Catálogo de publicaciones - libros

We investigate a new category of fuzzy-neural networks-Hybrid Fuzzy Polynomial Neural Networks (HFPNN). These networks consist of genetically optimized multi-layer with two kinds of heterogeneous neurons that are fuzzy set based polynomial neurons (FSPNs) and polynomial neurons (PNs). The augmented genetically optimized HFPNN (namely gHFPNN) results in a structurally optimized structure and comes with a higher level of flexibility in comparison to the one we encounter in the conventional HFPNN. The GA-based design procedure being applied at each layer of gHFPNN leads to the selection leads to the selection of preferred nodes (FSPNs or PNs) available within the HFPNN. The performance of the gHFPNN is quantified through experimentation using a benchmarking dataset–synthetic and experimental data already experimented with in fuzzy or neurofuzzy modeling.

This note describes a method for computation of non-redundant bases of attribute implications from data tables with fuzzy attributes. Attribute implications are formulas describing particular dependencies of attributes in data. A non-redundant basis is a minimal set of attribute implications such that each attribute implication which is true in a given data (semantically) follows from the basis. Our bases are uniquely given by so-called systems of pseudo-intents. We reduce the problem of computing systems of pseudo-intents to the problem of computing maximal independent sets in certain graphs. We outline theoretical foundations, the algorithm, and present demonstrating examples.

This paper proposes a new fuzzy rule-based classifier equipped with a Bayes rule consequent. The main features of our approach are no requirement on the covariance matrices structure and their avoidance of singularity; the expansion in unimodal densities to multimodal ones; and the fuzzy set analysis for measuring the qualities of features. Two tools are exploited in constructing the proposed classifier: the iterative pruning algorithm for removing the irrelevant features and the gradient descent method for training the related parameters.

In this paper a new linear matrix inequality (LMI) based design method for T-S fuzzy classifier is proposed. The various design factors including structure of fuzzy rule and various parameters should be considered to design T-S fuzzy classifier. To determine these design factors, we describe a new and efficient two-step approach that leads to good results for classification problem. At first, LMI based fuzzy clustering is applied to obtain compact fuzzy sets in antecedent. Then consequent parameters are optimized by a LMI optimization method.

An extension of the fuzzy c -means clustering algorithm based on an adaptive distance is presented. The proposed method furnishes a fuzzy partition and a prototype for each cluster by optimizing a criterion based on an adaptive L _2 distance that changes at each algorithm iteration. Experiments with real and synthetic data sets show the usefulness of this method.

We introduce information granulation-based fuzzy systems to carry out the model identification of complex and nonlinear systems. The proposed fuzzy model implements system structure and parameter identification with the aid of genetic algorithms (GAs) and information granulation (IG). The design methodology emerges as a hybrid structural optimization and parametric optimization. IG realized with Hard C-Means (HCM) clustering help determine the initial parameters of fuzzy. And the initial parameters are tuned effectively with the aid of the GAs and the least square method (LSM). And we use GAs to identify the structure of fuzzy rules.

This paper proposes an intelligent decision support framework for an effective e-service management. The proposed framework integrates case and rule based reasonings and multi criteria decision-making techniques in fuzzy environment for a real-time decision-making, which is dealing with uncertain and imprecise decision situations. The framework potentially leads to more accurate, flexible and efficient retrieval of alternatives that are most similar and most useful to the current decision situation.

In this paper we propose an approach to integrate protein information from various data sources by defining a Protein Ontology. Protein Ontology provides the technical and scientific infrastructure and knowledge to allow description and analysis of relationships between various proteins. Protein Ontology uses relevant protein data sources of information like PDB, SCOP, and OMIM. Protein Ontology describes: Protein Sequence and Structure Information, Protein Folding Process, Cellular Functions of Proteins, Molecular Bindings internal and external to Proteins, and Constraints affecting the Final Protein Conformation. Details about Protein Ontology are available online at http://www.proteinontology.info/ .

Many researchers have successfully shown that web content mining technics and web usage mining techniques can help to find out important patterns on the content and browsing behavior in a site. However, still it is an open problem how to reach a good interpretation of the cluster results after the mining process. We propose a technique called Reverse Clustering Analysis (RCA) applied to a Self Organizing Feature Map in order to identify the most representative Web Pages of the Site. Then use this information to perform enhancements in the site. Our mining process is based on the combination of WCM and WUM to find out the content that is most interesting for the visitors. We successfully test our proposal in a real web site.

In this paper, we present a new model for the skeleton driven limb animation, which is composed of three layers: linear bones, volumetric bones and coarse volumetric control lattice. Volumetric bones are driven by linear bones using geometric method and the coarse volumetric control lattice is driven by volumetric bones using finite element method. In order to compute faster, we perform linearized simulations. The limb is embedded in the volumetric control lattice, and the surface of limb is computed using linear interpolation method. PCG solver is used to solve the large linear system of equations. We can obtain realtime simulations and realize the motions such as bend and torsion of limb.