Relative order defines a topology for recurrent networks
Swanston, D.J., Kambhampati, C., Manchanda, S., Tham, M. and Warwick, K. (1995) Relative order defines a topology for recurrent networks. In: Fourth International Conference on Artificial Neural Networks, 26-28 June 1995, Cambridge, UK, pp. 256-261.
Full text not archived in this repository.
To link to this article DOI: 10.1049/cp:19950564
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.
Repository Staff Only: item control page