Local sample weighting inside random-forest splits and neural-network mini-batches sharpens feature importance under feature correlation and improves out-of-distribution accuracy in simulations.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Decorrelated feature importance from local sample weighting
Local sample weighting inside random-forest splits and neural-network mini-batches sharpens feature importance under feature correlation and improves out-of-distribution accuracy in simulations.