GKCM generalizes kernel CI testing to arbitrary regression models, provides uniform asymptotic level guarantees under stated conditions, and outperforms state-of-the-art methods in simulations when using tree-based regressors.
In particular, in settings 1 and 2, KCIT, GKCM KRR and GKCM RF now perform comparably (for most sample sizes with a small lead for the former)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
The Generalised Kernel Covariance Measure
GKCM generalizes kernel CI testing to arbitrary regression models, provides uniform asymptotic level guarantees under stated conditions, and outperforms state-of-the-art methods in simulations when using tree-based regressors.