Conditional diffusion models can replace GANs in the conditional randomization test, giving asymptotic type I error control and strong power on synthetic and real benchmarks.
B.; Wang, Y.; Barber, R
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Conditional Diffusion Models Based Conditional Independence Testing
Conditional diffusion models can replace GANs in the conditional randomization test, giving asymptotic type I error control and strong power on synthetic and real benchmarks.