Neural Networks for Functional Approximation and System Identification

Abstract

We construct generalized translation networks to approximate uniformly a class of nonlinear, continuous functionals defined on Lp([—1, 1]s) for integer s ≥ 1, 1 ≤ p < ∞, or C([—1, 1]s). We obtain lower bounds on the possible order of approximation for such functionals in terms of any approximation process depending continuously on a given number of parameters. Our networks almost achieve this order of approximation in terms of the number of parameters (neurons) involved in the network. The training is simple and noniterative; in particular, we avoid any optimization such as that involved in the usual backpropagation.

Cite

Text

Mhaskar and Hahm. "Neural Networks for Functional Approximation and System Identification." Neural Computation, 1997. doi:10.1162/NECO.1997.9.1.143

Markdown

[Mhaskar and Hahm. "Neural Networks for Functional Approximation and System Identification." Neural Computation, 1997.](https://mlanthology.org/neco/1997/mhaskar1997neco-neural/) doi:10.1162/NECO.1997.9.1.143

BibTeX

@article{mhaskar1997neco-neural,
  title     = {{Neural Networks for Functional Approximation and System Identification}},
  author    = {Mhaskar, Hrushikesh Narhar and Hahm, Nahmwoo},
  journal   = {Neural Computation},
  year      = {1997},
  pages     = {143-159},
  doi       = {10.1162/NECO.1997.9.1.143},
  volume    = {9},
  url       = {https://mlanthology.org/neco/1997/mhaskar1997neco-neural/}
}