{"id":"arxiv:2010.13749","kind":"preprint","title":"Meaningful uncertainties from deep neural network surrogates of large-scale numerical simulations","authors":["Gemma J. Anderson","Jim A. Gaffney","Brian K. Spears","Peer-Timo Bremer","Rushil Anirudh","Jayaraman J. Thiagarajan"],"author_count":6,"year":2020,"published":"2020-10-26","venue":null,"cited_by":null,"doi":null,"arxiv_id":"2010.13749","nct_id":null,"source_url":"https://arxiv.org/abs/2010.13749","archive_url":"https://space-kitty.com/papers/arxiv:2010.13749","updated":"2020-10-26","pmid":null,"pmcid":null,"openalex_id":null,"pdf_url":"https://arxiv.org/pdf/2010.13749","abstract":"Large-scale numerical simulations are used across many scientific disciplines to facilitate experimental development and provide insights into underlying physical processes, but they come with a significant computational cost. Deep neural networks (DNNs) can serve as highly-accurate surrogate models, with the capacity to handle diverse datatypes, offering tremendous speed-ups for prediction and many other downstream tasks. An important use-case for these surrogates is the comparison between simulations and experiments; prediction uncertainty estimates are crucial for making such comparisons meaningful, yet standard DNNs do not provide them. In this work we define the fundamental requirements for a DNN to be useful for scientific applications, and demonstrate a general variational inference approach to equip predictions of scalar and image data from a DNN surrogate model trained on inertial confinement fusion simulations with calibrated Bayesian uncertainties. Critically, these uncertainties are interpretable, meaningful and preserve physics-correlations in the predicted quantities.","abstract_license":"CC0-1.0 (arXiv descriptive metadata)","license":null,"is_open_access":false,"retracted":false,"peer_reviewed":false,"categories":["stat.ML","cs.LG","physics.plasm-ph"],"details":null,"provenance":"arxiv","attribution":"Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.","topics":[],"cited_in_dossiers":[]}