A search technique for rule extraction from trained neural networks

Authors
Krishnan, R Sivakumar, G Bhattacharya, P
Citation
R. Krishnan et al., A search technique for rule extraction from trained neural networks, PATT REC L, 20(3), 1999, pp. 273-280
Citations number
11
Categorie Soggetti
AI Robotics and Automatic Control
Journal title
PATTERN RECOGNITION LETTERS
ISSN journal
01678655 → ACNP
Volume
20
Issue
3
Year of publication
1999
Pages
273 - 280
Database
ISI
SICI code
0167-8655(199903)20:3<273:ASTFRE>2.0.ZU;2-A
Abstract
Search methods for rule extraction from neural networks work by finding tho se combinations of inputs that make the neuron active. By sorting the input weights to a neuron and ordering the weights suitably, it is possible to p rune the search space. Based on this observation, we present an algorithm f or rule extraction from feedforward neural networks with boolean inputs and analyze its properties. (C) 1999 Published by Elsevier Science B.V. All ri ghts reserved.