Sample Size Requirements for Feedforward Neural Networks
Abstract
We estimate the number of training samples required to ensure that the performance of a neural network on its training data matches that obtained when fresh data is applied to the network. Existing estimates are higher by orders of magnitude than practice indicates. This work seeks to narrow the gap between theory and practice by transforming the problem into determining the distribution of the supremum of a random field in the space of weight vectors, which in turn is attacked by application of a recent technique called the Poisson clumping heuristic.
Cite
Text
Turmon and Fine. "Sample Size Requirements for Feedforward Neural Networks." Neural Information Processing Systems, 1994.Markdown
[Turmon and Fine. "Sample Size Requirements for Feedforward Neural Networks." Neural Information Processing Systems, 1994.](https://mlanthology.org/neurips/1994/turmon1994neurips-sample/)BibTeX
@inproceedings{turmon1994neurips-sample,
title = {{Sample Size Requirements for Feedforward Neural Networks}},
author = {Turmon, Michael J. and Fine, Terrence L.},
booktitle = {Neural Information Processing Systems},
year = {1994},
pages = {327-334},
url = {https://mlanthology.org/neurips/1994/turmon1994neurips-sample/}
}