Representation selection for constraint satisfaction: a case study using n-queens

Representation selection for a constraint satisfaction problem (CSP) is addressed. The CSP problem class is introduced and the classic n-queens problem is used to show that many different CSP representations may exist for a given real-world problem. The complexities of solving these alternative representations are compared empirically and theoretically. The good agreement found is due to two key features of the analytic results, their generality and their precision (or instance specificity), which are also discussed. The n-queens problem is used only to provide a convenient case study; the approach to CSP representation selection applies to arbitrary problems that can be formulated in terms of CSP and, when the corresponding mathematical results are available, should also be readily applicable when selecting representations in domains other than CSP.<>