At the moment, ICL is capable of learning two types theories: in DNF and CNF format. Which one to use can be specified in the settings file S. The default is DNF.
This stand for Disjunctive Normal Form, ie. the theory is a disjunction of conjunctions. Each conjunct consists of a set of positive literals and a set of negative literals. The variables are existentially quantified.
This stands for Conjunctive Normal Form, ie. the theory is a conjunction of disjunctions. Each disjunction can be seen as a clause (positive literals in the body and the negative literals in the head). The variables are quantified universally.
The language bias is a very important part of the ICL learning system (as it is in most learning systems). It allows the user to guide the system by providing an outline of the search space of rules. The user should always include as much information as possible in the bias!
The only bias available at this moment is DLAB, Declaritive LAnguage Bias, which is also used in CLAUDIEN. The specification must be put in the L file.
The dlab_template contains a string of the form 'Neg <-- Pos' with Pos a DLAB specification for all positive literals, and Neg for all negative literals (this is mostly just false).
The dlab_template contains a string of the form 'Head <-- Body' with Body a DLAB specification for all literals in the body, and Head for all literals in the head of the clause.
Copyright 1998, Katholieke Universiteit Leuven, dept. Computerwetenschappen Information provider: KULeuven dept. Computerwetenschappen Comments for the authors: Wim Van Laer Page design: Wim Van Laer URL: http://www.cs.kuleuven.ac.be/language.html |