We consider the annealing diffusion process and investigate convergence rates. Namely, the diffusion
$dX_t = -\nabla V(X_t)dx+\sigma(t)dB_t$, where $(B_t)_t\leq 0$ is the $d$-dimensional Brownian motion and
$\sigma(t)$ decreases to zero, we prove a large deviation principle for $(V(X_t))$ and weak convergence
of $(\sigma^{-2}(t)(V(X_t)-\inf V)).$
Publié le : 1997-11-14
Classification:
Simulated annealing,
diffusion,
large deviations,
weak convergence,
62L20,
60F05,
60F10,
60J60
@article{1043862427,
author = {M\'arquez, David},
title = {Convergence rates for annealing diffusion processes},
journal = {Ann. Appl. Probab.},
volume = {7},
number = {1},
year = {1997},
pages = { 1118-1139},
language = {en},
url = {http://dml.mathdoc.fr/item/1043862427}
}
Márquez, David. Convergence rates for annealing diffusion processes. Ann. Appl. Probab., Tome 7 (1997) no. 1, pp. 1118-1139. http://gdmltest.u-ga.fr/item/1043862427/