Use of Simulated Annealing

Simulated annealing is well suited for problems where:

  1. The number of variables, and thus possible states, is very large. This makes exhaustive search infeasible.

  2. There is ``frustration'' : it is not possible to optimize all cost measures simultaneously.

  3. Significant improvements from a random starting position are possible.

  4. There are many good near-optimal solutions.

  5. Hill-climbing is likely to get stuck on local maxima.

Computation is proportional to N or a small power of N , while finding the exact optimum is often NP-complete.

Contents    Page-10    Prev    Next    Page+10    Index