Advertisement

Journal of Intelligent Information Systems

, Volume 4, Issue 1, pp 89–108 | Cite as

Discovering dynamics: From inductive logic programming to machine discovery

  • Saso Dzeroski
  • Ljupco Todorovski
Article

Abstract

Machine discovery systems help humans to find natural laws from collections of experimentally collected data. Most of the laws found by existing machine discovery systems describe static situations, where a physical system has reached equilibrium. In this paper, we consider the problem of discovering laws that govern the behavior of dynamical systems, i.e., systems that change their state over time. Based on ideas from inductive logic programming and machine discovery, we present two systems, QMN and LAGRANGE, for discovery of qualitative and quantitative laws from quantitative (numerical) descriptions of dynamical system behavior. We illustrate their use by generating a variety of dynamical system models from example behaviors.

Keywords

machine discovery machine learning dynamical system identification 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Babloyantz, A. (1986).Molecules, Dynamics, and Life. John Wiley & Sons, New York.Google Scholar
  2. Bielecki, M. W. (1992). Machine discovery approach to dynamic systems in the real laboratory.ML92 Workshop on Machine Discovery. Aberdeen, Scotland.Google Scholar
  3. Bohte, Z. (1991).Numerical Methods. The Society of Mathematicians, Physicists and Astronomers of Slovenia, Ljubljana, Slovenia.Google Scholar
  4. Bratko, I., Muggleton, S., and Varsek, A. (1992). Learning qualitative models of dynamic systems. In Muggleton, S., editor,Inductive Logic Programming, pages 437–452. Academic Press, London.Google Scholar
  5. Coiera, E. (1989). Learning qualitative models from example behaviors.Third International Workshop on Qualitative Physics. Stanford, CA.Google Scholar
  6. De Raedt, L. and Bruynooghe, M. (1993). A theory of clausal discovery. InProc. Thirteenth International Joint Conference on Artificial Intelligence, pages 1058–1063. Morgan Kaufmann, San Mateo, CA.Google Scholar
  7. Dzeroski, S. and Bratko, I. (1992). Handling noise in inductive logic programming.Second International Workshop on Inductive Logic Programming. Tokyo, Japan.Google Scholar
  8. Dzeroski, S. and Petrovski, I. (1994). Discovering dynamics with genetic programming. InProc. Seventh European Conference on Machine Learning. Springer, Berlin. To appear.Google Scholar
  9. Dzeroski, S. and Todorovski, L. (1993). Discovering dynamics. InProc. Tenth International Conference on Machine Learning, pages 97–103. Morgan Kaufmann, San Mateo, CA, 1993.Google Scholar
  10. Dzeroski, S., Muggleton, S., and Russell, S. (1992). PAC-learnability of determinate logic programs. InProc. Fifth ACM Workshop on Computational Learning Theory, pages 128–135. ACM Press, New York.Google Scholar
  11. Falkenheiner, B. and Michalski, R. (1990). Integrating quantitative and qualitative discovery in the ABACUS system. In Kodratoff, Y. and Michalski, R., editors,Machine Learning: An Artificial Intelligence Approach, pages 153–190. Morgan Kaufmann, San Mateo, CA.Google Scholar
  12. Geva, S. and Sitte, J. (1993). A cartpole experimental benchmark for trainable controllers.IEEE Control Systems, 13(5):40–51.Google Scholar
  13. Kraan, I., Richards, B., and Kuipers, B. (1991). Automatic abduction of qualitative models.Fifth International Workshop on Qualitative Physics. Austin, TX.Google Scholar
  14. Krizman, V. (1994) Handling noisy data in automated modeling of dynamical systems. MSc Thesis, Faculty of Electrical and Computer Engineering, University of Ljubljana, Slovenia.Google Scholar
  15. Kuipers, B. (1986). Qualitative simulation.Artificial Intelligence, 29(3):289–338.Google Scholar
  16. Langley, P., Simon, H., Bradshaw, G., and Zytkow, J. (1987).Scientific discovery. MIT Press, Cambridge, MA.Google Scholar
  17. Lavrac, N. and Dzeroski, S. (1994)Inductive Logic Programming: Techniques and Applications. Ellis Horwood, Chichester.Google Scholar
  18. Lavrac, N., Dzeroski, S., and Grobelnik, M. (1991). Learning nonrecursive definitions of relations with LINUS. InProc. Fifth European Working Session on Learning, pages 265–281. Springer, Berlin.Google Scholar
  19. Ljung, L. (1993). Modelling of industrial systems. InProc. Seventh International Symposium on Methodologies for Intelligent Systems, pages 338–349. Springer, Berlin.Google Scholar
  20. Muggleton, S., editor (1992).Inductive Logic Programming. Academic Press, London.Google Scholar
  21. Nordhausen, B. and Langley, P. (1990). A robust approach to numeric discovery. InProc. Seventh International Conference on Machine Learning, pages 411–418. Morgan Kaufmann, San Mateo, CA.Google Scholar
  22. Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T. (1986).Numerical Recipes. Cambridge University Press, Cambridge, MA.Google Scholar
  23. Urbancic, T. and Bratko, I. (1994) Learning to control dynamic systems. In Michie, D., Spiegelhalter, D., and Taylor, C., editors,Machine Learning, Neural and Statistical Classification. Ellis Horwood, Chichester. In press.Google Scholar
  24. Van Laer, W. and De Raedt, L. (1993). Discovering quantitative laws in inductive logic programming.MLnet Workshop on Machine Discovery. Blanes, Spain.Google Scholar
  25. Volk, W. (1958).Applied Statistics for Engineers. McGraw-Hill, New York.Google Scholar
  26. Zembowicz, R. and Zytkow, J. (1992). Discovery of equations: experimental evaluation of convergence. InProc. Tenth National Conference on Artificial Intelligence, pages 70–75. MIT Press, Cambridge, MA.Google Scholar
  27. Zytkow, J. and Zhu, J. (1991). Application of empirical discovery in knowledge acquisition. InProc. Fifth European Working Session on Learning, pages 101–117. Springer, Berlin.Google Scholar

Copyright information

© Kluwer Academic Publishers 1995

Authors and Affiliations

  • Saso Dzeroski
    • 1
  • Ljupco Todorovski
    • 1
  1. 1.Institut Jozef StefanLjubljanaSlovenia

Personalised recommendations