KPP is a new representation for tree ensembles that indexes features by nodes with a path metric, yielding a non-diagonal Gram matrix that unifies prediction, exact additive attribution, deterministic Lipschitz robust radius, and uniform Rademacher risk bounds under three conditioning regimes.
Minimax optimal rates for Mondrian trees and forests
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Kernel of Partition Paths: A Unified Representation for Tree Ensembles
KPP is a new representation for tree ensembles that indexes features by nodes with a path metric, yielding a non-diagonal Gram matrix that unifies prediction, exact additive attribution, deterministic Lipschitz robust radius, and uniform Rademacher risk bounds under three conditioning regimes.