Parameter Estimation with the Ordered l2 Regularization via an Alternating Direction Method of Multipliers

Regularization is a popular technique in machine learning for model estimation and for avoiding overfitting. Prior studies have found that modern ordered regularization can be more effective in handling highly correlated, high-dimensional data than traditional regularization. The reason stems from the fact that the ordered regularization can reject irrelevant variables and yield an accurate estimation of the parameters. How to scale up the ordered regularization problems when facing large-scale training data remains an unanswered question. This paper explores the problem of parameter estimation with the ordered <mml:semantics>l2</mml:semantics>-regularization via Alternating Direction Method of Multipliers (ADMM), called ADMM-O<mml:semantics>l2</mml:semantics>. The advantages of ADMM-O<mml:semantics>l2</mml:semantics> include (i) scaling up the ordered <mml:semantics>l2</mml:semantics> to a large-scale dataset, (ii) predicting parameters correctly by excluding irrelevant variables automatically, and (iii) having a fast convergence rate. Experimental results on both synthetic data and real data indicate that ADMM-O<mml:semantics>l2</mml:semantics> can perform better than or comparable to several state-of-the-art baselines.