No abstract

Abstract. In the design of support vector machines an important step is to select the optimal hyperparameters. One of the most used estimators of the performance is the Radius-Margin bound. Some modifications of this bound have been made to adapt it to soft margin problems, giving a convex optimization problem for the L2 soft margin formulation. However, it is still interesting to consider the L1 case due to the reduction in the support vector number. There have been some proposals to adapt the Radius-Margin bound to the L1 case, but the use of gradient descent to test them is not possible in some of them because these bounds are not differentiable. In this work we propose to use simulated annealing as a method to find the optimal hyperparameters when the bounds are not differentiable, have multiple local minima or the kernel is not differentiable with respect to its hyperparameters.

Abstract. Support vector machines, especially when using radial basis kernels, have given good results in the classification of different volatile compounds. We can achieve a feature extraction method adjusting the parameters of a modified radial basis kernel, giving more importance to those features that are important for classification proposes. However, the function that has to be minimized to find the best scaling factors is not derivable and has multiple local minima. In this work we propose to adapt the ideas of the ant colony optimization method to find an optimal value of the kernel parameters.

Abstract. The use of support vector machines for multi-category problems is still an open field to research. Most of the published works use the one-against-rest strategy, but with a one-against-one approach results can be improved. To avoid testing with all the binary classifiers there are some methods like the Decision Directed Acyclic Graph based on a decision tree. In this work we propose an optimization method to improve the performance of the binary classifiers using Particle Swarm Optimization and an automatic method to build the graph that improves the average number of operations needed in the test phase. Results show a good behavior when both ideas are used.

Electroencephalographic analysis techniques have become a very useful tool to assess brain activity and interactions between cerebral regions, that is, the so-called cerebral connectivity analysis. The effects of some drugs have, so far, been studied using spectral analysis and, to a lesser extent, some linear and nonlinear connectivity techniques. New indexes have recently been designed based on assumptions that make them more robust against volume conduction effects that could yield to spurious connectivity results.\ud These new indexes such as the imaginary coherence (IC) [Nolte et al., 2004], the phase-lag index (PLI) [Stam et al., 2007] and the weighted phase-lag index (WPLI) [Vinck et al., 2011] have proven very useful in several fields, for example in characterizing electroencephalographic (EEG) and magnetoencephalographic (MEG) activity of Alzheimer’s Disease patients compared to healthy controls.\ud However, these techniques have not been applied to study the effect of drugs on the brain. The main purpose of the current work was to assess the suitability and effectiveness of these innovative indexes to study the brain connectivity under psychoactive drug treatment, and concretely, the effects of a single dose of alprazolam, a short-acting drug of the benzodiazepine family. \ud Alprazolam is extensively prescribed for the treatment of anxiety and panic disorders, and peak plasma concentrations are obtained between 0.5 and 2 hours after intake [Greenblat and Wright, 1993]. Being a benzodiazepines, alprazolam induces an enhancement of the inhibitory pathways through their activity on the GABA A receptor complex, favouring the entrance to chloride ions into the neurons [Haefely, 1990]. Due to the enhancement of the inhibitory pathways a weakening or even an impairment of functional connectivity could be hypothesizedPostprint (author's final draft