Robust feature selection via simultaneous capped ¿¿¿2-norm and ¿¿¿2,1-norm minimization

摘要

High dimension is one of the key characters of big data. Feature selection plays a significant role in many machine learning applications dealing with high-dimensional data. To improve the robustness of feature selection, we propose a new robust feature selection method with emphasizing Simultaneous Capped ¿¿¿2-norm loss and ¿¿¿2, 1-norm regularizer Minimization (SCM). The capped ¿¿¿2-norm based loss function can effectively eliminate the influence of noise and outliers in regression and the ¿¿¿2, 1-norm regularization is used to select features across data sets with joint sparsity. Meanwhile, we propose an effective approach to solve the formulated minimization problem. Experimental studies on real-world data sets demonstrate the effectiveness of our method in comparison with other popular feature selection methods.