Stability of analogue neural classification networks
- 1 January 1994
- journal article
- research article
- Published by Taylor & Francis in International Journal of Systems Science
- Vol. 25 (1), 19-31
- https://doi.org/10.1080/00207729408928942
Abstract
A general frequency domain stability criterion is presented for autonomous nonlinear dynamical systems that possess several non-linearities and several equilibrium states. As special cases, the result is shown to contain well-known criteria for non-oscillatory behaviour of non-linear feedback loops, of non-linear electrical RLC-circuits and analogue neural networks. The proof relies on a single Liapunov function which can subsequently be used lo compute regions of attraction for each of the equilibrium states. Therefore the result is particularly useful for the analysis and the design of systems such as neural classification networks, which possess many equilibrium states.Keywords
This publication has 5 references indexed in Scilit:
- Neurons with graded response have collective computational properties like those of two-state neurons.Proceedings of the National Academy of Sciences, 1984
- The computation of stability regions for systems with many singular pointsInternational Journal of Control, 1973
- Least squares stationary optimal control and the algebraic Riccati equationIEEE Transactions on Automatic Control, 1971
- On the stability of systems having several equilibrium statesFlow, Turbulence and Combustion, 1969
- A theory of nonlinear networks. IQuarterly of Applied Mathematics, 1964