9
$\begingroup$

Settings Many algorithms operate on a single relation or table, while many real-world databases store information in multiple tables (Domingos, 2003).

Question What types of algorithms learn well from multiple (relational) tables. In particular, I am interested in the algorithms that are applicable to the regression and classification tasks (not the network analysis oriented ones, e.g. link prediction).


I am aware of several approaches listed bellow (but am sure that I am missing some):

  • Multi-Relational Data Mining (MRDM) (Dzeroski, 2002)
  • Inductive Logic Programming (ILP) (Muggleton, 1992)
  • Statistical Relational Learning (SRL) (Getoor, 2007)

Džeroski, S. (2003). Multi-relational data mining: an introduction. ACM SIGKDD Explorations Newsletter.

Getoor, Lise, and Ben Taskar, eds. Introduction to statistical relational learning. MIT press, 2007.

S. Muggleton and C. Feng. Efficient induction of logic programs. In Proceedings of the First Conference on Algorithmic Learning Theory, pages 368–381. Ohmsha, Tokyo, 1990.

$\endgroup$
2
$\begingroup$

I started studying this subject by reading this paper: Macskassy, S., & Provost, F. (2003). A simple relational classifier. My advisor told me it is the simplest classification approach in relational learning he knows.

$\endgroup$
  • $\begingroup$ Thanks. From first glance the paper seems quite interesting and practical. Will start reading it. $\endgroup$ – Neil Nov 8 '12 at 1:09
2
$\begingroup$

This is a good introduction book: De Raedt, Luc, ed. Logical and relational learning. Springer, 2008.

Try using ACE for TILDE and WARMR.

$\endgroup$
  • $\begingroup$ Could you summarize the main points of that book in relation to the OP? $\endgroup$ – chl Jul 10 '13 at 9:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.