A Novel Liapunov Function for Additive Neural Networks
Using the properties of nonlinear integral equations of the Hammerstein type, it is proven that a new Liapunov function for additive neural networks exists, provided two conditions on symmetry and positivity of the weight matrix hold. The function does not require monotonicity of the transfer function.
KeywordsTransfer Function Symmetric Part Absolute Stability Mathematical Literature Nonlinear Integral Equation
Unable to display preview. Download preview PDF.
- Krasnosel’skii, M., Topological Methods in the Theory of Nonlinear Integral Equations, Pergamon Press, Oxford, 1965Google Scholar
- Rall, L. B., Variational Methods for Nonlinear Integral Equations, in Anselone, P.M., Ed, Nonlinear Integral Equations; Proceedings of an Advanced Seminar at the University of Wisconsin, University of Wisconsin Press, Madison, 1964, pp. 155–189Google Scholar
- Jourjine, A. N., Large n Limit of Additive Neural Networks, Proceedings of Neuro-Nimes ‘92, Nimes, France, November 1992, pp. 635–643Google Scholar