Coping With Simulators That Don’t Always Return

Andrew Warrington, Frank Wood, Saeid Naderiparizi
; Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:1748-1758, 2020.

Abstract

Deterministic models are approximations of reality that are easy to interpret and often easier to build than stochastic alternatives. Unfortunately, as nature is capricious, observational data can never be fully explained by deterministic models in practice. Observation and process noise need to be added to adapt deterministic models to behave stochastically, such that they are capable of explaining and extrapolating from noisy data. We investigate and address computational inefficiencies that arise from adding process noise to deterministic simulators that fail to return for certain inputs; a property we describe as ’brittle’. We show how to train a conditional normalizing flow to propose perturbations such that the simulator succeeds with high probability, increasing computational efficiency.

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-warrington20a, title = {Coping With Simulators That Don’t Always Return}, author = {Warrington, Andrew and Wood, Frank and Naderiparizi, Saeid}, pages = {1748--1758}, year = {2020}, editor = {Silvia Chiappa and Roberto Calandra}, volume = {108}, series = {Proceedings of Machine Learning Research}, address = {Online}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/warrington20a/warrington20a.pdf}, url = {http://proceedings.mlr.press/v108/warrington20a.html}, abstract = {Deterministic models are approximations of reality that are easy to interpret and often easier to build than stochastic alternatives. Unfortunately, as nature is capricious, observational data can never be fully explained by deterministic models in practice. Observation and process noise need to be added to adapt deterministic models to behave stochastically, such that they are capable of explaining and extrapolating from noisy data. We investigate and address computational inefficiencies that arise from adding process noise to deterministic simulators that fail to return for certain inputs; a property we describe as ’brittle’. We show how to train a conditional normalizing flow to propose perturbations such that the simulator succeeds with high probability, increasing computational efficiency.} }
Endnote
%0 Conference Paper %T Coping With Simulators That Don’t Always Return %A Andrew Warrington %A Frank Wood %A Saeid Naderiparizi %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-warrington20a %I PMLR %J Proceedings of Machine Learning Research %P 1748--1758 %U http://proceedings.mlr.press %V 108 %W PMLR %X Deterministic models are approximations of reality that are easy to interpret and often easier to build than stochastic alternatives. Unfortunately, as nature is capricious, observational data can never be fully explained by deterministic models in practice. Observation and process noise need to be added to adapt deterministic models to behave stochastically, such that they are capable of explaining and extrapolating from noisy data. We investigate and address computational inefficiencies that arise from adding process noise to deterministic simulators that fail to return for certain inputs; a property we describe as ’brittle’. We show how to train a conditional normalizing flow to propose perturbations such that the simulator succeeds with high probability, increasing computational efficiency.
APA
Warrington, A., Wood, F. & Naderiparizi, S.. (2020). Coping With Simulators That Don’t Always Return. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in PMLR 108:1748-1758

Related Material