Deep recurrent neural networks for supernovae classification
We apply deep recurrent neural networks, which are capable of learning complex sequential information, to classify supernovae (code available at https://github.com/adammoss/supernovae). The observational time and filter fluxes are used as inputs to the network, but since the inputs are agnostic, add...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Published: |
American Astronomical Society
2017
|
| Subjects: | |
| Online Access: | https://eprints.nottingham.ac.uk/42324/ |
| _version_ | 1848796462707638272 |
|---|---|
| author | Charnock, Tom Moss, Adam |
| author_facet | Charnock, Tom Moss, Adam |
| author_sort | Charnock, Tom |
| building | Nottingham Research Data Repository |
| collection | Online Access |
| description | We apply deep recurrent neural networks, which are capable of learning complex sequential information, to classify supernovae (code available at https://github.com/adammoss/supernovae). The observational time and filter fluxes are used as inputs to the network, but since the inputs are agnostic, additional data such as host galaxy information can also be included. Using the Supernovae Photometric Classification Challenge (SPCC) data, we find that deep networks are capable of learning about light curves, however the performance of the network is highly sensitive to the amount of training data. For a training size of 50% of the representational SPCC data set (around 104 supernovae) we obtain a type-Ia versus non-type-Ia classification accuracy of 94.7%, an area under the Receiver Operating Characteristic curve AUC of 0.986 and an SPCC figure-of-merit F 1 = 0.64. When using only the data for the early-epoch challenge defined by the SPCC, we achieve a classification accuracy of 93.1%, AUC of 0.977, and F 1 = 0.58, results almost as good as with the whole light curve. By employing bidirectional neural networks, we can acquire impressive classification results between supernovae types I, II and III at an accuracy of 90.4% and AUC of 0.974. We also apply a pre-trained model to obtain classification probabilities as a function of time and show that it can give early indications of supernovae type. Our method is competitive with existing algorithms and has applications for future large-scale photometric surveys. |
| first_indexed | 2025-11-14T19:48:22Z |
| format | Article |
| id | nottingham-42324 |
| institution | University of Nottingham Malaysia Campus |
| institution_category | Local University |
| last_indexed | 2025-11-14T19:48:22Z |
| publishDate | 2017 |
| publisher | American Astronomical Society |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | nottingham-423242020-05-04T18:37:22Z https://eprints.nottingham.ac.uk/42324/ Deep recurrent neural networks for supernovae classification Charnock, Tom Moss, Adam We apply deep recurrent neural networks, which are capable of learning complex sequential information, to classify supernovae (code available at https://github.com/adammoss/supernovae). The observational time and filter fluxes are used as inputs to the network, but since the inputs are agnostic, additional data such as host galaxy information can also be included. Using the Supernovae Photometric Classification Challenge (SPCC) data, we find that deep networks are capable of learning about light curves, however the performance of the network is highly sensitive to the amount of training data. For a training size of 50% of the representational SPCC data set (around 104 supernovae) we obtain a type-Ia versus non-type-Ia classification accuracy of 94.7%, an area under the Receiver Operating Characteristic curve AUC of 0.986 and an SPCC figure-of-merit F 1 = 0.64. When using only the data for the early-epoch challenge defined by the SPCC, we achieve a classification accuracy of 93.1%, AUC of 0.977, and F 1 = 0.58, results almost as good as with the whole light curve. By employing bidirectional neural networks, we can acquire impressive classification results between supernovae types I, II and III at an accuracy of 90.4% and AUC of 0.974. We also apply a pre-trained model to obtain classification probabilities as a function of time and show that it can give early indications of supernovae type. Our method is competitive with existing algorithms and has applications for future large-scale photometric surveys. American Astronomical Society 2017-03-10 Article PeerReviewed Charnock, Tom and Moss, Adam (2017) Deep recurrent neural networks for supernovae classification. Astrophysical Journal, 837 (2). pp. 1-6. ISSN 1538-4357 methods: data analysis – supernovae: general – techniques: miscellaneous http://iopscience.iop.org/article/10.3847/2041-8213/aa603d/meta doi:10.3847/2041-8213/aa603d doi:10.3847/2041-8213/aa603d |
| spellingShingle | methods: data analysis – supernovae: general – techniques: miscellaneous Charnock, Tom Moss, Adam Deep recurrent neural networks for supernovae classification |
| title | Deep recurrent neural networks for supernovae classification |
| title_full | Deep recurrent neural networks for supernovae classification |
| title_fullStr | Deep recurrent neural networks for supernovae classification |
| title_full_unstemmed | Deep recurrent neural networks for supernovae classification |
| title_short | Deep recurrent neural networks for supernovae classification |
| title_sort | deep recurrent neural networks for supernovae classification |
| topic | methods: data analysis – supernovae: general – techniques: miscellaneous |
| url | https://eprints.nottingham.ac.uk/42324/ https://eprints.nottingham.ac.uk/42324/ https://eprints.nottingham.ac.uk/42324/ |