A Feature Screening Based On Variational Inequality For Fused Lasso

Feature screening method is one of the important links in the process of feature selection.By screening and removing redundant features in the data,the complex-ity of the model can be greatly simplified and the interpretability of the model can be improved.Aiming at a class of Fused Lasso models,this paper constructs a feature selection rule by using the necessary conditions of variational inequality for dual problems.First,this paper introduces the Lasso model and Fused Lasso model.The dual problem of the optimization problem is given by using the con-vex optimization theory,and the necessary conditions for the variational inequality form of the dual problem are derived.Secondly,a compact dual feasible region is constructed to contain the dual optimal solution.By estimating the upper bound of the dual constraint on this feasible region,establish a screening rule to find adjacent features with the same coefficient,and then achieve feature elimination.Finally,the application of the variable inequality in Fused Lasso model is compared with the traditional feature selection method.It shows that the method used in this paper is more accurate and effective.