Modify the implementation of sagpool so it is consistent with the description in paper #8562
+99
−3
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The current implementation of sagpool reuses the SelectTopK class, which introduces an extra learnable parameter that learns to take the negative value of the attention score or not.
The top k nodes should be selected by their attentions score attn, but in current implementation, the top k nodes are selected by attn*i/norm(i), where i is the extra learnable parameter. The value of i/norm(i) is 1 if i >0, -1 if i<0. That means when i<0 the current select function will return bottom k nodes.