Learning Graph Neural Networks with Noisy Labels

Hoang NT; Choong Jun Jin; Tsuyoshi Murata

Learning Graph Neural Networks with Noisy Labels

Machine Learning 2019-05-07 v1 Machine Learning

Authors: Hoang NT , Choong Jun Jin , Tsuyoshi Murata

Abstract

We study the robustness to symmetric label noise of GNNs training procedures. By combining the nonlinear neural message-passing models (e.g. Graph Isomorphism Networks, GraphSAGE, etc.) with loss correction methods, we present a noise-tolerant approach for the graph classification task. Our experiments show that test accuracy can be improved under the artificial symmetric noisy setting.

Keywords

learning with noisy labels graph representation learning graph neural network

Cite

@article{arxiv.1905.01591,
  title  = {Learning Graph Neural Networks with Noisy Labels},
  author = {Hoang NT and Choong Jun Jin and Tsuyoshi Murata},
  journal= {arXiv preprint arXiv:1905.01591},
  year   = {2019}
}

Comments

5 pages, 4 figures, 3 tables; Appeared as a poster presentation at Limited Labeled Data (LLD) Workshop, ICLR 2019

Related papers

View all related →