Particle Gibbs with Ancestor Sampling for Probabilistic Programs
<p>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 vari...
| Main Authors: | , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
2015
|
| _version_ | 1797059127916101632 |
|---|---|
| author | van de Meent, J Yang, H Mansinghka, V Wood, F |
| author_facet | van de Meent, J Yang, H Mansinghka, V Wood, F |
| author_sort | van de Meent, J |
| collection | OXFORD |
| description | <p>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.</p> |
| first_indexed | 2024-03-06T19:59:48Z |
| format | Conference item |
| id | oxford-uuid:26e26af2-5809-4623-8b7a-26516def3566 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T19:59:48Z |
| publishDate | 2015 |
| record_format | dspace |
| spelling | oxford-uuid:26e26af2-5809-4623-8b7a-26516def35662022-03-26T12:03:43ZParticle Gibbs with Ancestor Sampling for Probabilistic ProgramsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:26e26af2-5809-4623-8b7a-26516def3566EnglishOxford University Research Archive - Valet2015van de Meent, JYang, HMansinghka, VWood, F<p>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.</p> |
| spellingShingle | van de Meent, J Yang, H Mansinghka, V Wood, F Particle Gibbs with Ancestor Sampling for Probabilistic Programs |
| title | Particle Gibbs with Ancestor Sampling for Probabilistic Programs |
| title_full | Particle Gibbs with Ancestor Sampling for Probabilistic Programs |
| title_fullStr | Particle Gibbs with Ancestor Sampling for Probabilistic Programs |
| title_full_unstemmed | Particle Gibbs with Ancestor Sampling for Probabilistic Programs |
| title_short | Particle Gibbs with Ancestor Sampling for Probabilistic Programs |
| title_sort | particle gibbs with ancestor sampling for probabilistic programs |
| work_keys_str_mv | AT vandemeentj particlegibbswithancestorsamplingforprobabilisticprograms AT yangh particlegibbswithancestorsamplingforprobabilisticprograms AT mansinghkav particlegibbswithancestorsamplingforprobabilisticprograms AT woodf particlegibbswithancestorsamplingforprobabilisticprograms |