16 December 1992 Improving neural network performance by adapting node nonlinearities
Author Affiliations +
Abstract
It is known that using an infinite number of hidden layer nodes feedforward neural networks can approximate any continuous function with compact support arbitrarily well using very simple node nonlinearities. We investigate whether network architectures can be found that use more complicated node nonlinearities to achieve better approximation using a restricted number of nodes. Two methods are proposed, one based on modifying standard backpropagation networks, and one based on Kolmogorov's theorem. The feasibility of these networks is evaluated by considering their performance when predicting chaotic time series and memorizing the XOR mapping.
© (1992) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Frans M. Coetzee, Virginia L. Stonick, "Improving neural network performance by adapting node nonlinearities", Proc. SPIE 1766, Neural and Stochastic Methods in Image and Signal Processing, (16 December 1992); doi: 10.1117/12.130830; https://doi.org/10.1117/12.130830
PROCEEDINGS
12 PAGES


SHARE
RELATED CONTENT

Remote discrimination of clouds using a neural network
Proceedings of SPIE (December 16 1992)
Evolving neural network architecture
Proceedings of SPIE (December 16 1992)
Dimensionality reduction for nonlinear time series
Proceedings of SPIE (December 16 1992)
Pattern recognition by cooperating neural networks
Proceedings of SPIE (December 16 1992)

Back to Top