The NFDA-Nonsmooth Feasible Directions Algorithm applied to construction of Pareto Fronts of Ridge and Lasso Regressions

Ridge and Lasso regressions are types of linear regression, a machine learning tool for dealing with data. Based on multiobjective optimization theory, we transform Ridge and Lasso regression into bi-objective optimization problems. The Pareto fronts of the resulting problems provide a range of regression models from which the best one can be selected. We employ the NFDA-Nonsmooth Feasible Directions Algorithm devised for solving convex optimization problems to construct the Pareto fronts of Ridge and Lasso when regarded as bi-objective problems.