Catálogo de publicaciones - libros

Due to the non-stationarity of EEG signals, online training and adaptation is essential to EEG based brain-computer interface (BCI) systems. Asynchronous BCI offers more natural human-machine interaction, but it is a great challenge to train and adapt an asynchronous BCI online because the user’s control intention and timing are usually unknown. This paper proposes a novel motor imagery based asynchronous BCI for controlling a simulated robot in a specifically designed environment which is able to provide user’s control intention and timing during online experiments, so that online training and adaptation of motor imagery based asynchronous BCI can be effectively investigated. This paper also proposes an online training method, attempting to automate the process of finding the optimal parameter values of the BCI system to deal with non-stationary EEG signalsExperimental results have shown that the proposed methodfor online training of asynchronous BCI significantly improves the performance.

Fuzzy regression models has been traditionally considered as a problem of linear programming. The use of quadratic programming allows to overcome the limitations of linear programming as well as to obtain highly adaptable regression approaches. However, we verify the existence of multicollinearity in fuzzy regression and we propose a model based on Ridge regression in order to address this problem.

Group decision making with preference information on alternatives has become a very active research field over the last decade. Especially, the investigation on the group decision making problems based on different preference formats has attracted great interests from researchers recently and some approaches have been developed for dealing with these problems.However, the existing approaches can only be suitable for handling the subjective preference information. In this paper, we investigate the multiple attribute group decision making (MAGDM) problems, in which the attribute values (objective information) are given as non-negative real numbers, the information about attribute weights is to be determined, and the decision makers have their subjective preferences on alternatives. The provided subjective preference information can be represented in three well-known exact preference formats: 1) utility values; 2) fuzzy preference relations; and 3) multiplicative preference relations. We first set up three constrained optimization models integrating the given objective information and each of three preference formats respectively, and then based on these three models, we establish an integrated constrained optimization model to derive the attribute weights. The obtained attribute weights contain both the subjective preference information given by all the decision makers and the objective information. Thus, a straightforward and practical method is provided for MAGDM with multiple types of exact preference formats.

It is well known that global sustainability must begin with human actions. A reduction of the consumed energy in the heating systems is one of such possible actions. The higher the society prosperity the higher the required houses comfort, and the higher amount of energy. In Spain it is especially important as the construction rate is almost the half of that in Europe. To save energy is urgent, which means that the energy losses must be reduced.

In this paper, a multi agent system solution for the reduction of the energy consumption in heating systems of houses is presented. A control central unit (CCU) responsible of minimising the energy consumption interacts with the heaters. The CCU includes a Fuzzy Model (FM) and a Fuzzy controller (FC) and makes use of the concept of energy balance to distribute the energy between the heaters.

Results show the proposed system as a very promising solution for energy saving and comfort tracking in houses. This solution is the preliminary study to be included in a heating system product of a local company.

In this article different approximations of a local classifier algorithm are described and compared. The classification algorithm is composed by two different steps. The first one consists on the clustering of the input data by means of three different techniques, specifically a k-means algorithm, a Growing Neural Gas (GNG) and a Self-Organizing Map (SOM). The groups of data obtained are the input to the second step of the classifier, that is composed of a set of one-layer neural networks which aim is to fit a local model for each cluster. The three different approaches used in the first step are compared regarding several parameters such as its dependence on the initial state, the number of nodes employed and its performance. In order to carry out the comparative study, two artificial and three real benchmark data sets were employed.

Irreducible testors (also named typical testors) are a useful tool for feature selection in supervised classification problems with mixed incomplete data. However, the complexity of computing all irreducible testors of a training matrix has an exponential growth with respect to the number of columns in the matrix. For this reason different approaches like heuristic algorithms, parallel and distributed processing, have been developed. In this paper, we present the design and implementation of a custom architecture for BT algorithm, which allows computing testors from a given input matrix. The architectural design is based on a parallel approach that is suitable for high populated input matrixes. The architecture has been designed to deal with parallel processing of all matrix rows, automatic candidate generation, and can be configured for any size of matrix. The architecture is able to evaluate whether a feature subset is a testor of the matrix and to calculate the next candidate to be evaluated, in a single clock cycle. The architecture has been implemented on a Field Programmable Gate Array (FPGA) device. Results show that it provides significant performance improvements over a previously reported hardware implementation. Implementation results are presented and discussed.

Turing machine (TM) theory constitutes the theoretical basis for contemporary digital (von Neumann) computers. But it is problematic whether it could be an adequate theory of brain functions (computations) because, as it is widely accepted, the brain is a selectional device with blurred bounds between the areas responsible for data processing, control, and behavior. In this paper, by analogy with TMs, the optimal decoding algorithm of recent binary signal detection theory (BSDT) is presented in the form of a minimal one-dimensional abstract selectional machine (ASM). The ASM’s hypercomplexity is explicitly hypothesized, its optimal selectional and super-Turing computational performance is discussed. BSDT ASMs can contribute to a mathematically strict and biologically plausible theory of functional properties of the brain, mind/brain relations and super-Turing machines mimicking partially some cognitive abilities in animals and humans.

We present a face verification system using Parallel Gabor Principal Component Analysis (PGPCA) and fusion of Support Vector Machines (SVM) scores. The algorithm has been tested on two databases: XM2VTS (frontal images with frontal or lateral illumination) and FRAV2D (frontal images with diffuse or zenithal illumination, varying poses and occlusions). Our method outperforms others when fewer PCA coefficients are kept. It also has the lowest equal error rate (EER) in experiments using frontal images with occlusions. We have also studied the influence of wavelet frequency and orientation on the EER in a one-Gabor PCA. The high frequency wavelets are able to extract more discriminant information compared to the low frequency wavelets. Moreover, as a general rule, oblique wavelets produce a lower EER compared to horizontal or vertical wavelets. Results also suggest that the optimal wavelet orientation coincides with the illumination gradient.

The Naive Bayes Classifier is based on the (unrealistic) assumption of independence among the values of the attributes given the class value. Consequently, its effectiveness may decrease in the presence of interdependent attributes. In spite of this, in recent years, Naive Bayes classifier is worked for a privilege position due to several reasons [1]. We present DGW (Dependency Guided Wrapper), a wrapper that uses information about dependences to transform the data representation to improve the Naive Bayes classification. This paper presents experiments comparing the performance and execution time of 12 DGW variations against 12 previous approaches, as constructive induction of cartesian product attributes, and wrappers that perform a search for optimal subsets of attributes.

Experimental results show that DGW generates a new data representation that allows the Naive Bayes to obtain better accuracy more times than any other wrapper tested. DGW variations also obtain the best possible accuracy more often than the state of the art wrappers while often spending less time in the attribute subset search process.

Preference learning has recently received a lot of attention within the machine learning field, concretely learning by pairwise comparisons is a well-established technique in this field. We focus on the problem of learning the overall preference weights of a set of alternatives from the (possibly conflicting) uncertain and imprecise information given by a group of experts into the form of interval pairwise comparison matrices. Because of the complexity of real world problems, incomplete information or knowledge and different patterns of the experts, interval data provide a flexible framework to account uncertainty and imprecision. In this context, we propose a two-stage method in a distance-based framework, where the impact of the data certainty degree is captured. First, it is obtained the group preference matrix that best reflects imprecise information given by the experts. Then, the crisp preference weights and the associated ranking of the alternatives are derived from the obtained group matrix. The proposed methodology is made operational by using an Interval Goal Programming formulation.