Paper Info
Title
Rule-based explanation in connectionist networks
Abstract
Connectionist networks are interesting computational models that have been proved to be useful for pattern recognition, adaptive behaviour including machine learning and generalisation, probabilistic and plausible reasoning, classification and many more. Knowledge in connectionist networks is stored in distributed internal weights. Learning algorithms based on numerical optimisation techniques can adapt these weights for a particular task (e.g., pattern classification), While in some applications it may be sufficient to rely on the generalisation capability of a network, others may require that the knowledge be available in symbolic form. It is, therefore, necessary to interpret and understand the knowledge learned by connectionist networks. One of tile major criticisms to connectionist approach, however, is that knowledge generated by neural networks is not explicitly represented in the form of rules suitable for verification or interpretation by human, i.e., they are often regarded as black boxes. In this paper, we will discuss the ability of connectionist networks to transfer the role of learning into knowledge refinement and generate rule-based explanations. It will be shown that the exchange of information between symbolic and connectionist representations is accomplished in a simple recurrent network (SRN) for the task of prediction. Preliminary results show that our method is promising with regard to explaining networks' behaviours.
Year
Venue
Keywords
1997
SCAI '97 Proceedings of the sixth Scandinavian conference on Artificial intelligence
rule-based explanation,connectionist network,rule based
Field
DocType
Volume
Rule-based system,Computer science,Artificial intelligence,Connectionism,Machine learning
Conference
40
ISSN
ISBN
Citations 
0922-6389
90-5199-354-4
0
PageRank 
References 
Authors
0.34
0
2
Name
Order
Citations
PageRank
Xinyu Wu100.34
John G. Hughes232659.84