Timezone: »

 
A Graph Is More Than Its Nodes: Towards Structured Uncertainty-Aware Learning on Graphs
Hans Hao-Hsun Hsu · Yuesong Shen · Daniel Cremers

@
in New Frontiers in Graph Learning »
Event URL: https://openreview.net/forum?id=_4f6IaxMJeC »

Current graph neural networks (GNNs) that tackle node classification on graphs tend to only focus on nodewise scores and are solely evaluated by nodewise metrics. This limits uncertainty estimation on graphs since nodewise marginals do not fully characterize the joint distribution given the graph structure. In this work, we propose novel edgewise metrics, namely the edgewise expected calibration error (ECE) and the agree/disagree ECEs, which provide criteria for uncertainty estimation on graphs beyond the nodewise setting. Our experiments demonstrate that the proposed edgewise metrics can complement the nodewise results and yield additional insights. Moreover, we show that GNN models which consider the structured prediction problem on graphs tend to have better uncertainty estimations, which illustrates the benefit of going beyond the nodewise setting.

Keywords: Structured Prediction · graph neural networks · Uncertainty

Author Information

Hans Hao-Hsun Hsu (Technische Universität München)
Yuesong Shen (Technical University of Munich)
Daniel Cremers (Technical University of Munich)

More from the Same Authors