RaSE: A Variable Screening Framework via Random Subspace Ensembles

Variable screening methods have been shown to be effective in dimension reduction under the ultra-high dimensional setting. Most existing are designed rank predictors according their individual contributions response. As a result, variables that marginally independent but jointly dependent with response could missed. In this work, we propose new framework for variable screening, random subspace ensemble (RaSE), which works by evaluating quality of subspaces may cover multiple predictors. This can naturally combined any evaluation criterion, leads an array methods. The is capable identify signals no marginal effect or high-order interaction effects. It enjoy sure property and consistency. We also develop iterative version RaSE theoretical support. Extensive simulation studies real-data analysis show effectiveness framework.