GOLEM

Name	GOLEM
Description	The algorithm GOLEM (Muggleton) uses inverse resolution and least general generalization to induce a set of horn clauses from a set of positive examples, restricted to ground facts and background knowledge, given in the form of extensionally defined predicates (ground facts, e.g. relational databases). The base procedure: Randomly select two (positive) examples from the given example set. For both examples: Call the procedure "inverse resolution" to receive a horn clause that - together with the background knowledge - implies the example. For the inverse resolution the most specific inverse substitution (namely "∅") is used. Combine the two clauses generated in the last step by using least general generalization. The result is a horn clause that, together with the background knowledge, implies (at least) the two chosen examples.
Specialization	GOLEM - Software
Generalization	Bottom-Up Induction of Horn Clauses
Example Languages	Ground Facts
Hypothesis Language	Restricted First Order Logic
Dm Step	Characterization (Descriptive Setting) Concept Learning
Method Type	Algorithm
Theories	Inductive Logic Programming (ILP)