SimulatedAnnealingOptimizer Class

public class SimulatedAnnealingOptimizer : MultidimensionalOptimizer

Public Class SimulatedAnnealingOptimizer
	Inherits MultidimensionalOptimizer

public ref class SimulatedAnnealingOptimizer : public MultidimensionalOptimizer

type SimulatedAnnealingOptimizer = 
    class
        inherit MultidimensionalOptimizer
    end

Inheritance

Object  →  ManagedIterativeAlgorithm<Vector<Double>, Double, OptimizationSolutionReport>  →  MultidimensionalOptimizer<Vector<Double>, OptimizationSolutionReport>  →  MultidimensionalOptimizer  →  SimulatedAnnealingOptimizer

Use the SimulatedAnnealingOptimizer to find the global minimum of a multidimensional function. This optimizer is particularly useful for problems where the search space is large and complex, and traditional gradient-based methods are not effective due to the presence of many local minima.

Simulated Annealing is a powerful probabilistic optimization technique inspired by the physical process of annealing in metallurgy, where controlled cooling of metal leads to optimal crystal structures. The algorithm mimics this process by gradually reducing the probability of accepting worse solutions as it "cools down", allowing it to escape local optima while eventually converging to a good solution.

The behavior of the algorithm is controlled through several interacting parameters which you can set in the constructor or directly by changing optimizer properties: The initial temperature and cooling rate determine the exploration phase, while the neighborhood size and iterations per temperature level control local search intensity. As the temperature decreases according to the cooling schedule, the algorithm transitions from broad exploration to focused exploitation.

At high temperatures, the algorithm freely explores the solution space by accepting most moves, even those that temporarily worsen the objective value. The neighborhood size scales with temperature, allowing larger jumps early in the search. As the temperature drops, the acceptance probability for worse solutions decreases, and the neighborhood size shrinks, focusing the search on promising regions.

The number of iterations at each temperature level allows thorough exploration of the current neighborhood before cooling continues. The cooling rate controls how quickly the temperature drops - a slower cooling rate (closer to 1) provides more opportunities to escape local optima but requires more computation time. The minimum temperature and stagnation limit serve as stopping criteria, terminating the search when either sufficient cooling has occurred or no improvements are found for an extended period.

For best results, these parameters should be tuned based on the specific problem:

• Complex problems with many local optima benefit from higher initial temperatures and slower cooling rates.
• Larger neighborhood sizes help escape local optima but increase computation time.
• More iterations per temperature level provide more thorough exploration but slow convergence.
• The minimum temperature should be set low enough to ensure sufficient convergence.

Reference