Skip to content

Conversation

@vict0rsch
Copy link
Collaborator

@vict0rsch vict0rsch commented Jun 15, 2022

  • verify importance of tag 0 nodes in prediction
  • remove all tag 0 nodes
  • one super node per graph
  • one supernode per atom type
  • one supernode per atom type with min distance
  • positional embeddings

sn_pos = [b.pos[sub_nodes[i]].mean(0) for i in range(batch_size)]
# target relaxed position is the mean of the super node's aggregates
# (per batch)
sn_pos_relaxed = [
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same as sn_pos since tag-0 nodes are fixed. No need to compute it.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

indeed ; very small overhead though

@RolnickLab RolnickLab deleted a comment from vict0rsch Jun 22, 2022
@AlexDuvalinho AlexDuvalinho marked this pull request as ready for review July 13, 2022 15:13
@vict0rsch vict0rsch merged commit 4ca9b47 into main Sep 16, 2022
@vict0rsch vict0rsch deleted the supernode branch September 16, 2022 17:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants