Simulated annealing without potential in a finite set
From MaRDI portal
[For the entire collection see Zbl 0754.00008.] Convergence of simulated annealing is shown for continuous time Markov processes in a finite set for the general case, that their generators are not necessarily derived from a potential. The proof is based on estimates of the relative entropy with respect to the corresponding invariant measures and its time derivative. For the case of fixed temperature the techniques yield ergodicity for these Markov processes, which are in general not reversible.
Recommendations
- Note on the Convergence of Simulated Annealing Algorithms
- Rough large deviation estimates for the optimal convergence speed exponent of generalized simulated annealing algorithms
- scientific article; zbMATH DE number 4157608
- A study of subadmissible simulated annealing algorithms
- Recuit simulésans potentiel sur une variétériemannienne compacte
Cited in
(16)- Comportement asymptotique de l'énergie libre spécifique. Application à l'ergodicité et au recuit simulé en dimension infinie. (Asymptotical behaviour of specific free energy. Application to ergodicity and simulated annealing on infinite-dimensional spaces)
- On discrete inhomogeneous exit problems
- On the invariant measure of non-reversible simulated annealing
- Convergence results for a class of time-varying simulated annealing algorithms
- A study of subadmissible simulated annealing algorithms
- Remarks on the ergodicity of simulated annealing algorithms on a graph
- On the simulated annealing in \(\mathbb{R}^d\)
- On metastability
- Convergence of the kinetic annealing for general potentials
- A Remark on Simulated Annealing of Diffusion Processes
- scientific article; zbMATH DE number 1810270 (Why is no real title available?)
- On the Convergence and Applications of Generalized Simulated Annealing
- Une variante de l'inégalité de Cheeger pour les chaînes de Markov finies
- A concentration bound for stochastic approximation via Alekseev's formula
- Optimization by linear kinetic equations and mean-field Langevin dynamics
- On the complex eigenvalues of finite Markov generators
This page was built for publication: Simulated annealing without potential in a finite set
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4028964)