Active learning of deep surrogates for PDEs: application to metasurface design
Surrogate models for partial differential equations are widely used in the design of metamaterials to rapidly evaluate the behavior of composable components. However, the training cost of accurate surrogates by machine learning can rapidly increase with the number of variables. For photonic-device m...
| Main Authors: | , , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Published: |
Springer Science and Business Media LLC
2020
|
| Online Access: | https://hdl.handle.net/1721.1/128369 |
| _version_ | 1826209000007401472 |
|---|---|
| author | Pestourie, Raphael Mroueh, Youssef Nguyen, Thanh V. Das, Payel Johnson, Steven G |
| author2 | Massachusetts Institute of Technology. Department of Mathematics |
| author_facet | Massachusetts Institute of Technology. Department of Mathematics Pestourie, Raphael Mroueh, Youssef Nguyen, Thanh V. Das, Payel Johnson, Steven G |
| author_sort | Pestourie, Raphael |
| collection | MIT |
| description | Surrogate models for partial differential equations are widely used in the design of metamaterials to rapidly evaluate the behavior of composable components. However, the training cost of accurate surrogates by machine learning can rapidly increase with the number of variables. For photonic-device models, we find that this training becomes especially challenging as design regions grow larger than the optical wavelength. We present an active-learning algorithm that reduces the number of simulations required by more than an order of magnitude for an NN surrogate model of optical-surface components compared to uniform random samples. Results show that the surrogate evaluation is over two orders of magnitude faster than a direct solve, and we demonstrate how this can be exploited to accelerate large-scale engineering optimization. |
| first_indexed | 2024-09-23T14:15:48Z |
| format | Article |
| id | mit-1721.1/128369 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T14:15:48Z |
| publishDate | 2020 |
| publisher | Springer Science and Business Media LLC |
| record_format | dspace |
| spelling | mit-1721.1/1283692022-09-28T19:35:54Z Active learning of deep surrogates for PDEs: application to metasurface design Pestourie, Raphael Mroueh, Youssef Nguyen, Thanh V. Das, Payel Johnson, Steven G Massachusetts Institute of Technology. Department of Mathematics Surrogate models for partial differential equations are widely used in the design of metamaterials to rapidly evaluate the behavior of composable components. However, the training cost of accurate surrogates by machine learning can rapidly increase with the number of variables. For photonic-device models, we find that this training becomes especially challenging as design regions grow larger than the optical wavelength. We present an active-learning algorithm that reduces the number of simulations required by more than an order of magnitude for an NN surrogate model of optical-surface components compared to uniform random samples. Results show that the surrogate evaluation is over two orders of magnitude faster than a direct solve, and we demonstrate how this can be exploited to accelerate large-scale engineering optimization. 2020-11-05T21:02:08Z 2020-11-05T21:02:08Z 2020-10 2020-07 Article http://purl.org/eprint/type/JournalArticle 2057-3960 https://hdl.handle.net/1721.1/128369 Pestourie, Raphael et al. "Active learning of deep surrogates for PDEs: application to metasurface design." npj Computational Materials 6, 1 (October 2020): 164 © 2020 The Author(s) http://dx.doi.org/10.1038/s41524-020-00431-2 npj Computational Materials Creative Commons Attribution 4.0 International license https://creativecommons.org/licenses/by/4.0/ application/pdf Springer Science and Business Media LLC Raphael Pestourie |
| spellingShingle | Pestourie, Raphael Mroueh, Youssef Nguyen, Thanh V. Das, Payel Johnson, Steven G Active learning of deep surrogates for PDEs: application to metasurface design |
| title | Active learning of deep surrogates for PDEs: application to metasurface design |
| title_full | Active learning of deep surrogates for PDEs: application to metasurface design |
| title_fullStr | Active learning of deep surrogates for PDEs: application to metasurface design |
| title_full_unstemmed | Active learning of deep surrogates for PDEs: application to metasurface design |
| title_short | Active learning of deep surrogates for PDEs: application to metasurface design |
| title_sort | active learning of deep surrogates for pdes application to metasurface design |
| url | https://hdl.handle.net/1721.1/128369 |
| work_keys_str_mv | AT pestourieraphael activelearningofdeepsurrogatesforpdesapplicationtometasurfacedesign AT mrouehyoussef activelearningofdeepsurrogatesforpdesapplicationtometasurfacedesign AT nguyenthanhv activelearningofdeepsurrogatesforpdesapplicationtometasurfacedesign AT daspayel activelearningofdeepsurrogatesforpdesapplicationtometasurfacedesign AT johnsonsteveng activelearningofdeepsurrogatesforpdesapplicationtometasurfacedesign |