English

Combining fast inertial dynamics for convex optimization with Tikhonov regularization

Optimization and Control 2016-02-08 v1

Abstract

In a Hilbert space setting H\mathcal H, we study the convergence properties as t+t \to + \infty of the trajectories of the second-order differential equation \begin{equation*} \mbox{(AVD)}_{\alpha, \epsilon} \quad \quad \ddot{x}(t) + \frac{\alpha}{t} \dot{x}(t) + \nabla \Phi (x(t)) + \epsilon (t) x(t) =0, \end{equation*} where Φ\nabla\Phi is the gradient of a convex continuously differentiable function Φ:HR\Phi: \mathcal H \to \mathbb R, α\alpha is a positive parameter, and ϵ(t)x(t)\epsilon (t) x(t) is a Tikhonov regularization term, with limtϵ(t)=0\lim_{t \to \infty}\epsilon (t) =0. In this damped inertial system, the damping coefficient αt\frac{\alpha}{t} vanishes asymptotically, but not too quickly, a key property to obtain rapid convergence of the values. In the case ϵ()0\epsilon (\cdot) \equiv 0, this dynamic has been highlighted recently by Su, Boyd, and Cand\`es as a continuous version of the Nesterov accelerated method. Depending on the speed of convergence of ϵ(t)\epsilon (t) to zero, we analyze the convergence properties of the trajectories of \mbox(AVD)α,ϵ\mbox{(AVD)}_{\alpha, \epsilon}. We obtain results ranging from the rapid convergence of Φ(x(t))\Phi (x(t)) to minΦ\min \Phi when ϵ(t)\epsilon (t) decreases rapidly to zero, up to the strong ergodic convergence of the trajectories to the element of minimal norm of the set of minimizers of Φ\Phi, when ϵ(t)\epsilon (t) tends slowly to zero.

Cite

@article{arxiv.1602.01973,
  title  = {Combining fast inertial dynamics for convex optimization with Tikhonov regularization},
  author = {Hedy Attouch and Zaki Chbani},
  journal= {arXiv preprint arXiv:1602.01973},
  year   = {2016}
}
R2 v1 2026-06-22T12:44:10.119Z