Deep Convolutional Networks on Graph-Structured Data

TL;DR

This paper extends spectral networks to graph-structured data, enabling deep learning on non-Euclidean domains with less complexity and improved performance in large-scale classification tasks.

Contribution

It introduces a graph estimation-based extension of spectral networks for deep learning on non-Euclidean data, reducing parameter count and improving accuracy.

Findings

Matches or outperforms Dropout Networks in classification accuracy

Uses fewer parameters than traditional methods

Effective on large-scale graph data

Abstract

Deep Learning's recent successes have mostly relied on Convolutional Networks, which exploit fundamental statistical properties of images, sounds and video data: the local stationarity and multi-scale compositional structure, that allows expressing long range interactions in terms of shorter, localized interactions. However, there exist other important examples, such as text documents or bioinformatic data, that may lack some or all of these strong statistical regularities. In this paper we consider the general question of how to construct deep architectures with small learning complexity on general non-Euclidean domains, which are typically unknown and need to be estimated from the data. In particular, we develop an extension of Spectral Networks which incorporates a Graph Estimation procedure, that we test on large-scale classification problems, matching or improving over Dropout Networks with far less parameters to estimate.