Simulated annealing type metaheuristics - to cool or not to cool /
A popular class of metaheuristics are the hill climbing algorithms. In some variants, including simulated annealing (SA), moves to higher cost solutions are accepted only with a certain probability which depends on the parameter, usually called the temperature. An important part of implementation is...
Shranjeno v:
| Glavni avtor: | |
|---|---|
| Format: | Book Chapter |
| Jezik: | English |
| Teme: | |
| Sorodne knjige/članki: | Vsebovano v:
SOR '03 proceedings |
| Oznake: |
Označite
Brez oznak, prvi označite!
|
| Izvleček: | A popular class of metaheuristics are the hill climbing algorithms. In some variants, including simulated annealing (SA), moves to higher cost solutions are accepted only with a certain probability which depends on the parameter, usually called the temperature. An important part of implementation is the choice of a good temperature schedule. Among various temperature schedules, the constant temperature schedule for simulated "annealing" was found competitive or at least worth consideration. We report on some recent experiments with constant temeprature implementations of simulated annealing type algorithms for coloring and maximal independent sets. The results include improvements of the best upper bound for the Shannon capacity of ▫$C_7$▫. Furthermore, we discuss the results of a recent experimental study comparing various temperature schedules and ideas on learning the temperature on-line. |
|---|---|
| Fizični opis: | [6] str. |
| ISBN: | 9616165151 |