Rule Extraction from Recurrent Neural Networks: A Taxonomy and Review
Neural Computation
Hi-index | 0.00 |
This paper presents the use of a recurrent neuralnetwork to learn the dynamical behavior of the invertedpendulum and from this network to extract a finite stateautomata. Two clustering methods are compared for theautomata extraction: the K-means method, and theconstruction of fuzzy membership functions. It is shownthat the number of states for the fuzzy clustering methodinduces much less states than the K-means method.