Finite size scaling in neural networks
Disordered Systems and Neural Networks
2009-10-28 v1 adap-org
Adaptation and Self-Organizing Systems
Abstract
We demonstrate that the fraction of pattern sets that can be stored in single- and hidden-layer perceptrons exhibits finite size scaling. This feature allows to estimate the critical storage capacity \alpha_c from simulations of relatively small systems. We illustrate this approach by determining \alpha_c, together with the finite size scaling exponent \nu, for storing Gaussian patterns in committee and parity machines with binary couplings and up to K=5 hidden units.
Keywords
Cite
@article{arxiv.cond-mat/9611027,
title = {Finite size scaling in neural networks},
author = {Walter Nadler and Wolfgang Fink},
journal= {arXiv preprint arXiv:cond-mat/9611027},
year = {2009}
}
Comments
4 pages, RevTex, 5 figures, uses multicol.sty and psfig.sty