Beatriz Prieto

Parallel multiobjective memetic RBFNNs design and feature selection for function approximation problems

The design of radial basis function neural networks (RBFNNs) still remains as a difficult task when they are applied to classification or to regression problems. The difficulty arises when the parameters that define an RBFNN have to be set, these are: the number of RBFs, the position of their centers and the length of their radii. Another issue that has to be faced when applying these models to real world applications is to select the variables that the RBFNN will use as inputs. The literature presents several methodologies to perform these two tasks separately, however, due to the intrinsic parallelism of the genetic algorithms, a parallel implementation will allow the algorithm proposed in this paper to evolve solutions for both problems at the same time. The parallelization of the algorithm not only consists in the evolution of the two problems but in the specialization of the crossover and mutation operators in order to evolve the different elements to be optimized when designing RBFNNs. The subjacent genetic algorithm is the non-sorting dominated genetic algorithm II (NSGA-II) that helps to keep a balance between the size of the network and its approximation accuracy in order to avoid overfitted networks. Another of the novelties of the proposed algorithm is the incorporation of local search algorithms in three stages of the algorithm: initialization of the population, evolution of the individuals and final optimization of the Pareto front. The initialization of the individuals is performed hybridizing clustering techniques with the mutual information (MI) theory to select the input variables. As the experiments will show, the synergy of the different paradigms and techniques combined by the presented algorithm allow to obtain very accurate models using the most significant input variables.

PCA filtering and probabilistic SOM for network intrusion detection

The growth of the Internet and, consequently, the number of interconnected computers, has exposed significant amounts of information to intruders and attackers. Firewalls aim to detect violations according to a predefined rule-set and usually block potentially dangerous incoming traffic. However, with the evolution of attack techniques, it is more difficult to distinguish anomalies from normal traffic. Different detection approaches have been proposed, including the use of machine learning techniques based on neural models such as Self-Organizing Maps (SOMs). In this paper, we present a classification approach that hybridizes statistical techniques and SOM for network anomaly detection. Thus, while Principal Component Analysis (PCA) and Fisher Discriminant Ratio (FDR) have been considered for feature selection and noise removal, Probabilistic Self-Organizing Maps (PSOM) aim to model the feature space and enable distinguishing between normal and anomalous connections. The detection capabilities of the proposed system can be modified without retraining the map, but only by modifying the units activation probabilities. This deals with fast implementations of Intrusion Detection Systems (IDS) necessary to cope with current link bandwidths.

Separation of speech signals for nonlinear mixtures

This paper shows an approach to recover original speech signals from their nonlinear mixtures. Using a geometric method that makes a piecewise linear approximation of the nonlinear mixing space, and the fact that the speech distributions are Laplacian or Gamma type, a set of slopes is obtained as a set of linear mixtures.

Visualizing the evolution of a web-based social network

Weblogs are dynamic websites updated via easy-to-use content management systems and organized as a set of chronologically ordered stories, frequently built around a link or including links to other weblogs. Since they are managed by individuals, their links tend to mirror or, in some cases, establish new types of social relations, thereby creating a social network. Studying the evolution of this network allows the discovery of emerging social structures and their growth trends. In this paper, we demonstrate the advantages of using the self-organizing maps (SOM) to visualize the evolution of a social network formed by a set of blogs, from their beginning to their current state. By observing the position a weblog is mapped to, it is easy to see what communities it belongs to nowadays, and how and when it became a part of those communities. The proposed procedure gives some insight on how communities are formed and have evolved. In this study, we apply this method to Blogalia, a blog-hosting site from which we have obtained a complete set of data and, by using SOM projections, we have drawn some conclusions on what drives the evolution of its implicit social network.

Alzheimer's Diagnosis Using Eigenbrains and Support Vector Machines

An accurate and early diagnosis of the Alzheimers Disease (AD) is of fundamental importance for the patients medical treatment. Single Photon Emission Computed Tomography (SPECT) images are commonly used by physicians to assist the diagnosis, rating them by visual evaluations. In this work we present a computer assisted diagnosis tool based on a Principal Component Analysis (PCA) dimensional reduction of the feature space approach and a Support Vector Machine (SVM) classification method for improving the AD diagnosis accuracy by means of SPECT images. The most relevant image features were selected under a PCA compression, which diagonalizes the covariance matrix, and the extracted information was used to train a SVM classifier which could classify new subjects in an unsupervised manner.

Comparing different machine learning and mathematical regression models to evaluate multiple sequence alignments

The evaluation of multiple sequence alignments (MSAs) is still an open task in bioinformatics. Current MSA scores do not agree about how alignments must be accurately evaluated. Consequently, it is not trivial to know the quality of MSAs when reference alignments are not provided. Recent scores tend to use more complex evaluations adding supplementary biological features. In this work, a set of novel regression approaches are proposed for the MSA evaluation, comparing several supervised learning and mathematical methodologies. Therefore, the following models specifically designed for regression are applied: regression trees, a bootstrap aggregation of regression trees (bagging trees), least-squares support vector machines (LS-SVMs) and Gaussian processes. These algorithms consider a heterogeneous set of biological features together with other standard MSA scores in order to predict the quality of alignments. The most relevant features are then applied to build novel score schemes for the evaluation of alignments. The proposed algorithms are validated by using the BAliBASE benchmark. Additionally, an statistical ANOVA test is performed to study the relevance of these scores considering three alignment factors. According to the obtained results, the four regression models provide accurate evaluations, even outperforming other standard scores such as BLOSUM, PAM or STRIKE.

A Competitive Neural Network for Blind Separation of Sources Based on Geometric Properties

This contribution presents a new approach to recover original signals (“sources”) from their linear mixtures, observed by the same number of sensors. The algorithm proposed assume that the input distributions are bounded and the sources generate certain combinations of boundary values. The method is simpler than other proposals and is based on geometric algebra properties. We present a neural network approach to show that with two networks, one for the separation of sources and one for weight learning, running in parallel, it is possible to efficiently recover the original signals. The learning rule is unsupervised and each computational element uses only local information.

Issues on GPU Parallel Implementation of Evolutionary High-Dimensional Multi-objective Feature Selection

The interest on applications that analyse large volumes of high dimensional data has grown recently. Many of these applications related to the so-called Big Data show different implicit parallelism that can benefit from the efficient use, in terms of performance and power consumption, of Graphics Processing Unit (GPU) accelerators. Although the GPU microarchitectures make possible the acceleration of applications by exploiting parallelism at different levels, the characteristics of their memory hierarchy and the location of GPUs as coprocessors require a careful organization of the memory access patterns and data transferences to get efficient speedups. This paper aims to take advantage of heterogeneous parallel codes on GPUs to accelerate evolutionary approaches in Electroencephalogram (EEG) classification and feature selection in the context of Brain Computer Interface (BCI) tasks. The results show the benefits of taking into account not only the data parallelism achievable by GPUs, but also the memory access patterns, in order to increase the speedups achieved by superscalar cores.

Analyzing a web-based social network using Kohonen's SOM

In this paper the utility of using the Self Organizing Maps (SOM), in conjunction with U-matrix, to visualize the evolution of a social network community formed by a set of blogs is shown. Weblogs are dynamic websites updated via easy-to-use content management systems whose links tend to mirror or in some cases establish new types of social relations, thereby creating a social network. Analyzing the evolution of this network allows the discovery of emerging social structures and their trends in growth. Here we apply this method to Blogalia, a blog hosting site from which we have a complete set of data. The proposed procedure not only gives some insight on how communities form and evolve, but would also enable to predict the future paths that their members will take.

Neural networks

This paper presents a comprehensive overview of modelling, simulation and implementation of neural networks, taking into account that two aims have emerged in this area: the improvement of our understanding of the behaviour of the nervous system and the need to find inspiration from it to build systems with the advantages provided by nature to perform certain relevant tasks. The development and evolution of different topics related to neural networks is described (simulators, implementations, and real-world applications) showing that the field has acquired maturity and consolidation, proven by its competitiveness in solving real-world problems. The paper also shows how, over time, artificial neural networks have contributed to fundamental concepts at the birth and development of other disciplines such as Computational Neuroscience, Neuro-engineering, Computational Intelligence and Machine Learning. A better understanding of the human brain is considered one of the challenges of this century, and to achieve it, as this paper goes on to describe, several important national and multinational projects and initiatives are marking the way to follow in neural-network research.