Simulation-Based Stacking

Yuling Yao, Bruno Régaldo-Saint Blancard, Justin Domke
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, PMLR 238:4267-4275, 2024.

Abstract

Simulation-based inference has been popular for amortized Bayesian computation. It is typical to have more than one posterior approximation, from different inference algorithms, different architectures, or simply the randomness of initialization and stochastic gradients. With a consistency guarantee, we present a general posterior stacking framework to make use of all available approximations. Our stacking method is able to combine densities, simulation draws, confidence intervals, and moments, and address the overall precision, calibration, coverage, and bias of the posterior approximation at the same time. We illustrate our method on several benchmark simulations and a challenging cosmological inference task.

Cite this Paper

BibTeX
@InProceedings{pmlr-v238-yao24b, title = {Simulation-Based Stacking}, author = {Yao, Yuling and R\'{e}galdo-Saint Blancard, Bruno and Domke, Justin}, booktitle = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics}, pages = {4267--4275}, year = {2024}, editor = {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen}, volume = {238}, series = {Proceedings of Machine Learning Research}, month = {02--04 May}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v238/yao24b/yao24b.pdf}, url = {https://proceedings.mlr.press/v238/yao24b.html}, abstract = {Simulation-based inference has been popular for amortized Bayesian computation. It is typical to have more than one posterior approximation, from different inference algorithms, different architectures, or simply the randomness of initialization and stochastic gradients. With a consistency guarantee, we present a general posterior stacking framework to make use of all available approximations. Our stacking method is able to combine densities, simulation draws, confidence intervals, and moments, and address the overall precision, calibration, coverage, and bias of the posterior approximation at the same time. We illustrate our method on several benchmark simulations and a challenging cosmological inference task.} }
Endnote
%0 Conference Paper %T Simulation-Based Stacking %A Yuling Yao %A Bruno Régaldo-Saint Blancard %A Justin Domke %B Proceedings of The 27th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2024 %E Sanjoy Dasgupta %E Stephan Mandt %E Yingzhen Li %F pmlr-v238-yao24b %I PMLR %P 4267--4275 %U https://proceedings.mlr.press/v238/yao24b.html %V 238 %X Simulation-based inference has been popular for amortized Bayesian computation. It is typical to have more than one posterior approximation, from different inference algorithms, different architectures, or simply the randomness of initialization and stochastic gradients. With a consistency guarantee, we present a general posterior stacking framework to make use of all available approximations. Our stacking method is able to combine densities, simulation draws, confidence intervals, and moments, and address the overall precision, calibration, coverage, and bias of the posterior approximation at the same time. We illustrate our method on several benchmark simulations and a challenging cosmological inference task.
APA
Yao, Y., Régaldo-Saint Blancard, B. & Domke, J.. (2024). Simulation-Based Stacking. Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:4267-4275 Available from https://proceedings.mlr.press/v238/yao24b.html.

Related Material