arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2005.06195cs.LGstat.ML
收藏

靶向归一化和动量对ReLU染色的影响

The effect of Target Normalization and Momentum on Dying ReLU

Royal Institute of Technology (KTH)(皇家理工学院(KTH));Örebro University(厄勒布鲁大学)

Isac Arnekvist, J. Frederico Carvalho, Danica Kragic, Johannes A. Stork

摘要:

在神经网络(NN)回归中,使用动量优化参数、归一化数据值以及使用校正线性单元(ReLU)是流行的选择。虽然ReLU很受欢迎,但它们可以崩溃为常数函数并“死亡”,有效地从模型中删除它们的贡献。虽然已知一些缓解措施,但目前对ReLU在优化期间死亡的根本原因知之甚少。在本文中,我们考虑了目标归一化和动量对垂死ReLU的影响。我们发现经验,单位方差的目标是很好的动机和ReLU更容易死,当目标方差接近零。为了进一步研究这一问题,我们分析了一个离散时间线性自治系统,并从理论上展示了这与具有单个ReLU的模型的关系,以及常见属性如何导致ReLU死亡。我们还分析了单个ReLU模型的梯度,以识别对应于死亡ReLU的鞍点和区域,以及当使用动量时参数如何演变到这些区域。最后,我们的经验表明,这个问题仍然存在,并加剧,更深层次的模型,包括残差网络。

英文摘要:

Optimizing parameters with momentum, normalizing data values, and using rectified linear units (ReLUs) are popular choices in neural network (NN) regression. Although ReLUs are popular, they can collapse to a constant function and "die", effectively removing their contribution from the model. While some mitigations are known, the underlying reasons of ReLUs dying during optimization are currently poorly understood. In this paper, we consider the effects of target normalization and momentum on dying ReLUs. We find empirically that unit variance targets are well motivated and that ReLUs die more easily, when target variance approaches zero. To further investigate this matter, we analyze a discrete-time linear autonomous system, and show theoretically how this relates to a model with a single ReLU and how common properties can result in dying ReLU. We also analyze the gradients of a single-ReLU model to identify saddle points and regions corresponding to dying ReLU and how parameters evolve into these regions when momentum is used. Finally, we show empirically that this problem persist, and is aggravated, for deeper models including residual networks.

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。

↑