On search decision and the efficiency of polynomial-time algorithms

  • Authors:

  • M. R. Fellows;M. A. Langston

  • Affiliations:

  • Department of Computer Science, University of Idaho, Moscow, ID;Department of Computer Science, Washington State University, Pullman, WA

  • Venue:
  • STOC '89 Proceedings of the twenty-first annual ACM symposium on Theory of computing
  • Year:
  • 1989

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recent advances in well-partial-order theory, especially the seminal contributions of Robertson and Seymour, have troubling consequences for those who would equate tractability with polynomial-time decidability. Specifically:many problems are now known to be decidable in low-degree polynomial time, but only by decision algorithms with overwhelmingly astronomical constants of proportionality,the existence of such a polynomial-time decision algorithm alone does not ensure that a corresponding search problem can be solved efficiently, andeven if both a decision problem and a corresponding search problem can be shown to be polynomial-time computable, there is no guarantee that correct algorithms can be found or even recognized within any bounded amount of time.In this paper, we present a number of techniques for dealing with these remarkable features of algorithms based on well-partially-ordered sets. Our main results include a general strategy with which such algorithms can be made constructive. With the aid of this method, we demonstrate that low-degree polynomial-time algorithms are now known for almost all of the catalogued applications of RS posets. We also prove that, despite the nonconstructive nature of the well-partial-order theory on which this line of research is based, no RS poset application can settle P @@@@ N P non-constructively by any established method of argument.