Leveraging cooperation for parallel multi-objective feature selection in high-dimensional EEG data

摘要

Bioinformatics applications frequently involve high-dimensional model building or classification problemsrnthat require reducing dimensionality to improve learning accuracy while irrelevant inputs are removed.rnThus, feature selection has become an important issue on these applications. Moreover, several approachesrnfor supervised and unsupervised feature selections as a multi-objective optimization problem have beenrnrecently proposed to cope with issues on performance evaluation of classifiers and models. As parallel processingrnconstitutes an important tool to reach efficient approaches that make it possible to tackle complexrnproblems within reasonable computing times, in this paper, alternatives for the cooperation of subpopulationsrnin multi-objective evolutionary algorithms have been identified and classified, and several proceduresrnhave been implemented and evaluated on some synthetic and Brain–Computer Interface datasets. The resultsrnshow different improvements achieved in the solution quality and speedups, depending on the cooperationrnalternative and dataset. We show alternatives that even provide superlinear speedups with only smallrnreductions in the solution quality, besides another cooperation alternative that improves the quality of thernsolutions with speedups similar to, or only slightly higher than, the speedup obtained by the parallel fitnessrnevaluation in a master-worker implementation (the alternative used as reference that behaves as therncorresponding sequential multi-objective approach).