Pretrained GNN embeddings transfer across molecular graphs only under certain data-regime and feature-similarity conditions, and a proposed feature-structuralization method does not consistently improve transfer.
These rep- resentations are linearly transformed by component (d), and then summed along with the encoding of the explicit, original node features in output from component (e)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Towards Foundation Models on Graphs: An Analysis on Cross-Dataset Transfer of Pretrained GNNs
Pretrained GNN embeddings transfer across molecular graphs only under certain data-regime and feature-similarity conditions, and a proposed feature-structuralization method does not consistently improve transfer.