The simulated annealing algorithm found inspiration in the physical process of annealing. Material is heated to high temperature and a gradual slow reduction of temperature eliminate the internal defects. The gradual cooling process eliminate unstable particles, there is a stabilization of the crystal lattice and the material gets the required quality.
Interpretation in the way of evolutionary algorithms means that algorithm iterates by gradually reducing of the temperature. For each temperature is randomly generated a new solution which is subjected to a Metropolis criterion, whitch determines for the current temperature the probability of replacing the current solution to new .
Description of the simulated annealing algorithm:
| 1. | At beggining it generates a random solution and knows start temperature. |
| 2. | It generates a new random solution for the current temperature. It accept only those which are fulfill by Metropolis criterion. This step is repeated until parameter number of solutions for each temperature is reached. |
| 3. | It reduces temperature. |
| 4. | It continues with step 2 until parameter stop temperature is reached. |
Parameters description:
| Parameter | Description | Recommended Value |
|---|---|---|
| Specimen | Model of solution | Depends on the cost function |
| Cooldown method | "Continuous", "By step" | |
| Start temperature | Depends on the Metropolis algorithm, it should accept arround 50% of all solutions | Depends on the cost function |
| Final temperature | 0 cooldown by steps 0.001 - 1 continuous cool down |
|
| Number of solutions for each temperature | Depends on the cost function | |
| Cooldown step | Determines number of iterations for cooldown by steps method | Podle počáteční teploty, zhruba 1% - 10% <0.75, 0.95> |
| Cooldown multiplier | Determines number of iterations for continuous cool down method | <0.75, 0.95> |
| Elitism | Determines if elitism will be used or no | Yes / No |