Title :
Lie Algebra of Recurrent Neural Networks and Identifiability
Author :
Zbikowski, Rafal
Author_Institution :
Control Group, Dept. of Mechanical Eng., University of Glasgow, Glasgow G12 8QQ, Scotland
Abstract :
Lie algebra associated with recurrent neural networks is described in the context of the State Isomorphism Theorem approach to identifiability. It is shown that this leads to inconclusive results, due to conservative sufficient conditions involved, which are difficult to verify.
Keywords :
Adaptive control; Algebra; Controllability; Cyclic redundancy check; Equations; Observability; Parametric statistics; Recurrent neural networks; Sufficient conditions; Vectors;
Conference_Titel :
American Control Conference, 1993
Conference_Location :
San Francisco, CA, USA
Print_ISBN :
0-7803-0860-3
