Abstract :
In ant colony optimization methods, including ant system and max-min ant system, each ant stochastically generates its candidate solution, in a given iteration, based on the same pheromone tau and heuristic eta information as every other ant. In this paper, we propose a variation in which if an ant generates a particular candidate solution St-1 in iteration t - 1, then the solution components of St-1 will have a higher probability of being selected in the candidate solution St generated by that ant in iteration t. In other words, each ant will be biased in favor of its past decisions, i.e. it will be stubborn. We evaluate this variation in the context of max-min ant system and the traveling salesman problem (TSP), using different degrees of stubbornness, and applying the ANOVA test of statistical significance to determine the level of significance of the results.
Keywords :
artificial intelligence; minimax techniques; stochastic processes; travelling salesman problems; ANOVA test; ant colony optimization; max-min ant system; stochastic process; traveling salesman problem; Analysis of variance; Ant colony optimization; Cities and towns; Context modeling; Particle swarm optimization; Probability distribution; Proposals; System testing; Traveling salesman problems; USA Councils; ANOVA; Ant Colony Optimization; MAX-MIN Ant System; Particle Swarm Optimization; Search Diversity; Traveling Salesman Problem;
