BriefGPT.xyz
Feb, 2018
自编码架构中的依赖瓶颈：一项实证研究
"Dependency Bottleneck" in Auto-encoding Architectures: an Empirical Study
HTML
PDF
Denny Wu, Yixiu Zhao, Yao-Hung Hubert Tsai, Makoto Yamada, Ruslan Salakhutdinov
TL;DR
研究通过使用Hilbert-Schmidt独立准则（HSIC）作为依赖度量来测量DNN中图层之间的依赖关系，从而更准确地研究DNN中的信息瓶颈和泛化特性，并在实验中评估了该方法在auto-encoding的应用。
Abstract
Recent works investigated the generalization properties in
deep neural networks
(DNNs) by studying the
information bottleneck
in DNNs. However, the mea- surement of the
→