Abstract
The EBG learning technique has been mainly used in learning processes based on positive examples and successful experiences. However, several authors have demonstrated that failed proofs revealed to be quite useful as a form of avoiding future failures. The first attempts to learn from failure were based on the axiomatization of the problem-solver and on the creation of a specific meta-theory for all possible failures. Whenever there is a positive example of a failure, EBG is used to make operational the meta-theory.
Siqueira & Puget designed a new technique with a different philosophy to learn from counter-examples using only the domain theory. Their method finds a sufficient generalized condition from the failed proof of a goal. EBGF is still a fragile and incomplete technique as it doesn't cover all cases. The failure of a proof has specific characteristics which are not considered when we deal with positive proofs. In this paper we show the weaknesses of EBGF and we propose an improved technique to learn from failures in the presence of a counter-example. Our method is implemented in Prolog and its efficiency is currently under analysis.
Chapter PDF
References
Siqueira, J. CL. e Puget, J. F., Explanation-Based Generalization of Failures, Proceedings of the ECAI-88, pp 339–344, 1988.
Mitchell T. M., Keller R. M., & Kedar-Kabelli S. T., Explanation-based Generalization: a unifying view, Machine Learning 1:1, pp 47–80, 1986.
Minton, S. and Carbonell, J. G., Strategies for Learning Search Control Rules: An Explanation-based Approach. Proceedings of the 10th. IJCAI, Milan, pp 228–235, 1987.
Gupta, A., Explanation-Based Failure Recovery, Proceedings AAAI-87, pp 606–610, 1987.
Hirsh, H., Explanation-based generalization in a logic-programming environment. In Proceedings of the 10th. IJCAI, Milan, pp 221–227, 1987.
Hirsh H., Reasoning about operationality for Explanation-based Learning. In Proceedings of the Fourth International Workshop on Machine Learning, pp 214–220, 1988.
Kedar-Kabelli, S. T.& Mc Carty, CL. T., Explanation-Based Generalization as Resolution Theorem Proving, Proceedings of the Fourth International Workshop on Machine Learning, pp 383–389, 1987.
Hammond, J. K., Explanation and Repairing Plans that Fail, Proceedings of the 10th. IJCAI, Milan, pp 109–114, 1987.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1991 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Urbano, P. (1991). Learning by explanation of failures. In: Kodratoff, Y. (eds) Machine Learning — EWSL-91. EWSL 1991. Lecture Notes in Computer Science, vol 482. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0017024
Download citation
DOI: https://doi.org/10.1007/BFb0017024
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-53816-5
Online ISBN: 978-3-540-46308-5
eBook Packages: Springer Book Archive