Fix in PEs compute for full graph experiments #2
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This PR fixes the computation of positional encodings (PEs) for the full graph attention experiments, shown in the main paper (Table 1, Column 'Full Graph').
Due to the bug, for the full graph experiments, the PEs were computed on the fully connected (fully adjacent) graphs, and not the original sparse graphs.
With the correction, the PEs are calculated always on the original sparse graphs, which is the objective for PEs to capture original graph structure (hence positions as well) and inject them into the nodes.
--
Ps. Note that the full graph attention is not what the paper finds best for a graph transformer architecture, and this bug fix does not change the paper's main results, analysis and conclusion. The updated Table 1 will be on arxiv's next version of the paper.
Thanks to @Saro00 for pointing this out.