Extracting reduced logic programs from artificial neural networks

DMCA

Extracting reduced logic programs from artificial neural networks (2010)

Cached

Download Links

by Jens Lehmann , Sebastian Bader , Pascal Hitzler

Venue:	Applied Intelligence
Citations:	7 - 2 self

BibTeX

@ARTICLE{Lehmann10extractingreduced,
    author = {Jens Lehmann and Sebastian Bader and Pascal Hitzler},
    title = {Extracting reduced logic programs from artificial neural networks},
    journal = {Applied Intelligence},
    year = {2010},
    pages = {249--266}
}

OpenURL

Abstract

Artificial neural networks can be trained to perform excellently in many application areas. Whilst they can learn from raw data to solve sophisticated recognition and analysis problems, the acquired knowledge remains hidden within the network architecture and is not readily accessible for analysis or further use: Trained networks are black boxes. Recent research efforts therefore investigate the possibility to extract symbolic knowledge from trained networks, in order to analyze, validate, and reuse the structural insights gained implicitly during the training process. In this paper, we will study how knowledge in form of propositional logic programs can be obtained in such a way that the programs are as simple as possible — where simple is being understood in some clearly defined and meaningful way. 1 1

Keyphrases

artificial neural network logic program meaningful way symbolic knowledge structural insight trained network black box network architecture training process raw data analysis problem many application area recent research effort sophisticated recognition propositional logic program

DMCA