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 1802.07896cs.AIcs.LG
收藏

L2-Nonexpansive神经网络

L2-Nonexpansive Neural Networks

IBM Research(IBM研究(机构));Yorktown Heights, NY 10598, USA(美国纽约州约克镇高地(地址,非机构))

Haifeng Qian, Mark N. Wegman

摘要:

本文提出了一类条件良好的神经网络,其中输入的单位量的变化最多导致输出或任何内部层的单位量的变化。我们开发了已知的控制Lipschitz常数的方法,以实现其最大限度地提高鲁棒性的全部潜力,与一个新的正则化方案的线性层,新的方式来适应非线性和一个新的损失函数。使用MNIST和CIFAR-10分类器,我们展示了许多优点。在不需要任何对抗性训练的情况下,所提出的分类器在对抗白盒L2界对抗攻击的鲁棒性方面超过了现有技术。它们比普通网络更好地从带有部分随机标签的噪声数据中泛化。它们的产出在数量上是有意义的,并表明置信度和普遍性,以及其他可取的性质。

英文摘要:

This paper proposes a class of well-conditioned neural networks in which a unit amount of change in the inputs causes at most a unit amount of change in the outputs or any of the internal layers. We develop the known methodology of controlling Lipschitz constants to realize its full potential in maximizing robustness, with a new regularization scheme for linear layers, new ways to adapt nonlinearities and a new loss function. With MNIST and CIFAR-10 classifiers, we demonstrate a number of advantages. Without needing any adversarial training, the proposed classifiers exceed the state of the art in robustness against white-box L2-bounded adversarial attacks. They generalize better than ordinary networks from noisy data with partially random labels. Their outputs are quantitatively meaningful and indicate levels of confidence and generalization, among other desirable properties.

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。

↑