Constraint isomorphism and the generation of stochastic data

We consider the problem of generating random data under constraints that are expressed in terms of different parameter sets. These constraints must be consistent between the parameter sets. However, this requirement of constraint consistency has to date not received much attention in the literature. The major objective of this article is to propose a formal concept called constraint isomorphism to detect and help avoid inconsistencies between the constraints. The method presented here can be used as a verification technique for random-data generation. As a case study, we illustrate our methodology on the total-tardiness problem: a NP-hard job scheduling problem. Since generating random data under constraints is an extremely common problem, especially in the simulation arena, the technique has a wide spectrum of potential applications.

Acknowledgements

The authors are grateful to Professor David Goldsman, the anonymous referees and Professor Allan Waren of Cleveland State University for their valuable comments.