图上的谱网络和局部连接网络

Dec, 2013

Spectral Networks and Locally Connected Networks on Graphs

Joan Bruna, Wojciech Zaremba, Arthur Szlam, Yann LeCun

TL;DR本文探讨了如何将卷积神经网络推广到更广泛的定义域上。作者提出了两种构造方法，分别基于分层聚类和图拉普拉斯谱。结果表明，在低维图形上，可以实现独立于输入尺寸的卷积层学习，从而实现高效的深度网络结构。

Abstract

convolutional neural networks are extremely efficient architectures in image and audio recognition tasks, thanks to their ability to exploit the