Open Peer Review. Open Publishing. Open Access. Open Discussion. Open Directory. Open Recommendations. Open API. Open Source.
Spectral Networks and Locally Connected Networks on Graphs
Joan Bruna, Wojciech Zaremba, Arthur Szlam, Yann LeCun
Dec 25, 2013 (modified: Dec 25, 2013)ICLR 2014 conference submissionreaders: everyone
Decision:submitted, no decision
Abstract:Convolutional Neural Networks are extremely efficient architectures in image and audio recognition tasks, thanks to their ability to exploit the local translational invariance of signal classes over their domain. In this paper we consider possible generalizations of CNNs to signals defined on more general domains without the action of a translation group. In particular, we propose two constructions, one based upon a hierarchical clustering of the domain, and another based on the spectrum of the graph Laplacian. We show through experiments that for low-dimensional graphs it is possible to learn convolutional layers with $O(1)$ parameters, resulting in efficient deep architectures.
Enter your feedback below and we'll get back to you as soon as possible.