EPIA'03 - 11th Portuguese Conference on Artificial Intelligence

	Session: December 5, 17:15-18:45, Room B
Title:	Learning action theories with ramifications
	David Lorenzo
Abstract:	The representation of properties of actions is easier with logic programs due to the non-monotonic character of negation as failure, however for any moderately complex dynamic system, providing the logic program that represents the system behaviour is difficult. A logic programming formalization of action domains has been well-studied and some work exists on the combination with learning methods. We extend previous work to deal with the indirect effects of actions and to solve the problems derived from cyclic dependences. This will provide more expressiveness on the theories that are learned so that it can solve some problems that were unsolvable before.

Back to schedule.