Explainability Techniques for Graph Convolutional Networks

Federico Baldassarre; Hossein Azizpour

Explainability Techniques for Graph Convolutional Networks

Machine Learning 2019-06-03 v1 Artificial Intelligence Machine Learning

Authors: Federico Baldassarre , Hossein Azizpour

Abstract

Graph Networks are used to make decisions in potentially complex scenarios but it is usually not obvious how or why they made them. In this work, we study the explainability of Graph Network decisions using two main classes of techniques, gradient-based and decomposition-based, on a toy dataset and a chemistry task. Our study sets the ground for future development as well as application to real-world problems.

Keywords

graph generation graph neural network complex network analysis

Cite

@article{arxiv.1905.13686,
  title  = {Explainability Techniques for Graph Convolutional Networks},
  author = {Federico Baldassarre and Hossein Azizpour},
  journal= {arXiv preprint arXiv:1905.13686},
  year   = {2019}
}

Comments

Accepted at the ICML 2019 Workshop "Learning and Reasoning with Graph-Structured Representations" (poster + spotlight talk)

Related papers

View all related →