Gradient estimation of probabilistic programs reduces soundly to probabilistic inference after programmable coupling and factorization, enabling new low-variance estimators that beat baselines.
In Proceedings of the 26th ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL) , Andrew W
2 Pith papers cite this work, alongside 415 external citations. Polarity classification is still indexing.
2
Pith papers citing it
415
external citations · external index
fields
cs.PL 2representative citing papers
Extends logical relations to recursive session types for PSNI, proves soundness/completeness via biorthogonality with observation-index stratification, and gives an IFC refinement type system with secrecy polymorphism.
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.
-
Logical Relations for Session-Typed Concurrency
Extends logical relations to recursive session types for PSNI, proves soundness/completeness via biorthogonality with observation-index stratification, and gives an IFC refinement type system with secrecy polymorphism.