On the LearnAbility of Abstraction Theories from Observations for Relational Learning
The most common methodology in symbolic learning consists in inducing, given a set of observations, a general concept definition. It is widely known that the choice of the proper description language for a learning problem can affect the efficacy and effectiveness of the learning task. Furthermore, most real-world domains are affected by various kinds of imperfections in data, such as inappropriateness of the description language which does not contain/facilitate an exact representation of the target concept. To deal with such kind of situations, Machine Learning approaches moved from a framework exploiting a single inference mechanism, such as induction, towards one integrating multiple inference strategies such as abstraction. The literature so far assumed that the information needed to the learning systems to apply additional inference strategies is provided by a domain expert. The goal of this work is the automatic inference of such information.
The effectiveness of the proposed method was tested by providing the generated abstraction theories to the learning system INTHELEX as a background knowledge to exploit its abstraction capabilities. Various experiments were carried out on the real-world application domain of scientific paper documents, showing the validity of the approach.
KeywordsTarget Concept Inductive Logic Programming Inductive Learning Inference Strategy Unary Predicate
- 1.Ceri, S., Gottlöb, G., Tanca, L.: Logic Programming and Databases. Springer, Heidelberg (1990)Google Scholar
- 2.De Raedt, L.: Interactive Theory Revision - An Inductive Logic Programming Approach. Academic Press, London (1992)Google Scholar
- 3.Drastah, G., Czako, G., Raatz, S.: Induction in an abstraction space: A form of constructive induction. In: Proceedings of the International Joint Conference on Artificial Intelligence, pp. 708–712 (1989)Google Scholar
- 6.Flann, N.S., Dietterich, T.G.: Selecting appropriate representations for learning from examples. In: AAAI, pp. 460–466 (1986)Google Scholar
- 7.Giordana, A., Roverso, D., Saitta, L.: Abstracting concepts with inverse resolution. In: Proceedings of the 8th International Workshop on Machine Learning, Evanston, IL, pp. 142–146. Morgan Kaufmann, San Francisco (1991)Google Scholar
- 8.Giordana, A., Saitta, L.: Abstraction: A general framework for learning. In: Working Notes of the Workshop on Automated Generation of Approximations and Abstractions, Boston, MA, pp. 245–256 (1990)Google Scholar
- 9.Kanellakis, P.C.: Elements of relational database theory. In: Van Leeuwen, J. (ed.) Handbook of Theoretical Computer Science. Formal Models and Semantics, vol. B, pp. 1073–1156. Elsevier Science Publishers, Amsterdam (1990)Google Scholar
- 11.Rouveirol, C., Puget, J.: Beyond inversion of resolution. In: Proceedings of ICML 1997, Austin, TX, pp. 122–130. Morgan Kaufmann, San Francisco (1990)Google Scholar
- 13.Utgoff, P.E.: Shift of bias for inductive concept learning. In: Michalski, R.S., Carbonell, J.G., Mitchell, T.M. (eds.) Machine Learning: an artificial intelligence approach, vol. II, pp. 107–148. Morgan Kaufmann, Los Altos (1986)Google Scholar
- 14.Zucker, J.-D.: Semantic abstraction for concept representation and learning. In: Michalski, R.S., Saitta, L. (eds.) Proceedings of the 4th International Workshop on Multistrategy Learning, pp. 157–164 (1998)Google Scholar
- 16.Zucker, J.-D., Ganascia, J.-G.: Representation changes for efficient learning in structural domains. In: Saitta, L. (ed.) Proceedings of the 13th International Conference on Machine Learning, pp. 543–551. Morgan Kaufmann, San Francisco (1996)Google Scholar