Advertisement

Extreme Reactive Portfolio (XRP): Tuning an Algorithm Population for Global Optimization

  • Mauro BrunatoEmail author
  • Roberto Battiti
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 10079)

Abstract

Given the current glut of heuristic algorithms for the optimization of continuous functions, in some case characterized by complex schemes with parameters to be hand-tuned, it is an interesting research issue to assess whether competitive performance can be obtained by relying less on expert developers (whose intelligence can be a critical component of the success) and more on automated self-tuning schemes.

After a preliminary investigation about the applicability of record statistics, this paper proposes a fast reactive algorithm portfolio based on simple performance indicators: record value and iterations elapsed from the last record. The two indicators are used for a combined ranking and a stochastic replacement of the worst-performing members with a new searcher with random parameters or a perturbed version of a well-performing member.

The results on benchmark functions demonstrate a performance equivalent or better than that obtained by offline tuning schemes, which require a greater amount of CPU time and cannot take care of individual structural variations between different problem instances.

Notes

Acknowledgments

The research of Roberto Battiti was supported by the Russian Science Foundation, project no. 15–11–30022 “Global optimization, supercomputing computations, and applications.”

References

  1. 1.
    Battiti, R., Brunato, M., Campigotto, P.: Learning while optimizing an unknown fitness surface. In: Maniezzo, V., Battiti, R., Watson, J.-P. (eds.) LION 2007. LNCS, vol. 5313, pp. 25–40. Springer, Heidelberg (2008). doi: 10.1007/978-3-540-92695-5_3 CrossRefGoogle Scholar
  2. 2.
    Battiti, R., Brunato, M., Mascia, F.: Reactive Search and Intelligent Optimization. Operations research/Computer Science Interfaces, vol. 45. Springer, Heidelberg (2008)zbMATHGoogle Scholar
  3. 3.
    Battiti, R., Tecchiolli, G.: Learning with first, second, and no derivatives: a case study in high energy physics. Neurocomputing 6, 181–206 (1994)CrossRefGoogle Scholar
  4. 4.
    Battiti, R., Tecchiolli, G.: The continuous reactive tabu search: blending combinatorial optimization and stochastic search for global optimization. Ann. Oper. Res. - Metaheuristics Comb. Optim. 63, 153–188 (1996)zbMATHGoogle Scholar
  5. 5.
    Battiti, R., Brunato, M.: The LION way. Machine Learning plus Intelligent Optimization. LIONlab, University of Trento, Italy (2014)Google Scholar
  6. 6.
    Battiti, R., Campigotto, P.: Reinforcement learning and reactive search: an adaptive max-sat solver. In: Ghallab, M., Spyropoulos, C.D., N.F., Avouris, N. (eds.) Proceedings ECAI 08: 18th European Conference on Artificial Intelligence, Patras, Greece, 21–25 July 2008. IOS Press, Amsterdam (2008)Google Scholar
  7. 7.
    Birattari, M., Stützle, T., Paquete, L., Varrentrapp, K.: A racing algorithm for configuring metaheuristics. In: Langdon, W.B., et al. (eds.) Proceedings of the Genetic and Evolutionary Computation Conference, pp. 11–18. Morgan Kaufmann Publishers, San Francisco (2002). AIDA-2002-01 Technical report of Intellektik, Technische Universität Darmstadt, Darmstadt, GermanyGoogle Scholar
  8. 8.
    Brunato, M., Battiti, R., Pasupuleti, S.: A memory-based rash optimizer. In: Geffner, A.F.R.H.H. (ed.) Proceedings of AAAI-06 Workshop on Heuristic Search, Memory Based Heuristics and Their Applications, pp. 45–51, Boston, Mass. (2006). ISBN 978-1-57735-290-7Google Scholar
  9. 9.
    Cicirello, V.: Boosting stochastic problem solvers through online self-analysis of performance. Ph.D. thesis, Carnegie Mellon University (2003). Technical report CMU-RI-TR-03-27Google Scholar
  10. 10.
    Cicirello, V., Smith, S.: The max k-armed bandit: a new model for exploration applied to search heuristic selection. In: 20th National Conference on Artificial Intelligence (AAAI 2005), July 2005Google Scholar
  11. 11.
    Cicirello, V.A., Smith, S.F.: Heuristic selection for stochastic search optimization: modeling solution quality by extreme value theory. In: Wallace, M. (ed.) CP 2004. LNCS, vol. 3258, pp. 197–211. Springer, Heidelberg (2004). doi: 10.1007/978-3-540-30201-8_17 CrossRefGoogle Scholar
  12. 12.
    Fong, P.W.L.: A quantitative study of hypothesis selection. In: International Conference on Machine Learning, pp. 226–234 (1995). http://citeseer.ist.psu.edu/fong95quantitative.html Google Scholar
  13. 13.
    Gagliolo, M., Schmidhuber, J.: Learning dynamic algorithm portfolios. In: Proceedings AI and MATH 2006, Ninth International Symposium on Artificial Intelligence and Mathematics, Fort Lauderdale, Florida, January 2006Google Scholar
  14. 14.
    Horvitz, E., Ruan, Y., Gomes, C., Kautz, H., Selman, B., Chickering, D.M.: A bayesian approach to tackling hard computational problems. In: Seventeenth Conference on Uncertainty in Artificial Intelligence, pp. 235–244, Seattle, USA, August 2001Google Scholar
  15. 15.
    Huberman, B.A., Lukose, R.M., Hogg, T.: An economics approach to hard computational problems. Science 275, 51–54 (1997)CrossRefGoogle Scholar
  16. 16.
    Kautz, H., Horvitz, E., Ruan, Y., Gomes, C., Selman, B.: Dynamic restart policies. In: Eighteenth National Conference on Artificial Intelligence, pp. 674–681. American Association for Artificial Intelligence, Menlo Park (2002)Google Scholar
  17. 17.
    Lera, D., Sergeyev, Y.D.: Acceleration of univariate global optimization algorithms working with lipschitz functions and lipschitz first derivatives. SIAM J. Optim. 23(1), 508–529 (2013)MathSciNetCrossRefzbMATHGoogle Scholar
  18. 18.
    López-Ibánez, M., Dubois-Lacoste, J., Stützle, T., Birattari, M.: The irace package, iterated race for automatic algorithm configuration. Technical report TR/IRIDIA/2011-004, IRIDIA, Université Libre de Bruxelles, Belgium (2011)Google Scholar
  19. 19.
    Streeter, M.J., Smith, S.F.: An asymptotically optimal algorithm for the max k-armed bandit problem. In: Proceedings of the National Conference on Artificial Intelligence, vol. 21, pp. 135–142. AAAI Press, MIT Press/Menlo Park, Cambridge, London 1999 (2006)Google Scholar
  20. 20.
    Streeter, M.J., Smith, S.F.: A simple distribution-free approach to the max k-armed bandit problem. In: Benhamou, F. (ed.) CP 2006. LNCS, vol. 4204, pp. 560–574. Springer, Heidelberg (2006). doi: 10.1007/11889205_40 CrossRefGoogle Scholar
  21. 21.
    Zhigljavsky, A., Žilinskas, A.: Stochastic Global Optimization, vol. 9. Springer Science & Business Media, Heidelberg (2007)zbMATHGoogle Scholar

Copyright information

© Springer International Publishing AG 2016

Authors and Affiliations

  1. 1.Department of Computer Science and TelecommunicationsUniversity of TrentoTrentoItaly

Personalised recommendations