GNN Explainer for Regression Task

Questions
#1

Hi,
I want to use GNNExplainer for a Regression Task, using DGL Explainer but I didn’t find an example.
Examples available are dealing only with node classification, could you please help with this?
Thanks

#2

As explained here (GNNExplainer — DGL 2.2 documentation), GNNExpainer generates an explanation by learning an edge mask and a feature mask by optimizing the following objective function:

l(y, \hat{y}) + \alpha_1 \|M\|_1 + \alpha_2 H(M) + \beta_1 \|F\|_1 + \beta_2 H(F)

You can modify the first item to fit the regression task, e.g., replacing it with the MSE loss.

#3

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.