Title:On Solution Uniqueness and Robust Recovery for Sparse Regularization with a Gauge: from Dual Point of View

Abstract:In this paper, we focus on exploring solution uniqueness, sharpness, and robust recovery to sparse regularization with a gauge J. Our approach is based on the uniqueness of Lagrange multipliers for its dual problem. We observe that the Lagrange bidual problem of a gauge regularization problem is itself and its dual problem satisfies Slater's condition and hence the strong duality holds. Therefore, the solution set to the original gauge regularization problem is the Lagrange multiplier set for its dual problem and is the same for any solution to the dual problem. Based on criteria for the uniqueness of the Lagrange multipliers for the dual problem, we obtain several equivalent sufficient conditions for solution uniqueness to the gauge regularization problem which are also necessary conditions under some closedness condition. We also prove that these sufficient conditions are sufficient and necessary conditions for a sharp solution, which ensures the robust recovery with a linear rate. Some applications to weighted sorted $\ell_1$-norm sparse regularization, sparse analysis regularization, nuclear norm optimization, conic gauge optimization, and semidefinite conic gauge optimization are given.