Tid: 12 maj 1997 kl 1515-1700
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Harry Cohn, Melbourne University. List of publications
Titel: Global optimization by Markov chains: The simulated annealing algorithm
Sammanfattning: Suppose that we are given a large but finite set of points S, and a function f whose value in any point of S can be identified. We need to find the minimum of f over S. As the number of points of S is very large, a Markov chain with state space S, called a simulated annealing chain, provides a probabilistic tool for finding an optimal or near optimal point of f. This Markov chain depends on a parameter T called temperature which is lowered until a satisfactory solution is reached.
Some practical and theoretical aspects of the simulated annealing method will be discussed. Examples are provided by the travelling salesman problem.