Coordinate and subspace optimization methods for linear least squares with non-quadratic regularization (Q2465768): Difference between revisions

The authors consider the problem of linear least squares with non-quadratic regularization (RLS) in the form: minimize \(\{ \|Az - b\|^2 + \rho (z) \}\), for \(z \in\mathbb R^n\), where \(A\) is an \(m \times n, (m < n)\) full rank matrix. They prove that for this problem the parallel coordinate descent algorithm (previously proposed by one of the authors), the expectation-maximization one and a shrinkage minimization method are essentially equivalent for the numerical solution of the RLS problem. Some image denoising numerical experiments are also presented.