On-Line Identification and Rule Extraction of Finite State Automata with Recurrent Neural Networks
The on-line identification of an unknown finite state automaton with a generalized recurrent neural network and an on-line learning scheme, together with an on-line rule extraction algorithm is presented. Several tests were made on different, strongly connected automata with structures ranging between 2 and 32 states and the results of both training and extraction processes are very promising.
KeywordsRecurrent Neural Network Generalize Architecture Rule Extraction Finite State Automaton Deterministic Finite State Automaton
Unable to display preview. Download preview PDF.
- M. Tomita. Dynamic Construction of Finite Automata from Examples Using Hill-Climbing. Proceedings of the 4th Annual Cognitive Science Conference, pp. 105–108, Ann Arbor, MI 1982.Google Scholar
- A. U. Levin K. S. Narendra. Identification of Nonlinear Dynamical Systems Using Neural Networks. Neural Systems for Control, Edited by O. Omidvar D. L. Elliott. Academic Press, pp. 129–160, San Diego, 1997.Google Scholar
- H. T. Su T. Samad. Neuro-Control Design: Optimization Aspects. Neural Systems for Control, Edited by O. Omidvar D. L. Elliott. Academic Press, pp. 259–288, San Diego, 1997.Google Scholar
- I. Gabrijel A. Dobnikar. Adaptive RBF Neural Network. Proceedings of the SOCO’97 Conference Nimes, France, pp. 164–170, September 17-19, 1997.Google Scholar
- I. Gabrijel A. Dobnikar N. Steele. RBF Neural Networks and Fuzzy Logic Based Control — a Case Study. Proceedings of the 2nd IMACS International Multiconference CESA’98, Harnmamet, Tunisia, vol. 2, pp. 284–289, April 1-4, 1998.Google Scholar
- S. Haykin. Neural Networks — A Comprehensive Foundation — Second Edition. Prentice-Hall, Inc., New Jersey, 1999.Google Scholar