Fast simulated annealing in $\R^d$ and an application to maximum likelihood estimation

Abstract : Using classical simulated annealing to maximise a function $\psi$ defined on a subset of $\R^d$, the probability $\p(\psi(\theta_n)\leq \psi_{\max}-\epsilon)$ tends to zero at a logarithmic rate as $n$ increases; here $\theta_n$ is the state in the $n$-th stage of the simulated annealing algorithm and $\psi_{\max}$ is the maximal value of $\psi$. We propose a modified scheme for which this probability is of order $n^{-1/3}\log n$, and hence vanishes at an algebraic rate. To obtain this faster rate, the exponentially decaying acceptance probability of classical simulated annealing is replaced by a more heavy-tailed function, and the system is cooled faster. We also show how the algorithm may be applied to functions that cannot be computed exactly but only approximated, and give an example of maximising the log-likelihood function for a state-space model.
Type de document :
Article dans une revue
Stochastic Processes and their Applications, Elsevier, 2009, 119 (6), pp.1912-1931. 〈10.1016/j.spa.2008.09.007〉
Liste complète des métadonnées

Littérature citée [12 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-00093403
Contributeur : Sylvain Rubenthaler <>
Soumis le : mercredi 13 septembre 2006 - 14:24:20
Dernière modification le : vendredi 12 janvier 2018 - 01:54:51
Document(s) archivé(s) le : mardi 6 avril 2010 - 00:55:38

Identifiants

Collections

Citation

Sylvain Rubenthaler, Tobias Rydén, Magnus Wiktorsson. Fast simulated annealing in $\R^d$ and an application to maximum likelihood estimation. Stochastic Processes and their Applications, Elsevier, 2009, 119 (6), pp.1912-1931. 〈10.1016/j.spa.2008.09.007〉. 〈hal-00093403〉

Partager

Métriques

Consultations de la notice

303

Téléchargements de fichiers

123