UNI-MB - logo
UMNIK - logo
 
(UM)
  • Multiple predicate learning
    De Raedt, Luc ; Lavrač, Nada ; Džeroski, Sašo
    The problem with learning multiple predicates is that they are frequently not independent. In this case, some predicate definitions have to use others, which implies that the order in which clauses ... for different predicates are learned is important. In addition, an overgeneralized definition of one predicate can cause severe problems when learning the definitions of predicates that depend on it. While interactive state-of-the-art inductive logic programming (ILP) systems address this problem, empirical ILP systems largely ignore this problem or avoid it by working extensionally, using be learned. The paper identifies the problems with the extensional approaches and presents an empirical ILP systems for learning multiple predicates that effectively deals with the above problems.
    Vir: Proceedings of the Third International Workshop on Inductive Logic Programming [being also] ILP'93 : April 1-3, 1993, Bled Slovenija (Str. 221-240)
    Vrsta gradiva - prispevek na konferenci
    Leto - 1993
    Jezik - angleški
    COBISS.SI-ID - 2718230

Vnos na polico