We study the convergence of general descent methods applied to a lower semi-continuous and nonconvex function, which satisfies the Kurdyka-Łojasiewicz inequality in a Hilbert space. We prove that any precompact sequence converges to a critical point of the function, and obtain new convergence rates both for the values and the iterates. The analysis covers alternating versions of the forward-backward method with variable metric and relative errors. As an example, a nonsmooth and nonconvex version of the Levenberg-Marquardt algorithm is detailed.
Given a Hilbert space H and a closed convex function : H → R ∪ {+∞}, we consider the inertial proximal algorithmwhere (α n ) and (β n ) are nonnegative sequences. The notation ∂ stands for the subdifferential of in the sense of convex analysis. This algorithm can be viewed as the implicit discretization of a continuous gradient system involving a memory term. We give conditions that ensure that a suitable discrete energy decreases to inf as n → +∞. When has a unique minimum, the question of the convergence of (x n ) is solved. In the case of multiple minima, it is proved that if n k=1 α k ∈ l 1 and if a suitable geometric condition on the set argmin is fulfilled, then non stationary sequences of (A) cannot converge.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.