PiNet: Attention Pooling for Graph Classification

08/11/2020

∙

by   Peter Meltzer, et al.

∙

11

∙

share

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.