Let $ \mathcal{H} $ be a real Hilbert space. We investigate the long time behavior of the trajectories $ x(.) $ of the vanishing damped nonlinear dynamical system with regularizing term
$ \begin{equation} x^{\prime\prime}(t)+\gamma(t)x^{\prime}(t)+\nabla\Phi(x(t))+\varepsilon (t)\nabla U(x(t)) = 0, \;\;\;\;\;\;\;\;\;({\rm{GAVD}_{\gamma,\varepsilon}}) \end{equation} $
where $ \Phi, U:\mathcal{H}\rightarrow\mathbb{R} $ are two convex continuously differentiable functions, $ \varepsilon(.) $ is a decreasing function satisfying $ \lim\limits_{t\rightarrow+\infty}\varepsilon(t) = 0, $ and $ \gamma(.) $ is a nonnegative function which behaves, for $ t $ large enough, like $ \frac{K}{t^{\theta}} $ where $ K > 0 $ and $ 0\leq\theta\leq1. $ The main contribution of this paper is the following control result: If $ \int_{0}^{+\infty}\frac{\varepsilon(t)} {\gamma(t)}dt = +\infty, $ $ U $ is strongly convex and its unique minimizer $ x^{\ast} $ is also a minimizer of $ \Phi $ then every trajectory $ x(.) $ of (GAVD$ _{\gamma, \varepsilon} $) converges strongly to $ x^{\ast} $ and the rate of convergence to $ 0 $ of its energy function
$ W(t) = \frac{1}{2}\left\Vert x^{\prime}(t)\right\Vert ^{2}+\Phi(x(t))-\Phi ^{\ast}+\varepsilon(t)(U(x(t))-U^{\ast}) $
is of order to $ \circ(1/t^{1+\theta}) $. Moreover, we prove a new result concerning the weak convergence of the trajectories of (GAVD$ _{\gamma, \varepsilon} $) to a common minimizer of $ \Phi $ and $ U $ (if one exists) under a simple condition on the speed of decay of the regularizing factor $ \varepsilon(t) $ to $ 0 $.
Citation: Ramzi May, Chokri Mnasri, Mounir Elloumi. Asymptotic for a second order evolution equation with damping and regularizing terms[J]. AIMS Mathematics, 2021, 6(5): 4901-4914. doi: 10.3934/math.2021287
Let $ \mathcal{H} $ be a real Hilbert space. We investigate the long time behavior of the trajectories $ x(.) $ of the vanishing damped nonlinear dynamical system with regularizing term
$ \begin{equation} x^{\prime\prime}(t)+\gamma(t)x^{\prime}(t)+\nabla\Phi(x(t))+\varepsilon (t)\nabla U(x(t)) = 0, \;\;\;\;\;\;\;\;\;({\rm{GAVD}_{\gamma,\varepsilon}}) \end{equation} $
where $ \Phi, U:\mathcal{H}\rightarrow\mathbb{R} $ are two convex continuously differentiable functions, $ \varepsilon(.) $ is a decreasing function satisfying $ \lim\limits_{t\rightarrow+\infty}\varepsilon(t) = 0, $ and $ \gamma(.) $ is a nonnegative function which behaves, for $ t $ large enough, like $ \frac{K}{t^{\theta}} $ where $ K > 0 $ and $ 0\leq\theta\leq1. $ The main contribution of this paper is the following control result: If $ \int_{0}^{+\infty}\frac{\varepsilon(t)} {\gamma(t)}dt = +\infty, $ $ U $ is strongly convex and its unique minimizer $ x^{\ast} $ is also a minimizer of $ \Phi $ then every trajectory $ x(.) $ of (GAVD$ _{\gamma, \varepsilon} $) converges strongly to $ x^{\ast} $ and the rate of convergence to $ 0 $ of its energy function
$ W(t) = \frac{1}{2}\left\Vert x^{\prime}(t)\right\Vert ^{2}+\Phi(x(t))-\Phi ^{\ast}+\varepsilon(t)(U(x(t))-U^{\ast}) $
is of order to $ \circ(1/t^{1+\theta}) $. Moreover, we prove a new result concerning the weak convergence of the trajectories of (GAVD$ _{\gamma, \varepsilon} $) to a common minimizer of $ \Phi $ and $ U $ (if one exists) under a simple condition on the speed of decay of the regularizing factor $ \varepsilon(t) $ to $ 0 $.
[1] | F. Alvarez, On the minimizing properties of a second order dissipative system in Hilbert spaces, SIAM J. Cont. Optim., 38 (2000), 1102-1119. |
[2] | H. Attouch, A. Cabot, Asymptotic stabilization of inertial gradient dynamics wuth time-dependent viscosity, J. Differ. Equations, 263 (2017), 5412-5458. |
[3] | H. Attouch, Z. Chbani, H. Riahi, Combining fast inertial dynamics for convex optimization with Tikhonov regularization, J. Math. Anal. Appl., 457 (2018), 1065-1094. |
[4] | H. Attouch, Z. Chbani, J. Peypouquet, P. Redont, Fast convergence of inertial dynamics and algorithms with asymptotic vanishing viscosity, Math. Program., 168 (2018), 123-175. |
[5] | H. Attouch, M. O. Czarnecki, Asymptotic control and stabilization of nonlinear oscillators with non-isolated equilibria, J. Differ. Equations, 179 (2002), 278-310. |
[6] | H. Attouch, X. Goudou, P. Redont, The heavy ball with friction method, Ⅰ: The continuous dynamical system: Global exploration of the local minima of a real valued function by asymptotic analysis of a dissipative dynamical system, Commun. Contemp. Math., 2 (2000), 1-34. |
[7] | A. Cabot, The steepest descent dynamical system with control. Applications to constrained minimization, Control Optimization and Calculus of Variations, 10 (2004), 243-258. |
[8] | A. Cabot, P. Frankel, Asymptotics for some semilinear hyperbolic equations with non-autonomous damping, J. Differ. Equations, 252 (2012), 294-322. |
[9] | A. Haraux, M. A. Jendoubi, On a second order dissipative ODE in Hilbert space with an integrable source term, Acta Math. Sci., 32 (2012), 155-163. |
[10] | M. A. Jendoubi, R. May, On an asymptotically autonomous system with Tikhonov type regularizing term, Arch. Math., 95 (2010), 389-399. |
[11] | R. May, Asymptotic for a second order evolution equation with convex potential and vanishing damping, Turk. J. Math., 41 (2017), 681-685. |
[12] | Y. Nestrov, A method of solving a convex programming problem with convergence rate O(1/$k^2$), Sov. Math. Dokl., 27 (1983), 372-376. |
[13] | Z. Opial, Weak convergence of the sequence of successive approximation for nonexpansive mappings, Bull. Amer. Math. Soc., 73 (1967), 591-597. |
[14] | W. Su, S. Boyd, E. J. Candès, A differential equation for modeling Nestrov's accelerated gradient method: Theory and Insights, J. Mach. Learn. Res., 17 (2016), 1-43. |
[15] | A. N. Tikhonov, V. Y. Arsenin, Solutions of Ill-Posed Problems, New York: Halsted Press, 1977. |