Abstract
Artificial neural networks will be more widely accepted as standard engineering tools if their reasoning process can be made less opaque. This paper describes NetQuery, an explanation mechanism that extracts explanations from trained Radial Basis Function (RBF) networks. Standard RBF networks are modified to identify dependencies between the inputs, to be sparsely connected, and to have an easily interpretable output layer. Given these modifications, the network architecture can be used to extract “Why?” and “Why not?” explanations from the network in terms of excitatory and inhibitory inputs and their linear relationships, greatly simplified by a run-time pruning algorithm. These query results are validated by creating an expert system based on the explanations. NetQuery is also able to inform a user about a possible change in category for a given pattern by responding to a “How can I..?” query. This kind of query is extremely useful when analyzing the quality of a pattern set.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Hinton, G.E.: Learning Distributed Representations of Concepts. In: Proc. Annual Conference Cognitive Science Society, p. 1 (1986)
Thrun, S.B.: Extracting Rules From Artificial Neural Networks With Distributed Representation. In: Tesauro, G., Touretzky, D.S., Leen, T.K. (eds.) Advances in Neural Information Processing Systems 7, pp. 505–512. Morgan Kaufmann, San Mateo (1996)
Fu, L.: Rule Learning by Searching on Adapted Nets. In: Proc. 9th National Conference on Artificiallntelligence, p. 590 (1991)
Craven, M.W., Shavlik, J.W.: Extracting Tree-Structured Representations of trained networks. In: Touretzky, D.S., Mozer, M.C., Hasselmo, M.E. (eds.) Advances in Neural Information Processing Systems 8, pp. 24–30. MIT Press, Cambridge
Framling, K.: Explaining Results of Neural Networks by Contextual Importance and Utility. In: Proc. Rule Extraction From Trained Artificial Neural Networks Workshop, p. 43. Univ. of Sussex, Brighton (1996)
Gallant, S.I.: Connectionist Expert Systems. Communications of the ACM 31, 152–169 (1988)
Garson, D.G.: Interpreting Neural-Network Connection Weights. AI Expert, 47–51 (1991)
Srivastava, R.P.: Automating Judgmental Decisions Using Neural Networks: A Model for Processing Business Loan Applications. In: Proc. 1992 ACM Computer Science Conference, p. 351 (1992)
Moody, J., Darken, C.: Fast learning in a network of locally-tuned processing units. Neural Computation 1, 281–294 (1989)
Augusteijn, M.F., Shaw, K.A.: Radical Pruning: A Method to Construct Skeleton Radial Basis Function Networks. Int. J. Neural Systems (under review)
Le Cun, Y., Denker, J., Solla, S.: Optimal Brain Damage. In: Touretzky, D.S. (ed.) Advances in Neural Information Processing Systems 2, pp. 598–605 (1990)
Mozer, M.C., Smolensky, P.: Skeletonization: A Technique for Trimming the Fat from a Network via Relevance Assessment. In: Touretzky, D.S. (ed.) Advances in Neural Information Processing Systems 1, pp. 107–115 (1989)
Dickerson, J., Kosko, B.: Fuzzy Function Learning with Covariance Ellipsoids. In: IEEE Int. Conf. Neural Networks, vol. 3, p. 1162 (1993)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2000 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Augusteijn, M.F., Shaw, K. (2000). Constructing a Query Facility for RBF Networks. In: Cairó, O., Sucar, L.E., Cantu, F.J. (eds) MICAI 2000: Advances in Artificial Intelligence. MICAI 2000. Lecture Notes in Computer Science(), vol 1793. Springer, Berlin, Heidelberg. https://doi.org/10.1007/10720076_34
Download citation
DOI: https://doi.org/10.1007/10720076_34
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-67354-5
Online ISBN: 978-3-540-45562-2
eBook Packages: Springer Book Archive