion of a real energy distribution system. The minimization of financial cost and the maximization of reliability have been considered as the design objectives in this case.

In this paper we present three path relinking approaches for solving a bi-objective permutation flowshop problem. The path relinking phase is initialized by optimizing the two objectives using Ant Colony System. The initiating and guiding solutions of path relinking are randomly selected and some of the solutions along the path are intensified using local search. The three approaches differ in their strategy of defining the heuristic bounds for the local search, i.e., each approach allows its solutions to undergo local search under different conditions. These conditions are based on local nadir points. Several test instances are used to investigate the performances of the different approaches. Computational results show that the decision which allows solutions to undergo local search has an influence in the performance of path relinking. We also demonstrate that path relinking generates competitive results compared to the best known solutions of the test instances.

Multi-objective metaheuristics have previously been applied to partial classification, where the objective is to produce simple, easy to understand rules that describe subsets of a class of interest. While this provides a useful aid in descriptive data mining, it is difficult to see how the rules produced can be combined usefully to make a predictive classifier. This paper describes how, by using a more complex representation of the rules, it is possible to produce effective classifiers for two class problems. Furthermore, through the use of multi-objective genetic programming, the user can be provided with a selection of classifiers providing different trade-offs between the misclassification costs and the overall model complexity.