Towards OOD Detection in Graph Classification from Uncertainty Estimation Perspective
This addresses the challenge of detecting out-of-distribution examples in graph classification, which is important for improving model reliability in domains like chemistry or social networks, but it is incremental as it compares existing methods without introducing a new solution.
The paper tackled the problem of out-of-distribution detection in graph classification by comparing several uncertainty estimation methods, finding no universal approach and emphasizing the need to consider both graph representations and predictive distributions.
The problem of out-of-distribution detection for graph classification is far from being solved. The existing models tend to be overconfident about OOD examples or completely ignore the detection task. In this work, we consider this problem from the uncertainty estimation perspective and perform the comparison of several recently proposed methods. In our experiment, we find that there is no universal approach for OOD detection, and it is important to consider both graph representations and predictive categorical distribution.