Why Do Deep Residual Networks Generalize Better than Deep Feedforward Networks? - A Neural Tangent Kernel Perspective
Why Do Deep Residual Networks Generalize Better than Deep Feedforward Networks? - A Neural Tangent Kernel Perspective
复制标题
DOI：
--
复制
发表时间：
2020-02
期刊：
影响因子：
--
通讯作者：
Kaixuan Huang;Yuqing Wang-;Molei Tao;T. Zhao
中科院分区：
文献类型：
--
作者：
Kaixuan Huang;Yuqing Wang-;Molei Tao;T. Zhao
深度残差网络（ResNets）比深度前馈网络（FFNets）具有更好的泛化性能。然而，这种现象背后的理论在很大程度上仍然是未知的。本文从所谓的"神经切核"的角度研究了深度学习中的这个基本问题。具体来说，我们首先证明，在适当的条件下，随着宽度趋于无穷大，训练深度ResNets可以被视为学习具有某些核函数的再生核函数。然后，我们将深度ResNets的内核与深度FFNets的内核进行比较，发现由FFNets的内核诱导的函数类是渐进不可学习的，因为深度是无穷大的。相比之下，由ResNets的内核诱导的函数类并没有表现出这种退化。我们的发现部分证明了深度ResNets在泛化能力方面优于深度FFNets。数值结果支持我们的主张。
Deep residual networks (ResNets) have demonstrated better generalization performance than deep feedforward networks (FFNets). However, the theory behind such a phenomenon is still largely unknown. This paper studies this fundamental problem in deep learning from a so-called "neural tangent kernel" perspective. Specifically, we first show that under proper conditions, as the width goes to infinity, training deep ResNets can be viewed as learning reproducing kernel functions with some kernel function. We then compare the kernel of deep ResNets with that of deep FFNets and discover that the class of functions induced by the kernel of FFNets is asymptotically not learnable, as the depth goes to infinity. In contrast, the class of functions induced by the kernel of ResNets does not exhibit such degeneracy. Our discovery partially justifies the advantages of deep ResNets over deep FFNets in generalization abilities. Numerical results are provided to support our claim.