edge_mask=gnnexplainer_related_preds[1][prediction], NameError: name 'prediction' is not defined

[scottdd204]

See tutorials/KDD2022/xgraph_code_tutorial.ipynb

gnnexplainer_related_preds = \ gnnexplainer(data.x, data.edge_index, sparsity=sparsity, num_classes=num_classes, node_idx=node_idx) ax, G = gnnexplainer.visualize_graph(node_idx=node_idx, edge_index=data.edge_index, edge_mask=gnnexplainer_related_preds[1][prediction], y=data.y)

And therefore there is no example of visualisation for GNNExplainer.

[Oceanusity]

Hi, the prediction results is missed, and should add codes node_idx = node_indices[20], logits = model(data.x, data.edge_index) and prediction = logits[node_idx].argmax(-1).item().

Thanks for pointing it out.

[Oceanusity]

Or you can run the visualization for subgraphx first, and then you will have the prediction variable.