PiNet: Attention Pooling for Graph Classification

Abstract

We propose PiNet, a generalised differentiable attention-based pooling mechanism for utilising graph convolution operations for graph level classification. We demonstrate high sample efficiency and superior performance over other graph neural networks in distinguishing isomorphic graph classes, as well as competitive results with state of the art methods on standard chemo-informatics datasets.

4 pages, 3 figures 1 table

References in corpus (3)

• Simplifying Graph Convolutional Networks
• Predicting multicellular function through multi-layer tissue networks
• SoundNet: Learning Sound Representations from Unlabeled Video