Sparse Representation Using Stepwise Tikhonov Regularization With Offline Computations

TLDR: A novel algorithm for sparse reconstruction that relies on the recently proposed stepwise Tikhonov regularization (STIR) method to implement forward selection procedures such as Orthogonal least squares, orthogonal matching pursuit, and STIR.

Abstract: This letter describes a novel algorithm for sparse reconstruction. The method uses offline computations to reduce the computational burden of online execution. The approach relies on the recently proposed stepwise Tikhonov regularization (STIR) method to implement forward selection procedures such as orthogonal least squares (OLS), orthogonal matching pursuit (OMP), and STIR. Numerical simulations show the efficacy of the proposed approach, which is competitive against state-of-the-art implementation of OLS and OMP.