Approaching nonsmooth nonconvex optimization problems through first order dynamical systems with hidden acceleration and Hessian driven damping terms (Q1653320): Difference between revisions

In this paper the authors carry out an asymptotic analysis of the proximal-gradient dynamical system \[ \begin{cases} \dot x(t)+x(t)=prox_{\gamma f}[x(t)-\gamma\nabla \phi(x(t))-ax(t)-by(t)],\\ \dot y(t)+ax(t)+by(t)=0 \end{cases} \] where \(f\) is a proper, convex and lower semicontinuous function, \(\phi\) a possibly nonconvex smooth function and \(\gamma\), \(a\) and \(b\) are positive real numbers. They show that the generated trajectories approach the set of critical points of \(f+\phi\), here understood as zeros of its limiting subdifferential, under the premise that a regularization of this sum function satisfies the Kurdyka-Łojasiewicz property. The authors also establish convergence rates for the trajectories, formulated in terms of the Łojasiewicz exponent of the considered regularization function.