Using accuracy in scientific discovery

  • M. Moulet
Part 2: Discovery
Part of the Lecture Notes in Computer Science book series (LNCS, volume 482)


Learning by discovery aims at bringing to light laws from a set of numerical or symbolic data. Our work deals with the improvement of the discovery system ABACUS created by Michalski and Falkenhainer, and in particular, with the way the system makes use of informative accuracy of the data. ABACUS, like most others current discovery systems does not use this information in the real physical sense, that means accuracy given by the measure device. However, in experimental domains accuracy cannot obviously be separated from the data. In this paper, we show how, when used in a more realistic manner, this information can significantly improve not only the accuracy of the results but also the efficiency of the search algorithm. Several additional modifications to ABACUS to improve the robustness of the system without losing generality will also be described.


Scientific discovery learning by observation numeric — symbolic integration 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Davenport J., Siret Y., Tournier E. Calcul formel. Systèmes et algorithmes de manipulations algebriques. Collection etudes et recherche en informatique, Eds Masson, Paris, 1986.Google Scholar
  2. Eurin M., Guimiot H. Physique, Classiques HACHETTE, 1953.Google Scholar
  3. Falkenhainer B.C., Michalski R.S. Integrating Quantitative and Qualitative Discovery: The ABACUS system. Machine Learning Journal, vol. 3, 1986.Google Scholar
  4. Falkenhainer B.C., Michalski R.S. Integrating Quantitative and Qualitative Discovery. Machine Learning: An Artificial Intelligence Approach, vol III, R.S. Michalski, J.G. Carbonell, T.M. Mitchell (Eds.), 1990.Google Scholar
  5. Greene G.H. The ABACUS.2 system for quantitative discovery: Using dependencies to discover non-linear terms, MLI 88-17 TR-11-88, 1988.Google Scholar
  6. Joyal M. Cours de physique, Vol. 3 Electricite, Eds Masson & Cie, 1956.Google Scholar
  7. Langley P., Bradshaw G.L., Simon H. BACON.5: the discovery of conservation laws. Proceedings of the seventh International Joint Conference on Artificial Intelligence, p 121–126, 1985.Google Scholar
  8. Langley P., Zytkow J., Simon H.and Bradshaw G.L. The search for regularity: Four aspects of scientific discovery in Machine Learning: An Artificial Intelligence Approach, volume II, Michalski R.S., Carbonell J.G., Mitchell T.M.(Eds.), Tioga, Palo Alto, Calif., 1986.Google Scholar
  9. Langley P., Zytkow J., Simon H. and Bradshaw G.L. Scientific discovery. Computational explorations of the creative process. MIT press, Cambridge, MA, 1987.Google Scholar
  10. Nordhausen B., Langley P. A robust approach to Numeric Discovery", Proceedings of the seventh International Conference on Machine Learning, p 411–418, edited by B.W. Porter and R.J. Mooney, Morgan Kauffman Publishers, Austin, 1990.Google Scholar
  11. Zytkow J. M. Combining many searches in the FAHRENHEIT discovery system. Proceedings of the fourth International Workshop on Machine Learning, p 281–287, Morgan Kauffman Publishers, Irvine, 1987.Google Scholar
  12. Zytkow J. M., Zhu J and Hussam, A., Automated discovery in a chemistry laboratory. Proceedings of the AAAI-90, AAAI Press, p 889–894, 1990.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1991

Authors and Affiliations

  • M. Moulet
    • 1
  1. 1.Equipe Inférence et Apprentissage LRIUniversité Paris-SudOrsay CedexFrance

Personalised recommendations