ML Anthology

The Error Surface of the Simplest XOR Network Has Only Global Minima

Ida G. Sprinkhuizen-Kuyper, Egbert J. W. Boers
NeCo 1996 pp. 1301-1320
doi:10.1162/NECO.1996.8.6.1301 /neco/1996/sprinkhuizenkuyper1996neco-error/

Abstract

The artificial neural network with one hidden unit and the input units connected to the output unit is considered. It is proven that the error surface of this network for the patterns of the XOR problem has minimum values with zero error and that all other stationary points of the error surface are saddlepoints. Also, the volume of the regions in weight space with saddlepoints is zero, hence training this network on the four patterns of the XOR problem using, e.g., backpropagation with momentum, the correct solution with error zero will be reached in the limit with probability one.

NeCo Semantic Scholar

Cite

Text

Sprinkhuizen-Kuyper and Boers. "The Error Surface of the Simplest XOR Network Has Only Global Minima." Neural Computation, 1996. doi:10.1162/NECO.1996.8.6.1301

Markdown

[Sprinkhuizen-Kuyper and Boers. "The Error Surface of the Simplest XOR Network Has Only Global Minima." Neural Computation, 1996.](https://mlanthology.org/neco/1996/sprinkhuizenkuyper1996neco-error/) doi:10.1162/NECO.1996.8.6.1301

BibTeX

@article{sprinkhuizenkuyper1996neco-error,
  title     = {{The Error Surface of the Simplest XOR Network Has Only Global Minima}},
  author    = {Sprinkhuizen-Kuyper, Ida G. and Boers, Egbert J. W.},
  journal   = {Neural Computation},
  year      = {1996},
  pages     = {1301-1320},
  doi       = {10.1162/NECO.1996.8.6.1301},
  volume    = {8},
  url       = {https://mlanthology.org/neco/1996/sprinkhuizenkuyper1996neco-error/}
}