arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1802.07088cs.LGcs.CVstat.ML
收藏

i-RevNet:深度可逆网络

i-RevNet: Deep Invertible Networks

University of Tübingen(图宾根大学);CentraleSupélec(中央理工-高等电力学院);Université Paris-Saclay(巴黎-萨克雷大学);INRIA Saclay(法国国家信息与自动化研究所萨克雷中心);INRIA Lille(法国国家信息与自动化研究所里尔中心);ENS, Université PSL(巴黎文理研究大学高等师范学院);University of Amsterdam(阿姆斯特丹大学)

Jörn-Henrik Jacobsen, Arnold Smeulders, Edouard Oyallon

摘要:

人们普遍认为,深度卷积网络的成功是基于逐步丢弃关于手头问题的输入的无信息可变性。在大多数常用的网络架构中,从其隐藏表示中恢复图像的难度从经验上支持这一点。在本文中,我们通过一对一的映射表明,这种信息丢失并不是学习在复杂问题(如ImageNet)上推广良好的表示的必要条件。通过级联同胚层,我们构建了i-RevNet,这是一个可以完全反转到最终投影到类上的网络,即没有信息被丢弃。建立一个可逆的体系结构是困难的,因为局部逆是病态的,我们通过提供一个显式的逆来克服这一点。对i-RevNets学习表示的分析表明,深度网络的成功是通过渐进收缩和深度线性分离来实现的。为了揭示i-RevNet学习的模型的本质,我们在自然图像表示之间重建线性插值。

英文摘要:

It is widely believed that the success of deep convolutional networks is based on progressively discarding uninformative variability about the input with respect to the problem at hand. This is supported empirically by the difficulty of recovering images from their hidden representations, in most commonly used network architectures. In this paper we show via a one-to-one mapping that this loss of information is not a necessary condition to learn representations that generalize well on complicated problems, such as ImageNet. Via a cascade of homeomorphic layers, we build the i-RevNet, a network that can be fully inverted up to the final projection onto the classes, i.e. no information is discarded. Building an invertible architecture is difficult, for one, because the local inversion is ill-conditioned, we overcome this by providing an explicit inverse. An analysis of i-RevNets learned representations suggests an alternative explanation for the success of deep networks by a progressive contraction and linear separation with depth. To shed light on the nature of the model learned by the i-RevNet we reconstruct linear interpolations between natural image representations.

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。

↑