Conference item icon

Conference item

Particle Gibbs with ancestor sampling for probabilistic programs

Abstract:
Particle Markov chain Monte Carlo techniques rank among current state-of-the-art methods for probabilistic program inference. A drawback of these techniques is that they rely on importance resampling, which results in degenerate particle trajectories and a low effective sample size for variables sampled early in a program. We here develop a formalism to adapt ancestor resampling, a technique that mitigates particle degeneracy, to the probabilistic programming setting. We present empirical results that demonstrate nontrivial performance gains.
Publication status:
Published
Peer review status:
Reviewed (other)

Actions


Access Document


Files:

Authors


More by this author
Institution:
University of Oxford
Division:
MPLS
Department:
Computer Science
Role:
Author
More by this author
Institution:
University of Oxford
Role:
Author
More from this funder
Funding agency for:
Meent, J
More from this funder
Funding agency for:
Meent, J
More from this funder
Funding agency for:
Mansinghka, V
Expand funders...
Publisher:
Journal of Machine Learning Research Publisher's website
Journal:
Proceedings of the 18th International Conference on Artificial Intelligence and Statistics (AISTATS Journal website
Volume:
38
Pages:
986-994
Host title:
Proceedings of the 18th International Conference on Artificial Intelligence and Statistics (AISTATS)
Publication date:
2015-02-21
Acceptance date:
2015-01-11
EISSN:
1938-7228
ISSN:
2640-3498
Source identifiers:
581040
Pubs id:
pubs:581040
UUID:
uuid:37ee25a1-0fc0-47c8-a9cc-ed81743be51b
Local pid:
pubs:581040
Deposit date:
2016-01-03

Terms of use


Views and Downloads






If you are the owner of this record, you can report an update to it here: Report update to this record

TO TOP