A generalization of the sparse iterative covariance-based estimator

摘要: In this work, we extend the popular sparse iterative covariance-based estimator (SPICE) by generalizing formulation to allow for different norm constraint on signal and noise parameters in covariance model. For any choice of norms, resulting generalized SPICE method enjoys same benefits as regular method, including being hyperparameter free, although is shown govern sparsity solution. Furthermore, show that there a connection between penalized regression problem, both case were one allows differ each sample, when treating parameter equal. We examine performance choices compare results original showing using version. also provide way solving gridless which solves semi-definite programming problem.