mynkingannealing

Mynkingannealing is a stochastic optimization metaheuristic designed to locate high-quality solutions for difficult objective functions. The method extends simulated annealing by introducing a specialized candidate-generation operator called the Mynk operator, which combines local perturbations with selective larger jumps to improve diversification while preserving convergence potential.

The algorithm iteratively perturbs the current solution to produce a candidate, evaluates its objective value, and accepts or rejects it using a probabilistic criterion that depends on a temperature parameter. As the temperature decays, the procedure becomes less likely to accept worse solutions, steering the search toward promising regions.

The Mynk operator performs targeted mutations on a subset of decision variables, blending small, neighborhood moves with occasional larger reconfigurations. This bias toward downhill moves at higher temperatures helps escape shallow local minima while still exploring distant regions at the early stages of search.

Variants of mynkingannealing may employ multiple chains, adaptive cooling, or constraint penalties. Practical use requires choosing an initial temperature, a cooling schedule, and the size of the Mynk subset. Common principles derive from problem dimensionality, objective smoothness, and acceptable computational cost.

Applications include combinatorial problems such as scheduling and routing, as well as continuous optimization in engineering design and machine learning hyperparameter tuning. Limitations include parameter sensitivity and computational overhead for large-scale problems. See also simulated annealing, genetic algorithms, and quantum annealing.