ML Anthology

On Langevin Updating in Multilayer Perceptrons

Thorsteinn S. Rögnvaldsson
NeCo 1994 pp. 916-926
doi:10.1162/NECO.1994.6.5.916 /neco/1994/rognvaldsson1994neco-langevin/

Abstract

The Langevin updating rule, in which noise is added to the weights during learning, is presented and shown to improve learning on problems with initially ill-conditioned Hessians. This is particularly important for multilayer perceptrons with many hidden layers, that often have ill-conditioned Hessians. In addition, Manhattan updating is shown to have a similar effect.

NeCo Semantic Scholar

Cite

Text

Rögnvaldsson. "On Langevin Updating in Multilayer Perceptrons." Neural Computation, 1994. doi:10.1162/NECO.1994.6.5.916

Markdown

[Rögnvaldsson. "On Langevin Updating in Multilayer Perceptrons." Neural Computation, 1994.](https://mlanthology.org/neco/1994/rognvaldsson1994neco-langevin/) doi:10.1162/NECO.1994.6.5.916

BibTeX

@article{rognvaldsson1994neco-langevin,
  title     = {{On Langevin Updating in Multilayer Perceptrons}},
  author    = {Rögnvaldsson, Thorsteinn S.},
  journal   = {Neural Computation},
  year      = {1994},
  pages     = {916-926},
  doi       = {10.1162/NECO.1994.6.5.916},
  volume    = {6},
  url       = {https://mlanthology.org/neco/1994/rognvaldsson1994neco-langevin/}
}