Shift Aggregate Extract Networks

• Published in: Frontiers in robotics and AI Vol. 5; p. 42
• Main Authors: Orsini, Francesco, Baracchi, Daniele, Frasconi, Paolo
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Media S.A 10.04.2018
• Subjects: neural networks; relational learning; representation learning; Robotics and AI; social networks; supervised learning; representation learning; relational learning; social networks; supervised learning; neural networks
• ISSN: 2296-9144; 2296-9144
• DOI: 10.3389/frobt.2018.00042

We introduce an architecture based on deep hierarchical decompositions to learn effective representations of large graphs. Our framework extends classic R-decompositions used in kernel methods, enabling nested part-of-part relations. Unlike recursive neural networks, which unroll a template on input graphs directly, we unroll a neural network template over the decomposition hierarchy, allowing us to deal with the high degree variability that typically characterize social network graphs. Deep hierarchical decompositions are also amenable to domain compression, a technique that reduces both space and time complexity by exploiting symmetries. We show empirically that our approach is able to outperform current state-of-the-art graph classification methods on large social network datasets, while at the same time being competitive on small chemobiological benchmark datasets.