Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa
Abstract This study aimed to identify the optimal configuration for neural network (NN) emulators in numerical weather prediction, minimizing trial and error by comparing emulator performance across multiple hidden layers (1–5 layers), as automatically defined by the Sherpa library. Our findings rev...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2024-04-01
|
| Series: | Geoscience Letters |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40562-024-00336-8 |
| _version_ | 1827284275403161600 |
|---|---|
| author | Soonyoung Roh Park Sa Kim Hwan-Jin Song |
| author_facet | Soonyoung Roh Park Sa Kim Hwan-Jin Song |
| author_sort | Soonyoung Roh |
| collection | DOAJ |
| description | Abstract This study aimed to identify the optimal configuration for neural network (NN) emulators in numerical weather prediction, minimizing trial and error by comparing emulator performance across multiple hidden layers (1–5 layers), as automatically defined by the Sherpa library. Our findings revealed that Sherpa-applied emulators consistently demonstrated good results and stable performance with low errors in numerical simulations. The optimal configurations were observed with one and two hidden layers, improving results when two hidden layers were employed. The Sherpa-defined average neurons per hidden layer ranged between 153 and 440, resulting in a speedup relative to the CNT of 7–12 times. These results provide valuable insights for developing radiative physical NN emulators. Utilizing automatically determined hyperparameters can effectively reduce trial-and-error processes while maintaining stable outcomes. However, further experimentation is needed to establish the most suitable hyperparameter values that balance both speed and accuracy, as this study did not identify optimized values for all hyperparameters. |
| first_indexed | 2024-04-24T09:51:59Z |
| format | Article |
| id | doaj.art-33dafccb4b3a4b37a69591e4a2286d77 |
| institution | Directory Open Access Journal |
| issn | 2196-4092 |
| language | English |
| last_indexed | 2024-04-24T09:51:59Z |
| publishDate | 2024-04-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | Geoscience Letters |
| spelling | doaj.art-33dafccb4b3a4b37a69591e4a2286d772024-04-14T11:18:24ZengSpringerOpenGeoscience Letters2196-40922024-04-0111111410.1186/s40562-024-00336-8Streamlining hyperparameter optimization for radiation emulator training with automated SherpaSoonyoung Roh0Park Sa Kim1Hwan-Jin Song2Center for Atmospheric REmote Sensing (CARE), Kyungpook National UniversityNational Institute of Meteorological Sciences, Korea Meteorological AdministrationBK21 Weather Extremes Education and Research Team, Department of Atmospheric Sciences, Kyungpook National UniversityAbstract This study aimed to identify the optimal configuration for neural network (NN) emulators in numerical weather prediction, minimizing trial and error by comparing emulator performance across multiple hidden layers (1–5 layers), as automatically defined by the Sherpa library. Our findings revealed that Sherpa-applied emulators consistently demonstrated good results and stable performance with low errors in numerical simulations. The optimal configurations were observed with one and two hidden layers, improving results when two hidden layers were employed. The Sherpa-defined average neurons per hidden layer ranged between 153 and 440, resulting in a speedup relative to the CNT of 7–12 times. These results provide valuable insights for developing radiative physical NN emulators. Utilizing automatically determined hyperparameters can effectively reduce trial-and-error processes while maintaining stable outcomes. However, further experimentation is needed to establish the most suitable hyperparameter values that balance both speed and accuracy, as this study did not identify optimized values for all hyperparameters.https://doi.org/10.1186/s40562-024-00336-8Sherpa libraryHyperparameter optimizationNeural-network emulatorsNumerical weather predictionRRTMG-K radiation |
| spellingShingle | Soonyoung Roh Park Sa Kim Hwan-Jin Song Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa Geoscience Letters Sherpa library Hyperparameter optimization Neural-network emulators Numerical weather prediction RRTMG-K radiation |
| title | Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa |
| title_full | Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa |
| title_fullStr | Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa |
| title_full_unstemmed | Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa |
| title_short | Streamlining hyperparameter optimization for radiation emulator training with automated Sherpa |
| title_sort | streamlining hyperparameter optimization for radiation emulator training with automated sherpa |
| topic | Sherpa library Hyperparameter optimization Neural-network emulators Numerical weather prediction RRTMG-K radiation |
| url | https://doi.org/10.1186/s40562-024-00336-8 |
| work_keys_str_mv | AT soonyoungroh streamlininghyperparameteroptimizationforradiationemulatortrainingwithautomatedsherpa AT parksakim streamlininghyperparameteroptimizationforradiationemulatortrainingwithautomatedsherpa AT hwanjinsong streamlininghyperparameteroptimizationforradiationemulatortrainingwithautomatedsherpa |