Gradient estimation of probabilistic programs reduces soundly to probabilistic inference after programmable coupling and factorization, enabling new low-variance estimators that beat baselines.
Title resolution pending
2 Pith papers cite this work, alongside 21 external citations. Polarity classification is still indexing.
2
Pith papers citing it
21
external citations · OpenAlex
fields
cs.PL 2representative citing papers
CHAD is a homomorphic source-to-source transformation for forward- and reverse-mode AD on higher-order functional languages with arrays, proven correct via compositional logical relations.
citing papers explorer
-
GradInf: Gradient Estimation as Probabilistic Inference
Gradient estimation of probabilistic programs reduces soundly to probabilistic inference after programmable coupling and factorization, enabling new low-variance estimators that beat baselines.
-
CHAD: Combinatory Homomorphic Automatic Differentiation
CHAD is a homomorphic source-to-source transformation for forward- and reverse-mode AD on higher-order functional languages with arrays, proven correct via compositional logical relations.