FUTURE forgets training samples from tree ensembles by optimizing sigmoid-smoothed split thresholds and copying them back to the original discrete trees.
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
-
FUTURE: Flexible Unlearning for Tree Ensemble
FUTURE forgets training samples from tree ensembles by optimizing sigmoid-smoothed split thresholds and copying them back to the original discrete trees.