论文信息 - Determination of General Concept in Learning Default Rules

Determination of General Concept in Learning Default Rules

In this paper, we discuss a method to determine which concept, the target concept or its opposite, is more general in given examples when learning rules with exceptions, or default rules in the Inductive Logic Programming (ILP) framework. The ILP system learning default rules has to learn both the concepts in a three valued setting which clearly distinguishes what is true, what is false, and what is unknown. In order to learn hypotheses which holds as generally as possible in the domain, we should give a higher priority to the concept which is more general, or covers more examples than does the other. For this purpose, our method dynamically determines the general concept according to the ratio of positive examples covered by the hypothesis which is correct and most general in the hypothesis space.

[1] Stefan Wrobel,et al. Proceedings of the 8th European Conference on Machine Learning , 1995 .

[2] Giancarlo Ruffo,et al. Learning Logic Programs with Negation as Failure , 1996 .

[3] Luc De Raedt,et al. On Negation and Three-Valued Logic in Interactive Concept-Learning , 1990, ECAI.

[4] De Raedt,et al. Advances in Inductive Logic Programming , 1996 .

[5] Antonis C. Kakas,et al. Learning Non-Monotonic Logic Programs: Learning Exceptions , 1995, ECML.

[6] Saso Dzeroski,et al. Inductive Logic Programming: Techniques and Applications , 1993 .

[7] Noboru Babaguchi,et al. Non-monotonic inference system handling knowledge allowing classified exceptions , 1998, SMC'98 Conference Proceedings. 1998 IEEE International Conference on Systems, Man, and Cybernetics (Cat. No.98CH36218).