l1-norm quantile regression screening rule via the dual circumscribed sphere

摘要: l1-norm quantile regression is a common choice if there exists outlier or heavy-tailed error in high-dimensional data sets. However, it computationally expensive to solve this problem when the feature size of ultra high. As far as we know, existing screening rules can not speed up computation regression, which dues non-differentiability function/pinball loss. In paper, introduce dual circumscribed sphere technique and propose novel rule. Our rule expressed closed-form function given eliminates inactive features with low computational cost. Numerical experiments on some simulation real sets show that be used eliminate almost all features. Moreover, help reduce 23 times time, compared without our