Closed YOLO-jbc closed 2 years ago
Hello!! This is a graph classification task and all the samples are divided into 2 categories. I'm trying to use SubgraphX to explain my GCN model on one sample graph. Concretely,
_, explanation_results, related_preds = subgraphx(dataset[0].x, dataset[0].edge_index)
Then I get related_preds like this
related_preds
[{'masked': 0.997955322265625, 'maskout': 0.9543552398681641, 'origin': 0.9994699358940125, 'sparsity': 0.5}, {'masked': 0.0524115152657032, 'maskout': 0.0018024112796410918, 'origin': 0.000530135293956846, 'sparsity': 0.7}]
I am wondering what means 'masked', 'maskout', 'origin'? And how can I get the fidelity of the explaination result? Thanks a lot! ;)
Hello!! This is a graph classification task and all the samples are divided into 2 categories. I'm trying to use SubgraphX to explain my GCN model on one sample graph. Concretely,
Then I get
related_preds
like thisI am wondering what means 'masked', 'maskout', 'origin'? And how can I get the fidelity of the explaination result? Thanks a lot! ;)