ShareGNNs implement invariant-indexed weight sharing in an encoder-decoder MPNN, tying expressivity to the chosen invariants and reporting gains on synthetic, real-world, and subgraph counting tasks.
Each dataset is split into 10 predefined train/test folds
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Invariant-Based Weight Sharing for Message Passing
ShareGNNs implement invariant-indexed weight sharing in an encoder-decoder MPNN, tying expressivity to the chosen invariants and reporting gains on synthetic, real-world, and subgraph counting tasks.