Catálogo de publicaciones - libros

Typically the response of a multilayered perceptron (MLP) network on points which are far away from the boundary of its training data is not very reliable. When test data points are far away from the boundary of its training data, the network should not make any decision on these points. We propose a training scheme for MLPs which tries to achieve this. Our methodology trains a composite network consisting of two subnetworks : a mapping network and a vigilance network. The mapping network learns the usual input-output relation present in the data and the vigilance network learns a decision boundary and decides on which points the mapping network should respond. Though here we propose the methodology for multilayered perceptrons, the philosophy is quite general and can be used with other learning machines also.

This paper proposes the application of fault-tolerant control (FTC) using weighted fuzzy predictive control. The FTC approach is based on two steps, fault detection and isolation (FDI) and fault accommodation. Fault detection is performed by a model-based approach using fuzzy modeling. Fault isolation uses a fuzzy decision making approach. The model of the isolated fault is used in fault accommodation with a model predictive control (MPC) scheme. This paper uses a weighted fuzzy predictive control scheme, where fuzzy goals and fuzzy constraints are described in a fuzzy objective function. The criteria (goals or constraints) have an associated weight factor, which are chosen by the decision-maker. Two faults were simulated in a three tank benchmark and the respective fuzzy models were identified. The fuzzy FTC scheme proposed in this paper was able to accommodate the simulated faults.

In this paper, synchronization in coupled arrays of Cellular Neural Networks () is presented. In particular, synchronization of chaotic neural networks is obtained from complex systems theory. We consider two complex networks composed by second-order 3×4  array, and a with delay, the information interactions are defined via coupling law through of the first state of each cell. We impose the dynamics of a to of a complex network. We obtain synchronization in the complex network when the cells are globally coupled.

In the process of compound multi-agent negotiation a number of agents concurrently negotiate with one or more counterparts in order to satisfy the individual preferences that lead to the collective maximization of the overall utility function imposed on the compound service. In order to perform this task the overall utility function has to be decomposed into individual single-service utility functions. This problem is not trivial, especially in compound multi-agent negotiations involving more complex aggregation patters of negotiated issues. In this paper we propose an approach for derivation of the individual utility functions based of the principles of fuzzy set projection. We also propose a way of modifying the initially generated utility functions in the case where the agreement was not reached with those functions, what allows for reaching an agreement in repeated negotiation.

This paper presents a theory called conditional Dempster-Shafer theory (CDS) for uncertain knowledge updating. In this theory, knowledge about the value attained by an uncertain variable is modeled by a fuzzy measure and the evidence about the underlying uncertain variable is modeled by the Dempster-Shafer belief measure. Two operations in CDS are discussed in this paper, the conditioned combination rule and conditioning rule, which deal with evidence combining and knowledge updating, respectively. We show that conditioned combination rule and conditioning rule in CDS satisfy the property of Bayesian parallel combination.

In practice, classifiers are often build based on data or heuristic information. The number of potential features is usually large. One of the most important tasks in classification systems is to identify the most relevant features, because less relevant features can be interpreted as noise that reduces the classification accuracy, even for fuzzy classifiers which are somehow robust to noise. This paper proposes an ant colony optimization (ACO) algorithm for the feature selection problem. The goal is to find the set of features that reveals the best classification accuracy for a fuzzy classifier. The performance of the method is compared to other features selection methods based on tree search methods.

This paper presents the comparison results on the performance of the Artificial Bee Colony (ABC) algorithm for constrained optimization problems. The ABC algorithm has been firstly proposed for unconstrained optimization problems and showed that it has superior performance on these kind of problems. In this paper, the ABC algorithm has been extended for solving constrained optimization problems and applied to a set of constrained problems .

The distributed optimization paradigm based on Ant Colony Optimization (ACO) is a new management technique that uses the pheromone matrix to exchange information between the different subsystems to be optimized in the supply-chain. This paper proposes the use of the hybrid algorithm Beam-ACO, that fuses Beam-Search and ACO, to implement the same management concept. The Beam-ACO algorithm is used here to optimize the supplying, the distributer and the logistic agents of the supply-chain. Further, this paper implements the concept in a software platform that allows the pheromone matrix exchange through the different agents, using the TCP/IP protocol and data base systems. The results show that the distributed optimization paradigm can still be applied on supply chains where the different agents are optimized by different algorithms and that the use of the Beam-ACO in the supplying agent improves the local and the global results of the supply chain.

A cultural algorithm, together with a set of new operators for the timetabling problem(TP), is proposed in this paper. The new operators extract information about the problem during the evolutionary process, and they are combined with some previously proposed operators, in order to improve the performance of the algorithm. The proposed algorithm is tested with a benchmark of 20 instances, and compared with respect to three other algorithms: two evolutionary algorithms and a simulated annealing algorithm which won an international competition on TP.

Agents control has a broad spectrum of applications in computer science, communications and robotics. This paper focuses on formation of mobile agents, that is, configuration of points in the plane without kinematic restrictions of motion. Several goal formation strategies may be of interest. This paper summarizes control systems for achieving three basic formation structures, namely, absolute positioning with order, absolute positioning without order, and relative positioning with order. Mainly, the paper is devoted to describe each of the above schemes as well as control systems to deal with. Two of the control systems have been already reported in the literature and the remaining one is an original contribution of the paper. The control law proposed is based in the concept of associative memories. Also, simulations are presented to validate the expected behavior.