MDL for FCA: is there a place for background knowledge?

Tatiana Makhalova (1, 2) , Sergei O. Kuznetsov (1) , Amedeo Napoli (2)

1 CS-HSE - Faculty of Computer Science [Moscow]
2 ORPAILLEUR - Knowledge representation, reasonning

The Minimal Description Length (MDL) principle is a powerful and well founded approach, which has been successfully applied in a wide range of Data Mining tasks. In this paper we address the problem of pattern mining with MDL. We discuss how constraints-background knowledge on interestingness of patterns-can be embedded into MDL and argue the benefits of MDL over a simple selection of patterns based on measures.

Domaines: Informatique [cs]; Intelligence artificielle [cs.AI]

Tatiana Makhalova, Sergei O. Kuznetsov, Amedeo Napoli. MDL for FCA: is there a place for background knowledge?. IJCAI ECAI 2018 - 6th International Workshop "What can FCA do for Artificial Intelligence?", Jul 2018, Stockholm, Sweden. ⟨hal-01888440⟩