Is Feature Diversity Necessary in Neural Network Initialization?
Abstract
Standard practice in training neural networks involves initializing the weights in an independent fashion. The results of recent work suggest that feature "diversity" at initialization plays an important role in training the network. However, other initialization schemes with reduced feature diversity have also been shown to be viable. In this work, we conduct a series of experiments aimed at elucidating the importance of feature diversity at initialization. We show that a complete lack of diversity is harmful to training, but its effects can be counteracted by a relatively small addition of noise - even the noise in standard non-deterministic GPU computations is sufficient. Furthermore, we construct a deep convolutional network with identical features at initialization and almost all of the weights initialized at 0 that can be trained to reach accuracy matching its standard-initialized counterpart.
Keywords
Cite
@article{arxiv.1912.05137,
title = {Is Feature Diversity Necessary in Neural Network Initialization?},
author = {Yaniv Blumenfeld and Dar Gilboa and Daniel Soudry},
journal= {arXiv preprint arXiv:1912.05137},
year = {2020}
}
Comments
This paper has been substantially modified, updated, and expanded with additional content (arXiv:2007.01038). To avoid confusion, we are withdrawing the old version of this article