Metaheuristic Searches
Metaheuristic algorithms are high-level strategies used to guide the search process toward an optimal or near-optimal solution in complex optimization problems.
Key Characteristics of Metaheuristics
- Guided Search Strategy: Metaheuristics determine the path and direction of the search process.
- Stochastic Nature: Most metaheuristic algorithms use randomness in their search process.
- Escape from Local Optima: They incorporate mechanisms to avoid getting stuck in local minima (or maxima) and continue exploring better solutions.
- General-Purpose Methods: Unlike problem-specific algorithms, metaheuristics can be applied to a variety of optimization problems.
- May Use Domain Knowledge: Some metaheuristics integrate domain-specific heuristics at a lower level to improve efficiency.
What Does Stochastic Mean?
- A stochastic algorithm involves randomness in its decision-making process.
- Instead of following a fixed deterministic path, it explores solutions using probability-based choices.
- This randomness helps avoid local optima and increases the chances of finding a globally optimal solution.
Examples of Single-Point Metaheuristic Searches
These algorithms maintain a single candidate solution at any given time and iteratively improve it.
- Simulated Annealing (SA) – Inspired by thermodynamics, it probabilistically accepts worse solutions to escape local optima.
- Tabu Search (TS) – Uses memory structures (tabu list) to avoid revisiting previously explored solutions.
- Iterated Local Search (ILS) – Perturbs the current solution and applies local search to refine it.
- Variable Neighbourhood Search (VNS) – Dynamically changes the neighborhood structure to explore different regions of the search space.