Projection neural network for a class of sparse regression problems with cardinality penalty

Abstract In this paper, we consider a class of sparse regression problems, whose objective function is the summation convex loss and cardinality penalty. By constructing smoothing for function, propose projection neural network design correction method solving problem. The solution proposed unique, global existent, bounded globally Lipschitz continuous. Besides, prove that all accumulation points have common support set unified lower bound nonzero entries. Combining with method, any corrected point local minimizer considered Moreover, analyze equivalence on minimizers between problem another Finally, some numerical experiments are provided to show efficiency in problems.