Stein Variational Adaptive Importance Sampling

Jun Han, Qiang Liu
Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, PMLR R15:711-720, 2017.

Abstract

We propose a novel adaptive importance sam- pling algorithm which incorporates Stein vari- ational gradient decent algorithm (SVGD) with importance sampling (IS). Our algorithm lever- ages the nonparametric transforms in SVGD to iteratively decrease the KL divergence between importance proposals and target distributions. The advantages of our algorithm are twofold: 1) it turns SVGD into a standard IS algorithm, al- lowing us to use standard diagnostic and ana- lytic tools of IS to evaluate and interpret the re- sults, and 2) it does not restrict the choice of the importance proposals to predefined distribu- tion families like traditional (adaptive) IS meth- ods. Empirical experiments demonstrate that our algorithm performs well on evaluating partition functions of restricted Boltzmann machines and testing likelihood of variational auto-encoders.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR15-han17a, title = {{S}tein Variational Adaptive Importance Sampling}, author = {Han, Jun and Liu, Qiang}, booktitle = {Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence}, pages = {711--720}, year = {2017}, editor = {Elidan, Gal and Kersting, Kristian}, volume = {R15}, series = {Proceedings of Machine Learning Research}, month = {11--15 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r15/main/assets/han17a/han17a.pdf}, url = {https://proceedings.mlr.press/r15/han17a.html}, abstract = {We propose a novel adaptive importance sam- pling algorithm which incorporates Stein vari- ational gradient decent algorithm (SVGD) with importance sampling (IS). Our algorithm lever- ages the nonparametric transforms in SVGD to iteratively decrease the KL divergence between importance proposals and target distributions. The advantages of our algorithm are twofold: 1) it turns SVGD into a standard IS algorithm, al- lowing us to use standard diagnostic and ana- lytic tools of IS to evaluate and interpret the re- sults, and 2) it does not restrict the choice of the importance proposals to predefined distribu- tion families like traditional (adaptive) IS meth- ods. Empirical experiments demonstrate that our algorithm performs well on evaluating partition functions of restricted Boltzmann machines and testing likelihood of variational auto-encoders.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Stein Variational Adaptive Importance Sampling %A Jun Han %A Qiang Liu %B Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2017 %E Gal Elidan %E Kristian Kersting %F pmlr-vR15-han17a %I PMLR %P 711--720 %U https://proceedings.mlr.press/r15/han17a.html %V R15 %X We propose a novel adaptive importance sam- pling algorithm which incorporates Stein vari- ational gradient decent algorithm (SVGD) with importance sampling (IS). Our algorithm lever- ages the nonparametric transforms in SVGD to iteratively decrease the KL divergence between importance proposals and target distributions. The advantages of our algorithm are twofold: 1) it turns SVGD into a standard IS algorithm, al- lowing us to use standard diagnostic and ana- lytic tools of IS to evaluate and interpret the re- sults, and 2) it does not restrict the choice of the importance proposals to predefined distribu- tion families like traditional (adaptive) IS meth- ods. Empirical experiments demonstrate that our algorithm performs well on evaluating partition functions of restricted Boltzmann machines and testing likelihood of variational auto-encoders. %Z Reissued by PMLR on 04 October 2026.
APA
Han, J. & Liu, Q.. (2017). Stein Variational Adaptive Importance Sampling. Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R15:711-720 Available from https://proceedings.mlr.press/r15/han17a.html. Reissued by PMLR on 04 October 2026.

Related Material