李创,中山大学

Best subset selection is an important problem in regression analysis,which has many applications in computer science and medicine. However, the existing best subset selection methods have some limitations, such as strict conditions on modeling the relationship or their performances rely on the forms of models. Motivated by these problems, we propose a novel selection procedure to directly identify the best subset of predictors via distance covariance. Based on it, we develop a computational efficient algorithm that can be available to high-dimensional data with guaranteed convergence. We show that the estimator from the proposed algorithm is consistency in the sparsity selection under wild regularity conditions.