Probabilistic surrogate networks for simulators with unbounded randomness

Andreas Munk, Berend Zwartsenberg, Adam Ścibior, Atılım Güneş G. Baydin, Andrew Stewart, Goran Fernlund, Anoush Poursartip, Frank Wood
Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, PMLR 180:1423-1433, 2022.

Abstract

We present a framework for automatically structuring and training fast, approximate, deep neural surrogates of stochastic simulators. Unlike traditional approaches to surrogate modeling, our surrogates retain the interpretable structure and control flow of the reference simulator. Our surrogates target stochastic simulators where the number of random variables itself can be stochastic and potentially unbounded. Our framework further enables an automatic replacement of the reference simulator with the surrogate when undertaking amortized inference. The fidelity and speed of our surrogates allow for both faster stochastic simulation and accurate and substantially faster posterior inference. Using an illustrative yet non-trivial example we show our surrogates’ ability to accurately model a probabilistic program with an unbounded number of random variables. We then proceed with an example that shows our surrogates are able to accurately model a complex structure like an unbounded stack in a program synthesis example. We further demonstrate how our surrogate modeling technique makes amortized inference in complex black-box simulators an order of magnitude faster. Specifically, we do simulator-based materials quality testing, inferring safety-critical latent internal temperature profiles of composite materials undergoing curing.

Cite this Paper


BibTeX
@InProceedings{pmlr-v180-munk22a, title = {Probabilistic surrogate networks for simulators with unbounded randomness}, author = {Munk, Andreas and Zwartsenberg, Berend and \'Scibior, Adam and Baydin, At{\i}l{\i}m G{\"u}ne{\c s} G. and Stewart, Andrew and Fernlund, Goran and Poursartip, Anoush and Wood, Frank}, booktitle = {Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence}, pages = {1423--1433}, year = {2022}, editor = {Cussens, James and Zhang, Kun}, volume = {180}, series = {Proceedings of Machine Learning Research}, month = {01--05 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v180/munk22a/munk22a.pdf}, url = {https://proceedings.mlr.press/v180/munk22a.html}, abstract = {We present a framework for automatically structuring and training fast, approximate, deep neural surrogates of stochastic simulators. Unlike traditional approaches to surrogate modeling, our surrogates retain the interpretable structure and control flow of the reference simulator. Our surrogates target stochastic simulators where the number of random variables itself can be stochastic and potentially unbounded. Our framework further enables an automatic replacement of the reference simulator with the surrogate when undertaking amortized inference. The fidelity and speed of our surrogates allow for both faster stochastic simulation and accurate and substantially faster posterior inference. Using an illustrative yet non-trivial example we show our surrogates’ ability to accurately model a probabilistic program with an unbounded number of random variables. We then proceed with an example that shows our surrogates are able to accurately model a complex structure like an unbounded stack in a program synthesis example. We further demonstrate how our surrogate modeling technique makes amortized inference in complex black-box simulators an order of magnitude faster. Specifically, we do simulator-based materials quality testing, inferring safety-critical latent internal temperature profiles of composite materials undergoing curing.} }
Endnote
%0 Conference Paper %T Probabilistic surrogate networks for simulators with unbounded randomness %A Andreas Munk %A Berend Zwartsenberg %A Adam Ścibior %A Atılım Güneş G. Baydin %A Andrew Stewart %A Goran Fernlund %A Anoush Poursartip %A Frank Wood %B Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2022 %E James Cussens %E Kun Zhang %F pmlr-v180-munk22a %I PMLR %P 1423--1433 %U https://proceedings.mlr.press/v180/munk22a.html %V 180 %X We present a framework for automatically structuring and training fast, approximate, deep neural surrogates of stochastic simulators. Unlike traditional approaches to surrogate modeling, our surrogates retain the interpretable structure and control flow of the reference simulator. Our surrogates target stochastic simulators where the number of random variables itself can be stochastic and potentially unbounded. Our framework further enables an automatic replacement of the reference simulator with the surrogate when undertaking amortized inference. The fidelity and speed of our surrogates allow for both faster stochastic simulation and accurate and substantially faster posterior inference. Using an illustrative yet non-trivial example we show our surrogates’ ability to accurately model a probabilistic program with an unbounded number of random variables. We then proceed with an example that shows our surrogates are able to accurately model a complex structure like an unbounded stack in a program synthesis example. We further demonstrate how our surrogate modeling technique makes amortized inference in complex black-box simulators an order of magnitude faster. Specifically, we do simulator-based materials quality testing, inferring safety-critical latent internal temperature profiles of composite materials undergoing curing.
APA
Munk, A., Zwartsenberg, B., Ścibior, A., Baydin, A.G.G., Stewart, A., Fernlund, G., Poursartip, A. & Wood, F.. (2022). Probabilistic surrogate networks for simulators with unbounded randomness. Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 180:1423-1433 Available from https://proceedings.mlr.press/v180/munk22a.html.

Related Material