Goal-directed simulated annealing and simulated sintering

Recently, Sastry et al. [1] proposed a general procedure to estimate the cost of a minimal solution to a combinatorial optimization problem instance. Such an estimate can be used to determine termination criteria for general purpose combinatorial optimization techniques. We have extended this idea and designed an adaptive cooling schedule for simulated annealing and simulated sintering based on the cost of a minimal solution to a problem instance [2]. With our approach, a schedule of target-costs is determined based on the cost of an easily located solution and the cost (exact, estimated or lower-bound) of a minimal solution. For each target-cost, we search for a solution of cost less than or equal to the target-cost using a range of values for the control-parameter temperature.