Complexity of Unconstrained L_2-L_p Minimization

We consider the unconstrained $L2$-$Lp$ minimization: find a minimizer of $|Ax-b|^2_2+\lambda |x|^p_p$ for given $A \in R^{m\times n}$, $b\in R^m$ and parameters $\lambda>0$, $p\in [0,1)$. This problem has been studied extensively in variable selection and sparse least squares fitting for high dimensional data. Theoretical results show that the minimizers of the $L2$-$Lp$ problem have various attractive features due to the concavity and non-Lipschitzian property of the regularization function $|\cdot|^p_p$. In this paper, we show that the $Lq$-$Lp$ minimization problem is strongly NP-hard for any $p\in [0,1)$ and $q\ge 1$, including its smoothed version. On the other hand, we show that, by choosing parameters $(p,\lambda)$ carefully, a minimizer, global or local, will have certain desired sparsity. We believe that these results provide new theoretical insights to the studies and applications of the concave regularized optimization problems.