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 1811.05850cs.LGstat.ML
收藏

Drop-Activation:隐式参数约简与调和正则化

Drop-Activation: Implicit Parameter Reduction and Harmonic Regularization

Purdue University(普渡大学);The University of Chicago(芝加哥大学);National University of Singapore(新加坡国立大学)

Senwei Liang, Yuehaw Khoo, Haizhao Yang

摘要:

过度拟合经常发生在深度学习中。在本文中,我们提出了一种新的正则化方法称为Drop激活,以减少过拟合和提高泛化能力。其关键思想是通过在训练时间内将非线性激活函数随机设置为恒等函数来丢弃非线性激活函数。在测试过程中,我们使用了一个确定性网络与一个新的激活函数编码的平均效果随机丢弃激活。我们的理论分析支持Drop-Activation作为隐式参数减少的正则化效应,并验证了其与批量归一化一起使用的能力(Ioffe和Szegedy 2015)。在CIFAR-10、CIFAR-100、SVHN、EMNIST和ImageNet上的实验结果表明,Drop-Activation通常可以提高图像分类任务的流行神经网络架构的性能。此外,作为正则化器,Drop-Activation可以与标准训练和正则化技术(如批量归一化和自动增强)协调使用(Cubuk et al. 2019)。代码可在\url{https://github.com/LeungSamWai/Drop-Activation}上获得。

英文摘要:

Overfitting frequently occurs in deep learning. In this paper, we propose a novel regularization method called Drop-Activation to reduce overfitting and improve generalization. The key idea is to drop nonlinear activation functions by setting them to be identity functions randomly during training time. During testing, we use a deterministic network with a new activation function to encode the average effect of dropping activations randomly. Our theoretical analyses support the regularization effect of Drop-Activation as implicit parameter reduction and verify its capability to be used together with Batch Normalization (Ioffe and Szegedy 2015). The experimental results on CIFAR-10, CIFAR-100, SVHN, EMNIST, and ImageNet show that Drop-Activation generally improves the performance of popular neural network architectures for the image classification task. Furthermore, as a regularizer Drop-Activation can be used in harmony with standard training and regularization techniques such as Batch Normalization and Auto Augment (Cubuk et al. 2019). The code is available at \url{https://github.com/LeungSamWai/Drop-Activation}.

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。

↑