Simulated Simulated Annealing

SYNOPTIC ABSTRACTThe principal shortcoming of simulated annealing (SA) is that it takes too much computer time. We present a few “swindling” ideas for speeding up SA by simulating its action on a problem. The increase in speed is attained at the cost of decreasing generality — the methods all require the use of problem-specific information. We illustrate the ideas with examples, and discuss our computational experience. We conclude with remarks on the usefulness of simulated annealing.