Catálogo de publicaciones - libros

In this paper an innovative approach to Spectral Pattern Recognition for multispectral images based on Genetic Programming is introduced. The problem is faced in terms of unsupervised pixel classification. Given an image consisting in bands, the goal is to find the optimal number of clusters and the positions of their centres in the -dimensional hyperspace, which allow the best possible description of the image. The pixels are then assigned to the clusters according to “minimum distance to means” principle. Furthermore the system is endowed with mechanisms able to avoid that cluster centres may be too close one another, which would favour an excessive increase in their number. As a result a good-quality clustered image is achieved. The output consists of the image divided into clusters, the proposed number of clusters, the centre coordinates and the spectral signature for any such cluster and solution fitness value. The results are compared against those achieved by another system, MultiSpec, which performs supervised classification, yet it is endowed with some features typical of an unsupervised classification system.

In this paper a system based on Genetic Programming for forecasting nonlinear time series is outlined. Our system is endowed with two features. Firstly, at any given time , it performs a τ-steps ahead prediction (i.e. it forecasts the value at time + τ) based on the set of input values for the time steps preceding . Secondly, the system automatically finds among the past input variables the most useful ones to estimate future values. The effectiveness of our approach is evaluated on El Niño 3.4 time series on the basis of a 12-month-ahead forecast.

The tuning of parameter values in parametric modelling can be viewed as an optimization problem where the outcome of the optimal model is to be as similar as possible to the experimental data. We give a general formulation of the problem with different fitness function definitions, both in terms of single-objective and multi-objective evolutionary optimization. As a test case we show results on a parametric model of combustion in a combustion chamber of a diesel engine. Such an optimal model will therefore be used for a controller design. The outputs (pressure inside combustion chamber versus rotation angle) of the resulting optimal models are compared to experimental data. Results of different optimization runs with Differential Evolution (DE) and Evolution Strategy (ES) as search algorithms and with different fitness definitions are compared.

A hybrid network, based on the integration of Fuzzy ARTMAP (FAM) and the Rectangular Basis Function Network (RecBFN), is proposed for rule learning and extraction problems. The underlying idea for such integration is that FAM operates as a classifier to cluster data samples based on similarity, while the RecBFN acts as a “compressor” to extra and refine knowledge learned by the trained FAM network. The hybrid network is capable of classifying data samples incrementally as well as of acquiring rules directly from data samples for explaining its predictions. To evaluate the effectiveness of the hybrid network, it is applied to a fault detection and diagnosis task by using a set of real sensor data collected from a Circulating Water (CW) system in a power generation plant. The rules extracted from the network are analyzed and discussed, and are found to be in agreement with experts’ opinions used in maintaining the CW system.

In this paper, several design procedures presented in the process control literature for the PID controller, based on fuzzy control systems, are reviewed.

In order to make the fuzzy logic control less dependent on the quality of the expert knowledge, four techniques for improving the fuzzy PID controllers performance, by adding some kind of adaptation feature when facing nonlinear processes, were presented.

From simulation results, it was possible to show that all four adaptive controllers had better responses than the FPID controller. Adaptive fuzzy PID controllers had a smooth response and a more conservative control action than the non-adaptive fuzzy PID controller.

As a future work, the next step is to assess the adaptive fuzzy PID system on a nonlinear experimental setup. Other fuzzy control systems combined with advanced control techniques, such as, auto-tuning, minimum variance and predictive strategies are also some future considerations.

The combination of wavelets with neural networks can hopefully remedy each others weaknesses, resulting in wavelet based neural network capable of handling system identification problems of a moderately large dimension. A wavelet based neural network is a nonlinear regression structure that represents nonlinear mappings as the superposition of dilated and translated versions of a function, which is found both in the space and frequency domains. In this paper, a wavelet-based neural network is introduced for the nonlinear identification of dynamic systems with chaotic behavior (chaotic time series). The structure of the wavelet based neural network is similar to that of radial basis function neural networks, except that here the activation function of the hidden nodes is replaced by wavelet functions. The proposed wavelet-based neural network is evaluated on two case studies: (i) the Hénon map, and (ii) the Rössler system. Simulation results demonstrate the accuracy and the reliability of the proposed identification methodology based on a wavelet based neural network.

The central component of most knowledge-based (HRM) systems is a model of the actual or required knowledge and abilities of employees, applicants and job positions. The notion of has been used in many of them to describe levels of skills and knowledge as applied to concrete work situations. Nonetheless, the imprecise nature of relationships and interactions between competences has been neglected in existing approaches. In this paper, a model for imprecise and composition relationships between competences is described, aimed at coming up with more detailed and realistic selection processes. A concrete case study is also described, illustrating how the Hr-Xml canonical format for competency definition and interchange can be extended to give support to those relationships.

The advances of mobile devices and wireless telecommunication infrastructure have allowed mobile users to conveniently transmit information through a wireless environment. To enhance the information services via the wireless transmission, the WAP Forum was founded and a new protocol was proposed to support wireless applications. However, as in the World Wide Web, the increasing information leads to the problem of information overload. One method to overcome such a problem is to model individual users to provide personalized information services. By this way, only the information predicted as user-interested can reach the end user. This paper presents a multi-agent framework in which a decision tree-based approach is employed to learn a user’s preferences. To assess the proposed framework, a mobile phone simulator is used to represent a wireless environment and a series of experiments are conducted. The experimental studies have concentrated on how to deliver appropriate information to the individual user, and on how the system can adapt to a user’s most recent preferences. The results and analysis show that based on our framework theWAPbased personalized information services can be successfully achieved.

The central component of most knowledge-based (HRM) systems is a model of the actual or required knowledge and abilities of employees, applicants and job positions. The notion of has been used in many of them to describe levels of skills and knowledge as applied to concrete work situations. Nonetheless, the imprecise nature of relationships and interactions between competences has been neglected in existing approaches. In this paper, a model for imprecise and composition relationships between competences is described, aimed at coming up with more detailed and realistic selection processes. A concrete case study is also described, illustrating how the Hr-Xml canonical format for competency definition and interchange can be extended to give support to those relationships.

In this paper we introduce UPoet, also known as the Ubiquitous Poet, the prototype of a 3D virtual character resembling a hieratical zen monk, being able to play small minimalist japanese poetic compositions, namely haiku and to chat with users or explain them haiku history and tradition.

UPoet facial animations are used to express UPoet reactions to users requests, and to convey its feelings whilst the text of the poetry is being generated.