Robust feature selection for high-dimensional datasets using a GPU-accelerated ensemble of cooperative coevolutionary optimizers

Feature selection is an increasingly important step in the application of machine learning and knowledge discovery techniques to high-dimensional datasets. However, the growing complexity and size of datasets have made feature selection increasingly challenging, as selecting an optimal subset of features can be computationally very expensive, especially when a robust solution is required. To address this issue, we present an approach based on ensembles of cooperative coevolutionary optimisers and its parallelisation for hybrid multi-core CPU and GPU computation. The application of the developed algorithm to some typical high-dimensional datasets is discussed in the paper. According to the preliminary results, the proposed framework represents a valuable tool for addressing the computational challenges faced in feature selection, and it can be potentially applied to a wide range of machine learning and knowledge discovery tasks.