Catálogo de publicaciones - libros

In this paper a new evolutionary algorithm (EA) is described for the unconstrained Binary Quadratic Problem, which is to be used with small, medium and large scale problems as well. This method can be divided into two stages, where each stage is a steady-state EA. The first stage improves the quality of the initial population. The second stage uses concatenated, complex neighbourhood structures for the mutations and improves the quality of the solutions with a randomized k-opt local search procedure. The bit selection by mutation is based on an explicit collective memory (EC-memory) that is a modification of the flee-mutation operator ( Sebag et al. 1997 ). We tested our algorithm on all the benchmark problems of the OR-Library. Comparing the results with other heuristic methods, we can conclude that our algorithm belongs to the best methods of this problem scope.

Genetic algorithms (GA) and their capabilities for solving optimization problems have been thoroughly investigated recently in a number of application areas. These algorithms represent a class of stochastic algorithms based on methods of biological evolution. The chromosome is one of the smallest particles in the kernel of every cell and it defines the genetics of living organisms. Nature has the ability of finding suitable chromosomes through natural selection. Analogically, genetic algorithms can find optimal solutions by means of rational computational iterations [ 1 ], [ 2 ], [ 3 ], [ 4 ].

This work explores the utilization of the island-model within the context of Particle Swarm Optimization (PSO). The well-known notions of decentralized evolutionary algorithms are extended to this context, resulting in the definition of a multi-swarm. The influence that different parameterizations of the model, namely, the number of swarms, their interconnection topology, the policy for selecting particles to migrate, and the policy for accepting incoming particles is studied. Four continuous optimization problems are used for this purpose. The experimental results indicate that a moderate number of swarms arranged in a fully-connected topology provide the best results.

Directed mutation can improve the efficiency of processing many optimization problems. The directed mutation operator presented in this paper is based on the Skew-Normal Distribution. It is the first one that is not defined by case differentiation. Its expectation as well as its variance are convergent for all degrees of skewness and random number generation is simple and fast. An appropriate recombination scheme is given, and experimental results using this directed mutation are presented.

In order to extend fuzzy if-then rules bases, we propose to make use of a method which has been developed for the interpolation of crisp data — the multivariate spline interpolation. Among the various possibilities of how to accomplish the necessary generalisations, we describe here the probably simplest method: We apply spline interpolation to fuzzy data which itself is approximated by vectors of a finite-dimensional real linear space.

Many papers propound algorithms for extraction of knowledge from numerical data. But, few works have been developed for design of experiments and datum plane’s cover.

This paper describes the modelling of fuzzy rule systems using a multiresolution strategy that handles the problem of granularization of the input space by using multiresolution linguistic terms. Models of different resolutions are chained by antecedents because linguistic terms of a level j are obtained by refinements of linguistic terms of a superior level j + 1. The models can also be chained by consequents using aggregation procedures. The family of models are called Multiresolution Fuzzy Rule Systems. A metasemantics based on linguistic operators is proposed for the interpretation of the refinements as a rule specialization. Interesting models result allowing local refinement of rules that preserve the semantic interpretation.

The complete sequence of bacterial genomes provides new perspectives for the study of gene expression and gene function. DNA array experiments allow measuring the expression levels for all genes of an organism in a single hybridization experiment.

We introduce in this paper a new formulation of the regularized fuzzyc-means (FCM) algorithm which allows us to set automatic ally the actual number of clusters. The approach is based on the minimization of an objective function which mixes, via a particular parameter, a classic al FCM term and an entropy regularizer. The method uses a new exponential form of the fuzzy memberships which ensures the consistency of their bounds and makes it possible to interpret the mixing parameter as the variance (or scale) of the clusters. This variance closely related to the number of clusters, provides us with a more intuitive and an easy to set parameter. We will discuss the proposed approach from the regularization point-of-view and we will demonstrate its validity both analytic ally and experimentally. We conducted preliminary experiments both on simple toy examples as well as challenging image segmentation problems.

Time series forecasting is often done “one step ahead” with statistical models or numerical machine learning algorithms. It is possible to extend those predictive models to a few steps ahead and iterate the predictions thus allowing further forecasting. However, it is not possible to do this for thousands of data points because cumulative error tends to make the long term forecasting unreliable. Such uncertainty can be conveied by the use of fuzzy forecasting where the forecasted value is a fuzzy set rather than a number. The end-user can only appreciate the uncertainty of the forecast if the forecasting model is easy to understand. Contrary to common “black-box” models, we use symbolic machine learning on symbolic representations of time-series. In this paper, we tackle the real-world issue of forecasting electric load for one year, sampled every ten minutes, with data available for the past few years. In this context, future values are not only related to their short term previous values, but also to temporal attributes (the day of the week, holidays ...). We use a symbolic machine learning algorithm (decision tree) to extract this kind of knowledge and predict future pattern occurences. Those patterns are learnt when building a symbolic representation of the time series, by clustering episodes showing similar patterns and making the cluster a symbolic attribute of the episodes. Intra-class variations result in forecasting uncertainty that we model through fuzzy prototypes. Those prototypes are then used to construct a fuzzy forecasting easily understood by the end-user.