This is a borderline paper. The reviewers suggest, and I agree with them, that the paper can be improved by a convincing real world application, stronger motivation of the problem, discussion of alternative approaches, and a more thorough discussion of practical aspects of the algorithm.