English

PiNet: Attention Pooling for Graph Classification

Machine Learning 2020-08-12 v1 Computer Vision and Pattern Recognition Machine Learning

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.

Keywords

Cite

@article{arxiv.2008.04575,
  title  = {PiNet: Attention Pooling for Graph Classification},
  author = {Peter Meltzer and Marcelo Daniel Gutierrez Mallea and Peter J. Bentley},
  journal= {arXiv preprint arXiv:2008.04575},
  year   = {2020}
}

Comments

4 pages, 3 figures 1 table

R2 v1 2026-06-23T17:46:19.598Z