Variable Selection and Regularization via Arbitrary Rectangle-range Generalized Elastic Net

TL;DR

This paper introduces ARGEN, a new regularization method for constrained variable selection in high-dimensional models, with proven consistency, an efficient algorithm, and demonstrated superior performance over elastic net in simulations and real-world stock tracking.

Contribution

The paper proposes ARGEN, a novel generalized elastic net penalty with arbitrary rectangle constraints, including theoretical properties, an efficient solving algorithm, and practical applications.

Findings

ARGEN outperforms elastic net in simulation studies.

The algorithm MU-QP-RR-W-$l_1$ efficiently solves ARGEN.

Application to stock index tracking demonstrates ARGEN's practical utility.

Abstract

We introduce the arbitrary rectangle-range generalized elastic net penalty method, abbreviated to ARGEN, for performing constrained variable selection and regularization in high-dimensional sparse linear models. As a natural extension of the nonnegative elastic net penalty method, ARGEN is proved to have variable selection consistency and estimation consistency under some conditions. The asymptotic behavior in distribution of the ARGEN estimators have been studied. We also propose an algorithm called MU-QP-RR-W-$l_1$ to efficiently solve ARGEN. By conducting simulation study we show that ARGEN outperforms the elastic net in a number of settings. Finally an application of S&P 500 index tracking with constraints on the stock allocations is performed to provide general guidance for adapting ARGEN to solve real-world problems.