Symbolic conditioning of arrays in probabilistic programs
Authors:
Open Access
- 29 August 2017
- journal article
- research article
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Programming Languages
- Vol. 1 (ICFP) , 1-25
Abstract
Probabilistic programming systems make machine learning more modular by automatinginference. Recent work by Shan and Ramsey makes inference more modular by automatingconditioning. Their technique uses a symbolic program transformation that treats conditioning generally via the measure-theoretic notion ofdisintegration. This technique, however, is limited to conditioning a single scalar variable. As a step towards modular inference for realistic machine learning applications, we have extended the disintegration algorithm to symbolically condition arrays in probabilistic programs. The extended algorithm implementslifted disintegration, where repetition is treated symbolically and without unrolling loops. The technique uses a language ofindex variablesfor tracking expressions at various array levels. We find that the method works well for arbitrarily-sized arrays of independent random choices, with the conditioning step taking time linear in the number of indices needed to select an element.Keywords
Funding Information
- DARPA (FA8750-14-2-0007)
- NSF (CNS-0723054)
This publication has 15 references indexed in Scilit:
- PSI: Exact Symbolic Inference for Probabilistic Programs2016
- Probabilistic Inference by Program Transformation in Hakaru (System Description)2016
- Simplifying Probabilistic Programs Using Computer Algebra2016
- StanJournal of Educational and Behavioral Statistics, 2015
- Purely functional lazy nondeterministic programmingJournal of Functional Programming, 2011
- Noncomputable Conditional Distributions2011
- Preserving Sharing in the Partial Evaluation of Lazy Functional ProgramsLecture Notes in Computer Science, 2008
- A note on Metropolis-Hastings kernels for general state spacesThe Annals of Applied Probability, 1998
- Conditioning as disintegrationStatistica Neerlandica, 1997
- A natural semantics for lazy evaluation1993