Active learning for computationally efficient distribution of binary evolution simulations

Rocha, K. A. et al. (2022) Active learning for computationally efficient distribution of binary evolution simulations. Astrophysical Journal, 938(1), 64.

[img] Text
277256.pdf - Published Version
Available under License Creative Commons Attribution.

1MB

Abstract

Binary stars undergo a variety of interactions and evolutionary phases, critical for predicting and explaining observations. Binary population synthesis with full simulation of stellar structure and evolution is computationally expensive, requiring a large number of mass-transfer sequences. The recently developed binary population synthesis code POSYDON incorporates grids of MESA binary star simulations that are interpolated to model large-scale populations of massive binaries. The traditional method of computing a high-density rectilinear grid of simulations is not scalable for higher-dimension grids, accounting for a range of metallicities, rotation, and eccentricity. We present a new active learning algorithm, psy-cris, which uses machine learning in the data-gathering process to adaptively and iteratively target simulations to run, resulting in a custom, high-performance training set. We test psy-cris on a toy problem and find the resulting training sets require fewer simulations for accurate classification and regression than either regular or randomly sampled grids. We further apply psy-cris to the target problem of building a dynamic grid of MESA simulations, and we demonstrate that, even without fine tuning, a simulation set of only ∼1/4 the size of a rectilinear grid is sufficient to achieve the same classification accuracy. We anticipate further gains when algorithmic parameters are optimized for the targeted application. We find that optimizing for classification only may lead to performance losses in regression, and vice versa. Lowering the computational cost of producing grids will enable new population synthesis codes such as POSYDON to cover more input parameters while preserving interpolation accuracies.

Item Type:Articles
Additional Information:K.A.R. is supported by the Gordon and Betty Moore Foundation (PI Kalogera, grant award GBMF8477). K.A.R. also thanks the LSSTC Data Science Fellowship Program, which is funded by LSSTC, NSF Cybertraining Grant No. 1829740, the Brinson Foundation, and the Moore Foundation; their participation in the program has benefited this work. J.J.A. acknowledges support from CIERA and Northwestern University through a Postdoctoral Fellowship. P.M. acknowledges support from the FWO junior postdoctoral fellowship No. 12ZY520N. C.P.L.B. was supported by ADD, and Z.D. is supported by the CIERA Board of Visitors Research Professorship. V.K. was partially supported through a CIFAR Senior Fellowship and a Guggenheim Fellowship. S.B., T.F., K.K., D.M., Z.X. and E.Z. were supported by a Swiss National Science Foundation Professorship grant (PI Fragos, project number PP00P2 176868). K.K. was partially supported by the Federal Commission for Scholarships for Foreign Students for the Swiss Government Excellence Scholarship (ESKAS No. 2021.0277). Z.X. was supported by the Chinese Scholarship Council (CSC). The computations were performed at Northwestern University on the Trident computer cluster (funded by the GBMF8477 award). This research was supported in part through the computational resources and staff contributions provided for the Quest high performance computing facility at Northwestern University which is jointly supported by the Office of the Provost, the Office for Research, and Northwestern University Information Technology.
Status:Published
Refereed:Yes
Glasgow Author(s) Enlighten ID:Berry, Dr Christopher
Authors: Rocha, K. A., Andrews, J. J., Berry, C. P.L., Doctor, Z., Marchant, P., Kalogera, V., Coughlin, S., Bavera, S. S., Dotter, A., Fragos, T., Kovlakas, K., Misra, D., Xing, Z., and Zapartas, E.
College/School:College of Science and Engineering > School of Physics and Astronomy
Research Centre:College of Science and Engineering > School of Physics and Astronomy > Institute for Gravitational Research
Journal Name:Astrophysical Journal
Publisher:IOP Publishing
ISSN:0004-637X
ISSN (Online):1538-4357
Published Online:13 October 2022
Copyright Holders:Copyright © 2022. The Author(s)
First Published:First published in Astrophysical Journal 938(1):64
Publisher Policy:Reproduced under a Creative Commons license
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record