Wim VanLaer, Luc Dehaspe, Luc DeRaedt
The clausal discovery engine CLAUDIEN is presented. CLAUDIEN discovers regularities in data and is s representative :of the inductive logic programming paradigm. As such, it represent s data and regu!aritles by means of first order clausal theories. Because the search space of c~ausal theories is larger-than that of attribute value representation, CLAUDIEN alSO accepts as input a declarative specification of the langu~sge bias, which determines the Rt of syntactically well-formed regularities. Whereas other papers on CLAUDIEN fOCUSS on the semantics or logical problem specification of CLAUDIEN, on the discovery algorithm, Or the PAC-learning aspects, this paper wants to illustrate the power of the resulting technique. In order to achieve this aim, we show how CL^UmEN can be used to learn I) integrity conattaints in data.bases, 2) functional dependencies ~nd determinations, 3) properties of sequences, 4) mixed quantitative and qualitative laws, S) reverse engineering, and 6) classification rules.