English

Multifractal Analysis of the Coupling Space of Feed-Forward Neural Networks

Condensed Matter 2009-10-28 v1 q-bio

Abstract

Random input patterns induce a partition of the coupling space of feed-forward neural networks into different cells according to the generated output sequence. For the perceptron this partition forms a random multifractal for which the spectrum f(α)f(\alpha) can be calculated analytically using the replica trick. Phase transition in the multifractal spectrum correspond to the crossover from percolating to non-percolating cell sizes. Instabilities of negative moments are related to the VC-dimension.

Keywords

Cite

@article{arxiv.cond-mat/9512176,
  title  = {Multifractal Analysis of the Coupling Space of Feed-Forward Neural Networks},
  author = {A. Engel and M. Weigt},
  journal= {arXiv preprint arXiv:cond-mat/9512176},
  year   = {2009}
}

Comments

10 pages, Latex, submitted to PRL