PiNet: Attention Pooling for Graph Classification
Meltzer, Peter, Marcelo Daniel Gutierrez Mallea, Peter J. Bentley · arXiv (Cornell University) · 2020
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.