High-Dimensional Model-Based Optimization Based on Noisy Evaluations of Computer Games

  • Mike Preuss
  • Tobias Wagner
  • David Ginsbourger
Part of the Lecture Notes in Computer Science book series (LNCS, volume 7219)

Abstract

Most publications on surrogate models have focused either on the prediction quality or on the optimization performance. It is still unclear whether the prediction quality is indeed related to the suitability for optimization. Moreover, most of these studies only employ low-dimensional test cases. There are no results for popular surrogate models, such as kriging, for high-dimensional (n > 10) noisy problems. In this paper, we analyze both aspects by comparing different surrogate models on the noisy 22-dimensional car setup optimization problem, based on both, prediction quality and optimization performance. In order not to favor specific properties of the model, we run two conceptually different modern optimization methods on the surrogate models, CMA-ES and BOBYQA. It appears that kriging and random forests are very good modeling techniques with respect to both, prediction quality and suitability for optimization algorithms.

Keywords

Computer Games Design and Analysis of Computer Experiments Kriging Model-Based Optimization Sequential Parameter Optimization The Open Racing Car Simulator 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Biermann, D., Weinert, K., Wagner, T.: Model-based optimization revisited: Towards real-world processes. In: Michalewicz, Z., Reynolds, R.G. (eds.) Proc. 2008 IEEE Congress on Evolutionary Computation (CEC 2008), pp. 2980–2987. IEEE Press, Piscataway (2008)Google Scholar
  2. 2.
    Breiman, L.: Random forests. Machine Learning 45(1), 5–32 (2001)MATHCrossRefGoogle Scholar
  3. 3.
    Cardamone, L., Loiacono, D., Lanzi, P.L.: Car Setup Optimization. Competition Software Manual. Tech. Rep. 2010.1, Dipartimento di Elettronica e Informazione, Politecnico di Milano, Italy (2010)Google Scholar
  4. 4.
    Cressie, N.: The origins of kriging. Mathematical Geology 22(3), 239–252 (1990)MATHMathSciNetCrossRefGoogle Scholar
  5. 5.
    Forrester, A.I.J., Keane, A.J., Bressloff, N.W.: Design and analysis of ’noisy’ computer experiments. AIAA Journal 44(10), 2331–2339 (2006)CrossRefGoogle Scholar
  6. 6.
    Ginsbourger, D., Picheny, V., Roustant, O., Richet, Y.: Kriging with heterogeneous nugget effect for the approximation of noisy simulators with tunable fidelity. In: Proc. Joint Meeting of the Statistical Society of Canada and the Société Francaise de Statistique (2008)Google Scholar
  7. 7.
    Hansen, N., Ostermeier, A.: Completely derandomized self-adaptation in evolution strategies. Evolutionary Computation 9(2), 159–195 (2001)CrossRefGoogle Scholar
  8. 8.
    Hastie, T., Tibshirani, R.: Generalized additive models. Statistical Science 1(3), 297–318 (1986)MathSciNetCrossRefGoogle Scholar
  9. 9.
    Hastie, T., Tibshirani, R., Friedman, J.: The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd edn. Springer (2009)Google Scholar
  10. 10.
    Horton, R.L.: The General Linear Model. McGraw-Hill, New York (1978)MATHGoogle Scholar
  11. 11.
    Huang, D., Allen, T.T., Notz, W.I., Zheng, N.: Global optimization of stochastic black-box systems via sequential kriging meta-models. Journal on Global Optimization 34(4), 441–466 (2006)MATHMathSciNetCrossRefGoogle Scholar
  12. 12.
    Jin, Y.: A comprehensive survey of fitness approximation in evolutionary computation. Soft Computing 9(1), 3–12 (2005)CrossRefGoogle Scholar
  13. 13.
    Jones, D.R.: A taxonomy of global optimization methods based on response surfaces. Journal of Global Optimization 21(4), 345–383 (2001)MATHMathSciNetCrossRefGoogle Scholar
  14. 14.
    Jones, D.R., Schonlau, M., Welch, W.J.: Efficient global optimization of expensive black-box functions. Journal of Global Optimization 13(4), 455–492 (1998)MATHMathSciNetCrossRefGoogle Scholar
  15. 15.
    Kemmerling, M.: Optimierung der Fahrzeugabstimmung auf Basis verrauschter Daten einer Autorennsimulation. Diplomarbeit, TU Dortmund (2010) (in German)Google Scholar
  16. 16.
    Kemmerling, M., Preuss, M.: Automatic adaptation to generated content via car setup optimization in torcs. In: Proceedings of the 2010 IEEE Conference on Computational Intelligence and Games, CIG 2010, pp. 131–138. IEEE (2010)Google Scholar
  17. 17.
    Kleijnen, J.P.C.: Kriging metamodeling in simulation: A review. European Journal of Operational Research 192(3), 707–716 (2009)MATHMathSciNetCrossRefGoogle Scholar
  18. 18.
    Powell, M.J.D.: The bobyqa algorithm for bound constrained optimization without derivatives. Tech. Rep. DAMTP 2009/NA06, Centre for Mathematical Sciences, University of Cambridge, UK (2009)Google Scholar
  19. 19.
    Preuss, M., Rudolph, G., Wessing, S.: Tuning optimization algorithms for real-world problems by means of surrogate modeling. In: Proc. 12th Annual Conference on Genetic and Evolutionary Computation (GECCO 2010), pp. 401–408. ACM, New York (2010)CrossRefGoogle Scholar
  20. 20.
    Rasmussen, C.E., Williams, C.K.I.: Gaussian processes for machine learning. Springer (2006)Google Scholar
  21. 21.
    Sacks, J., Welch, W.J., Mitchell, T.J., Wynn, H.P.: Design and analysis of computer experiments. Statistical Science 4(4), 409–423 (1989)MATHMathSciNetCrossRefGoogle Scholar
  22. 22.
    Tenne, Y., Izui, K., Nishiwaki, S.: Dimensionality-reduction frameworks for computationally expensive problems. In: Fogel, G., Ishibuchi, H. (eds.) Proc. 2010 IEEE Congress on Evolutionary Computation (CEC 2010), pp. 1–8. IEEE Press, Piscataway (2010)Google Scholar
  23. 23.
    Wang, G.G., Shan, S.: Review of metamodeling techniques in support of engineering design optimization. Journal of Mechanical Design 129(4), 370–380 (2007)MathSciNetCrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Mike Preuss
    • 1
  • Tobias Wagner
    • 2
  • David Ginsbourger
    • 3
  1. 1.Chair of Algorithm Engineering (LS 11)Technische Universität DortmundDortmundGermany
  2. 2.Institute of Machining Technology (ISF)Technische Universität DortmundDortmundGermany
  3. 3.Institute of Mathematical Statistics and Actuarial ScienceUniversity of BernBernSwitzerland

Personalised recommendations