Assumptions Behind GNN Explainers | How Does Explanation Become Computationally Possible?

[AnkitSSaxena]

https://youtu.be/B4E7ir--34M

'Assumptions' play a very important role in explainer tools. They make the explanation process computationally feasible and shape what an explainer can (and cannot) reveal. There are several assumptions an explainer can adopt. Of course, the right assumptions always depend on the explainer’s goal.

In my new YouTube video, I break down the most common assumptions used by explainer tools for Graph Neural Networks (GNNs).

Do watch and share your thoughts in the comments.

https://youtu.be/B4E7ir--34M