arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.09104cs.LGstat.ML
收藏

深度学习中规范化方法的新解释

New Interpretations of Normalization Methods in Deep Learning

Huawei Noah Ark’s Lab(华为诺亚方舟实验室);Xi’an Jiaotong University(西安交通大学)

Jiacheng Sun, Xiangyong Cao, Hanwen Liang, Weiran Huang, Zewei Chen, Zhenguo Li

摘要:

近年来,人们提出了多种归一化方法来帮助训练神经网络,如批量归一化(BN)、层归一化(LN)、权重归一化(WN)、组归一化(GN)等。在本文中,我们首先提出了一个引理来定义一些必要的工具。然后,我们利用这些工具对目前流行的规范化方法进行了深入的分析,得出以下结论:1)大多数规范化方法都可以在一个统一的框架下进行解释,即将预激活或权值规范化到球面上; 2)由于现有的归一化方法大多是尺度不变的,我们可以在去除尺度对称性的球面上进行优化,3)我们证明了使用这些归一化方法进行训练可以使权值的范数增加,这可能会导致对抗脆弱性,因为它放大了攻击。最后,一系列的实验进行了验证这些索赔。

英文摘要:

In recent years, a variety of normalization methods have been proposed to help train neural networks, such as batch normalization (BN), layer normalization (LN), weight normalization (WN), group normalization (GN), etc. However, mathematical tools to analyze all these normalization methods are lacking. In this paper, we first propose a lemma to define some necessary tools. Then, we use these tools to make a deep analysis on popular normalization methods and obtain the following conclusions: 1) Most of the normalization methods can be interpreted in a unified framework, namely normalizing pre-activations or weights onto a sphere; 2) Since most of the existing normalization methods are scaling invariant, we can conduct optimization on a sphere with scaling symmetry removed, which can help stabilize the training of network; 3) We prove that training with these normalization methods can make the norm of weights increase, which could cause adversarial vulnerability as it amplifies the attack. Finally, a series of experiments are conducted to verify these claims.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑