ML Anthology

Analog Versus Discrete Neural Networks

Bhaskar DasGupta, Georg Schnitger
NeCo 1996 pp. 805-818
doi:10.1162/NECO.1996.8.4.805 /neco/1996/dasgupta1996neco-analog/

Abstract

We show that neural networks with three-times continuously differentiable activation functions are capable of computing a certain family of n-bit Boolean functions with two gates, whereas networks composed of binary threshold functions require at least Ω(log n) gates. Thus, for a large class of activation functions, analog neural networks can be more powerful than discrete neural networks, even when computing Boolean functions.

NeCo Semantic Scholar

Cite

Text

DasGupta and Schnitger. "Analog Versus Discrete Neural Networks." Neural Computation, 1996. doi:10.1162/NECO.1996.8.4.805

Markdown

[DasGupta and Schnitger. "Analog Versus Discrete Neural Networks." Neural Computation, 1996.](https://mlanthology.org/neco/1996/dasgupta1996neco-analog/) doi:10.1162/NECO.1996.8.4.805

BibTeX

@article{dasgupta1996neco-analog,
  title     = {{Analog Versus Discrete Neural Networks}},
  author    = {DasGupta, Bhaskar and Schnitger, Georg},
  journal   = {Neural Computation},
  year      = {1996},
  pages     = {805-818},
  doi       = {10.1162/NECO.1996.8.4.805},
  volume    = {8},
  url       = {https://mlanthology.org/neco/1996/dasgupta1996neco-analog/}
}