Intelligent Engineering Systems through Artificial Neural Networks, Volume 20
34 Developing a Kakuro Puzzle Solver Using Swarm Intelligence
Download citation file:
Many applications involve highly coupled attributes that can range broadly in values. The Kakuro puzzle, an NP-Complete problem, mimics the situation but in a form of numerical combinations. It is expected that an efficient solver can assist in several technological needs, such as finding the optimal data storage utilization, circuit wiring and multiprocessor scheduling. The existing solver is algorithmic-based using lookup tables and exercising recursions. They are good for small size puzzles whose entries have few possible combinations. This motivates the search of alternative methods. In this paper, a swarm intelligent learning strategy is proposed. A set of artificial agents obey rules and follow heuristics to communicate with each other without lookup. The iteratively gained knowledge guides the agents toward the solution. Experiments show the advantage of learning in handling abundant value combinations.