Efficient and Scalable Induction of Logic Programs using a Deductive Database System
Michel Ferreira, Ricardo Rocha, Tiago Soares and Nuno A. Fonseca
August 2006
Abstract
In this work we present an
user transparent approach to couple
an ILP system with a relational database by using a Deductive Database
(DDB) system. Our proposal uses the DDB engine to transparently
translate the generated hypotheses to SQL statements with only minor
changes to the implementation of the ILP system. By transferring as
much as possible the evaluation of hypotheses to the RDBMS, we show
how this coupled environment provides an excellent framework for the
efficient and scalable execution of ILP algorithms. Being able to
abstract the Prolog to SQL translation by using the DDB, we
concentrate on evaluating several high-level schemes of interaction
between the ILP system and the RDBMS, with different distributions of
work between the logic system and the database system. Our results
indicate that the execution time of ILP algorithms can be effectively
reduced and that the size of the problems solved can be significantly
increased due to a non-memory storage of the data-sets. Best
performance is achieved when we use a scheme that transforms the
hypotheses into an equivalent SQL count query.
Bibtex
@InProceedings{ferreira-ilp06-short,
author = {M. Ferreira and R. Rocha and T. Soares and Nuno A. Fonseca},
title = {{Efficient and Scalable Induction of Logic Programs using a Deductive Database System}},
booktitle = {Proceedings of the 16th International Conference on Inductive Logic
Programming (ILP 2006) - Short Papers},
pages = {74--76},
editor = {S. Muggleton and R. Otero},
month = {August},
year = {2006},
address = {Santiago de Compostela, Spain},
publisher = {University of Corunna},
}
Download Paper
PDF file