Optimization Theory
Simulated annealing
- Crystallization Analogy
- Carefully lower temperature (annealing)
- Algorithm
- Given objective function and concise description of configuration, generate rearrangements of configuration
- If ?E<=0, accept new configuration, if ?Eɬ, accept with probability ~exp(-?E/T)
- T follows annealing schedule
- Results
- Temperatures prevents algorithm from getting stuck, provides extra control, finds better solutions
- Applied to a wide variety of NP-complete problems, proves a “natural framework for heuristic design”