Fernando Rojas Ruiz

Hybridizing Genetic Algorithms with ICA in Higher Dimension

In this paper we present a novel method for blindly separating unobservable independent component signals from their linear mixtures, using genetic algorithms (GA) to minimize the nonconvex and nonlinear cost functions. This approach is very useful in many fields such as forecasting indexes in financial stock markets where the search for independent components is the major task to include exogenous information into the learning machine. The GA presented in this work is able to extract independent components with faster rate than the previous independent component analysis algorithms based on Higher Order Statistics (HOS) as input space dimension increases showing significant accuracy and robustness.

Evaluation of electro-oculography data for ataxia SCA-2 classification: A blind source separation approach

Spino Cerebellar Ataxia type 2 is an autosomal dominant cerebellar hereditary ataxia with the highest prevalence in Cuba. Typical symptoms in patients of SCA2 ataxia include modifications in latency, peak velocity, and deviation in visual saccadic movements. After applying some electro-oculography based tests to both healthy and SCA2 afflicted individuals, differences in saccade morphology were found, mainly for 60º of stimulus amplitude. In this work, independent component analysis (ICA) is used to decompose the pulse and step components for the control of saccadic ocular movements, finding significant differences in time and amplitude ratio of both components. The obtained results point out the validity of the proposed methodology in order to study the effects of ataxia SCA2 in the oculomotor system.

Geometrical ICA-Based Method for Blind Separation of Super-Gaussian Signals

This work explains a new method for blind separation of a linear mixture of sources, based on geometrical considerations concerning the observation space. This new method is applied to a mixture of several sources and it obtains the estimated coefficients of the unknown mixture matrix A and separates the unknown sources. In this work, the principles of the new method and a description of the algorithm are shown.

Plugging an Histogram-Based Contrast Function on a Genetic Algorithm for Solving PostNonLinear-BSS

This paper proposes a novel Independent Component Analysis algorithm based on the use of a genetic algorithm intended for its application to the problem of blind source separation on post-nonlinear mixtures. We present a simple though effective contrast function which evaluates individuals of each population (candidate solutions) based on estimating the probability densities of the outputs through histogram approximation. Although more sophisticate methods for probability density function approximation exist, such as kernel-based methods or k-nearest-neighbor estimation, the histogram presents the advantage of its simplicity and easy calculation if an appropriate number of samples is available.