Konvergenzverhalten gewisser Markov-Prozesse

Ulrich Faigle

Séminaire lotharingien de combinatoire, Tome 19 (1988)

Résumé

Probabilistic methods which are summarised under the term Simulated Annealing want to model the minimisation of energy states in thermodynamics. The idea of this method goes back to Metropolis et al., who observed that in Markov processes with symmetric transition probabilities, in the limit an element is produced according to a Boltzmann distribuution.

The exerience with Simulated Annealung soo far shoows that the success is very implemention-dependent. However, the theoretical background for good implementations is largely missing. In this paper, some new results on the convergence of Simulating-Annealing-processes are discussed.