Feature Selection For Weighted Equal Group Lasso

The Weighted Equal Group Lasso model is a effective regression technique by combining the L2,1 and L1 norms,the sparseness of coefficient differences within and between groups in the solution of the optimization problem is constrained.When processing high-dimensional data,due to the complexity of the regularizers of the model,the computational cost is too expensive.In view of the effectiveness of the TLFre method on the Sparse Group Lasso problem,this paper applies the TLFre method on the Weighted Equal Group Lasso model.Based on Fenchel's conjugate duality theory,it gives the more complex duality problem caused by the specific difference weighted matrices.Depending on the results in convex analysis and variational analysis,the dual feasible set would be decomposed.Finally,a concise and effective feature selection rule is obtained.