FIGNN adds per-feature Top-K masking branches to a frozen GNN surrogate, producing variable-specific spatial attributions and error budgets for climate and fluid dynamics forecasts.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
FIGNN: Feature-Specific Interpretability for Graph Neural Network Surrogate Models
FIGNN adds per-feature Top-K masking branches to a frozen GNN surrogate, producing variable-specific spatial attributions and error budgets for climate and fluid dynamics forecasts.