Title :
Relative order defines a topology for recurrent networks
Author :
Swanston, D.J. ; Kambhampati, C. ; Manchanda, S. ; Tham, Mau-Luen ; Warwick, K.
Author_Institution :
Reading Univ., UK
Abstract :
This paper uses techniques from control theory in the analysis of trained recurrent neural networks. Differential geometry is used as a framework, which allows the concept of relative order to be applied to neural networks. Any system possessing finite relative order has a left-inverse. Any recurrent network with finite relative order also has an inverse, which is shown to be a recurrent network
Keywords :
differential geometry; learning (artificial intelligence); neural net architecture; neurocontrollers; recurrent neural nets; Hopfield network; control theory; differential geometry; finite relative order; left inverse; neural network architecture; neural network training; neurocontrol; recurrent network topology; recurrent neural networks; relative order;
Conference_Titel :
Artificial Neural Networks, 1995., Fourth International Conference on
Conference_Location :
Cambridge
Print_ISBN :
0-85296-641-5
DOI :
10.1049/cp:19950564
