Analytical learning guided by empirical technology: An approach to integration (Extended abstract)
An approach to integration of the analytical and empirical learning strategies for solving the theory-based concept specialization problem is proposed. The emphasis is on learning correct concepts within conjunctive description language isolated from supplied domain theory. In order to overcome expressive limitations of such language, the analytical learning component is guided by a specific empirical version space technology. The approach learns incrementally a correct pure conjunctive or DNF concept definition in dependance on the domain theory.
- 1.N.S. Flann and T.G. Dietterich, A Study of Explanation-Based Methods for Inductive Learning, Machine Learning, 4(2): 187–226, 1989.Google Scholar
- 2.H. Hirsh, Theoretical Underpinnings of Version Spaces, In Proc. of the Twelfth Int. Joint Conf. on Artificial Intelligence, IJCAI-91, pp.665–670, Sydney, Australia, 1991.Google Scholar
- 3.H. Hirsh, Generalizing Version Spaces, Machine Learning, 17(1):5–46, 1994.Google Scholar
- 4.E.N. Smirnov, Space Fragmenting-A Method For Disjunctive Concept Acquisition, In B. duBoulay and V. Sgurev, eds., Artificial Intelligence V-Methodology, Systems, Applications, pp.97–104, Elsevier Science Publishers, 1992.Google Scholar
- 5.E.N. Smirnov and N.I. Nikolaev, Multiple Explanation-Based Learning Guided by the Candidate Elimination Algorithm, In P. Jorrand and V. Sgurev, eds., Artificial Intelligence VI-Methodology, Systems, Applications, pp. 247–256, World Scientific Publ., Singapore, 1994.Google Scholar