Representation, constraint satisfaction and the knapsack problem

This paper uses the simple structure of the knapsack problem to study the issues of representation and constraint satisfaction in evolutionary algorithms. We discuss a wide range of possible representations and constraint satisfaction schemes and demonstrate the importance of these choices. The best results confirm the results of an earlier paper that a redundant mapping and using a decoder incorporating a simple heuristic give better results, suggesting that using indirect mappings for discrete constrained optimisation problems is beneficial