Annals of Operations Research

, Volume 156, Issue 1, pp 61–82

A machine learning approach to algorithm selection for \(\mathcal{NP}\) -hard optimization problems: a case study on the MPE problem

Article

Abstract

Given one instance of an \(\mathcal{NP}\) -hard optimization problem, can we tell in advance whether it is exactly solvable or not? If it is not, can we predict which approximate algorithm is the best to solve it? Since the behavior of most approximate, randomized, and heuristic search algorithms for \(\mathcal{NP}\) -hard problems is usually very difficult to characterize analytically, researchers have turned to experimental methods in order to answer these questions. In this paper we present a machine learning-based approach to address the above questions. Models induced from algorithmic performance data can represent the knowledge of how algorithmic performance depends on some easy-to-compute problem instance characteristics. Using these models, we can estimate approximately whether an input instance is exactly solvable or not. Furthermore, when it is classified as exactly unsolvable, we can select the best approximate algorithm for it among a list of candidates. In this paper we use the MPE (most probable explanation) problem in probabilistic inference as a case study to validate the proposed methodology. Our experimental results show that the machine learning-based algorithm selection system can integrate both exact and inexact algorithms and provide the best overall performance comparing to any single candidate algorithm.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Abdelbar, A. M., & Hedetniemi, S. M. (1998). Approximating MAPs for belief networks is \(\mathcal{NP}\) -hard and other theorems. Artificial Intelligence, 102, 21–38. CrossRefGoogle Scholar
  2. Breese, J. S., & Horvitz, E. (1990). Ideal reformulation of belief networks. In UAI90 (pp. 129–144). Google Scholar
  3. Cooper, G., & Herskovits, E. (1992). A Bayesian method for the induction of probabilistic networks from data. Machine Learning, 9(4), 309–347. Google Scholar
  4. Fink, E. (1998). How to solve it automatically: selection among problem-solving methods. In Proceedings of the fourth international conference on artificial intelligence planning systems (pp. 128–136). Google Scholar
  5. Fung, R., & Chang, K. C. (1989). Weighting and integrating evidence for stochastic simulation in Bayesian networks. Uncertainty in Artificial Intelligence, 5, 209–219. Google Scholar
  6. Gent, I. P., & Walsh, T. (1993). An empirical analysis of search in GSAT. Journal of Artificial Intelligence Research, 1, 47–59. Google Scholar
  7. Glover, F., & Laguna, M. (1997). Tabu search. Boston: Kluwer Academic. Google Scholar
  8. Gomes, C. P., & Selman, B. (1997). Algorithm portfolio design: theory vs. practice. In UAI97 (pp. 190–197). Google Scholar
  9. Guo, H. (2003). Algorithm selection for sorting and probabilistic inference: a machine learning-based approach. PhD thesis, Kansas State University. Google Scholar
  10. Guo, H., Boddhireddy, P., & Hsu, W. (2004). Using Ant algorithm to solve MPE. In The 17th Australian joint conference on artificial intelligence, Dec. 2004, Cairns, Australia. Google Scholar
  11. Hooker, J. (1994). Needed: an empirical science of algorithms. Operations Research, 42, 201–212. CrossRefGoogle Scholar
  12. Hoos, H., & Stutzle, T. (1998). Evaluating Las Vegas algorithms—pitfalls and remedies. In UAI98. Google Scholar
  13. Hoos, H., & Stutzle, T. (2000). Local search algorithms for SAT: an empirical evaluation. Journal of Automated Reasoning, 24(4), 421–481. CrossRefGoogle Scholar
  14. Horvitz, E. (1990). Computation and action under bounded resources. PhD thesis, Stanford University. Google Scholar
  15. Horvitz, E., Ruan, Y., Kautz, H., Selman, B., & Chickering, D. M. (2001). A Bayesian approach to tackling hard computational problems. In UAI01 (pp. 235–244). Google Scholar
  16. Houstis, E. N., Catlin, A. C., Rice, J. R., Verykios, V. S., Ramakrishnan, N., & Houstis, C. (2000). PYTHIA-II: a knowledge/database system for managing performance data and recommending scientific software. ACM Transactions on Mathematical Software, 26(2), 227–253. CrossRefGoogle Scholar
  17. Hutter, F. (2005). Stochastic local search for solving the most probable explanation problem in Bayesian networks. M.S. thesis, Intellectics Group, Darmstadt University of Technology. Google Scholar
  18. Ide, J. S., & Cozman, F. G. (2002). Random generation of Bayesian networks. In Brazilian symposium on artificial intelligence, Pernambuco, Brazil. Google Scholar
  19. Jensen, F. V., Olesen, K. G., & Anderson, K. (1990). An algebra of Bayesian belief universes for knowledge-based systems. Networks, 20, 637–659. CrossRefGoogle Scholar
  20. Jitnah, N., & Nicholson, A. E. (1998). Belief network algorithms: a study of performance based on domain characterization. In Learning and reasoning with complex representations (Vol. 1359, pp. 169–188). New York: Springer. Google Scholar
  21. Johnson, D. S. (2002). A theoretician’s guide to the experimental analysis of algorithms. In M. H. Goldwasser, D. S. Johnson & C. C. McGeoch (Eds.), Data structures, near neighbor searches, and methodology: fifth and sixth DIMACS implementation challenges (pp. 215–250). Google Scholar
  22. Kask, K., & Dechter, R. (1999). Stochastic local search for Bayesian networks. In Workshop on AI and statistics 99 (pp. 113–122). Google Scholar
  23. Kirkpatrick, S., Gelatt, C. D., & Vecchi, M. P. (1983). Optimization by simulated annealing. Science, 220(4598), 671–680. CrossRefGoogle Scholar
  24. Lagoudakis, M., & Littman, M. (2001). Learning to select branching rules in the DPLL procedure for satisfiability. Electronic notes in discrete mathematics (ENDM): Vol. 9. LICS 2001 workshop on theory and applications of satisfiability testing (SAT 2001), Boston, MA, June 2001. Google Scholar
  25. Lagoudakis, M., Littman, M., & Parr, R. (2001). Selection the right algorithm. In Proceedings of the 2001 AAAI fall symposium series: using uncertainty within computation, Cape Cod, MA, November 2001. Google Scholar
  26. Lauritzen, S. L., & Spiegelhalter, D. J. (1988). Local computations with probabilities on graphical structures and their application to expert systems (with discussion). Journal of the Royal Statistical Society Series B, 50, 157–224. Google Scholar
  27. Leyton-Brown, K., Nudelman, E., & Shoham, Y. (2002). Learning the empirical hardness of optimization problems: the case of combinatorial auctions. In Constraint programming 2002 (CP-02). Google Scholar
  28. Leyton-Brown, K., Nudelman, E., Andrew, G., McFadden, J., & Shoham, Y. (2003a). A portfolio approach to algorithm selection. In IJCAI. Google Scholar
  29. Leyton-Brown, K., Nudelman, E., Andrew, G., McFadden, J., & Shoham, Y. (2003b). Boosting as a metaphor for algorithm design. Preprint. Google Scholar
  30. Littman, M. (1999). Initial experiments in stochastic search for Bayesian networks. In Proceedings of the sixteenth national conference on artificial intelligence (pp. 667–672). Google Scholar
  31. Lobjois, L., & Lema, M. (1998). Branch and bound algorithm selection by performance prediction. In Proceedings of the fifteenth national/tenth conference on AI/innovative applications of AI (pp. 353–358). Google Scholar
  32. Lucks, M., & Gladwell, I. (1992). Automated selection of mathematical software. ACM Transactions on Mathematical Software, 18(1), 11–34. CrossRefGoogle Scholar
  33. Mannila, H. (1985). Instance complexity for sorting and NP-complete problems. PhD thesis, Department of Computer Science, University of Helsinki. Google Scholar
  34. McGeoch, C. C. (1986). Experimental analysis of algorithms. PhD thesis, Carnegie-Mellon University. Google Scholar
  35. Mengshoel, O. J. (1999). Efficient Bayesian network inference: genetic algorithms, stochastic local search, and abstraction. Computer Science Department, University of Illinois at Urbana-Champaign. Google Scholar
  36. Moret, B. M. E. (2002). Towards a discipline of experimental algorithmics. In Data structures, near neighbor searches, and methodology: fifth and sixth DIMACS implementation challenges. DIMACS monographs (Vol. 59, pp. 197–213). Google Scholar
  37. Orponen, P., Ko, K., Schoning, U., & Watanabe, O. (1994). Instance complexity. Journal of the ACM, 41(1), 96–121. CrossRefGoogle Scholar
  38. Park, J. D. (2002). Using weighted MAX-SAT engines to solve MPE. In Proceedings of the 18th national conference on artificial intelligence (AAAI) (pp. 682–687). Google Scholar
  39. Pearl, J. (1988). Probabilistic reasoning in intelligent systems: networks of plausible inference. San Mateo: Morgan Kaufmann. Google Scholar
  40. Ramakrishnan, N., & Valdes-perez, R. E. (2000). Note on generalization in experimental algorithmics. ACM Transactions on Mathematical Software, 26(4), 568–580. CrossRefGoogle Scholar
  41. Rardin, R. L., & Uzsoy, R. (2001). Experimental evaluation of heuristic optimization algorithms: a tutorial. Journal of Heuristics, 7(3), 261–304. CrossRefGoogle Scholar
  42. Rice, J. R. (1976). The algorithm selection problem. In M. V. Zelkowitz (Ed.), Advances in computers (Vol. 15, pp. 65–118). Google Scholar
  43. Ruan, Y., Kautz, H., & Horvitz, E. (2004). The backdoor key: a path to understanding problem hardness. In Nineteenth national conference on artificial intelligence, San Jose, CA, 2004. Google Scholar
  44. Russell, S., & Norvig, P. (2003). Artificial intelligence: a modern approach. Englewood Cliffs: Prentice-Hall. Google Scholar
  45. Sanders, P. (2002). Presenting data from experiments in algorithmics. In Experimental algorithmics: from algorithm design to robust and efficient software (pp. 181–196). New York: Springer. Google Scholar
  46. Santos, E. (1991). On the generation of alternative explanations with implications for belief revision. In UAI 91 (pp. 339–347). Google Scholar
  47. Santos, E., Shimony, S. E., & Williams, E. (1995). On a distributed anytime architecture for probabilistic reasoning (Technique report AFIT/EN/TR94-06). Department of Electrical and Computer Engineering, Air Force Institute of Technology. Google Scholar
  48. Shafer, G., & Shenoy, P. (1990). Probability propagation. Annals of Mathematics and Artificial Intelligence, 2, 327–352. CrossRefGoogle Scholar
  49. Shimony, S. E., & Charniak, E. (1999). A new algorithm for finding MAP assignments to belief network. In UAI 99 (pp. 185–193). Google Scholar
  50. Shimony, S. E., & Domshlak, C. (2003). Complexity of probabilistic reasoning in directed-path singly connected Bayes networks. Artificial Intelligence, 151, 213–225. CrossRefGoogle Scholar
  51. Witten, I. H., & Frank, E. (1999). Data mining: practical machine learning tools and techniques with Java implementations. Los Altos: Morgan Kaufmann. Google Scholar
  52. Zilberstein, S. (1993). Operational rationality through compilation of anytime algorithms. PhD thesis, University of California at Berkeley. Google Scholar

Copyright information

© Springer Science+Business Media, LLC 2007

Authors and Affiliations

  1. 1.Computer Science and Techology ProgramBNU-HKBU United International CollegeZhuHaiChina
  2. 2.Department of Computing and Information SciencesKansas State UniversityManhattanUSA

Personalised recommendations