Rational Parametrizations of Neural Networks

Abstract

A connection is drawn between rational functions, the realization theory of dynamical systems, and feedforward neural networks. This allows us to parametrize single hidden layer scalar neural networks with (almost) arbitrary analytic activation functions in terms of strictly proper rational functions. Hence, we can solve the uniqueness of parametrization problem for such networks.

Cite

Text

Helmke and Williamson. "Rational Parametrizations of Neural Networks." Neural Information Processing Systems, 1992.

Markdown

[Helmke and Williamson. "Rational Parametrizations of Neural Networks." Neural Information Processing Systems, 1992.](https://mlanthology.org/neurips/1992/helmke1992neurips-rational/)

BibTeX

@inproceedings{helmke1992neurips-rational,
  title     = {{Rational Parametrizations of Neural Networks}},
  author    = {Helmke, Uwe and Williamson, Robert C.},
  booktitle = {Neural Information Processing Systems},
  year      = {1992},
  pages     = {623-630},
  url       = {https://mlanthology.org/neurips/1992/helmke1992neurips-rational/}
}