A Gauss-Seidel type inertial proximal alternating linearized minimization for a class of nonconvex optimization problems (Q2307755)

The authors formulate an inertial version of the PALM method for minimizing the sum of two nonsmooth (possibly also nonconvex) functions in separate variables connected by a smooth coupling function. The inertial step is performed once of the two subproblems is addressed, it is formulated in the spirit of a Gauss-Seidel type method. Subsequence convergence and, in the Kurdyka-Lojasiewicz setting, global convergence of the generated sequence to a critical point are proved. Comparisons to PALM and to another inertial version of it are carried out in the context of solving signal recovery, image denoising and nonnegative matrix factorization models.