Using Deep Learning to Correct Theoretically-derived Models. Watson, P. A. G. In Deep learning for the Earth Sciences, pages 315–327. John Wiley & Sons, Ltd, 2021. Section: 22 _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781119646181.ch22![Using Deep Learning to Correct Theoretically-derived Models [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Machine learning holds great promise for identifying better equations of motion to use in Earth system simulators. However, learning a full set of equations for any component, such as the atmosphere or ocean, is likely to be a very difficult task. Faster progress may be made by keeping the existing components that have been derived using theory and experience with Earth simulations and using machine learning to compose an additive term that corrects their errors. This approach also has several other advantages. Here, studies that have used this approach are reviewed, and additional results are presented to display how well it performs at improving simulation of the Lorenz '96 dynamical equations. The outlook for applying this approach to Earth system modeling and simulating climate change is discussed.
@incollection{watson_using_2021,
title = {Using {Deep} {Learning} to {Correct} {Theoretically}-derived {Models}},
isbn = {978-1-119-64618-1},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1002/9781119646181.ch22},
abstract = {Machine learning holds great promise for identifying better equations of motion to use in Earth system simulators. However, learning a full set of equations for any component, such as the atmosphere or ocean, is likely to be a very difficult task. Faster progress may be made by keeping the existing components that have been derived using theory and experience with Earth simulations and using machine learning to compose an additive term that corrects their errors. This approach also has several other advantages. Here, studies that have used this approach are reviewed, and additional results are presented to display how well it performs at improving simulation of the Lorenz '96 dynamical equations. The outlook for applying this approach to Earth system modeling and simulating climate change is discussed.},
language = {en},
urldate = {2021-08-28},
booktitle = {Deep learning for the {Earth} {Sciences}},
publisher = {John Wiley \& Sons, Ltd},
author = {Watson, Peter A. G.},
year = {2021},
doi = {10.1002/9781119646181.ch22},
note = {Section: 22
\_eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781119646181.ch22},
keywords = {artificial neural networks, deep learning algorithms, machine learning algorithms, root mean square error, theoretically-derived models},
pages = {315--327},
}
Downloads: 0
{"_id":"ugKvWpBuGsHt3DH4y","bibbaseid":"watson-usingdeeplearningtocorrecttheoreticallyderivedmodels-2021","author_short":["Watson, P. A. G."],"bibdata":{"bibtype":"incollection","type":"incollection","title":"Using Deep Learning to Correct Theoretically-derived Models","isbn":"978-1-119-64618-1","url":"https://onlinelibrary.wiley.com/doi/abs/10.1002/9781119646181.ch22","abstract":"Machine learning holds great promise for identifying better equations of motion to use in Earth system simulators. However, learning a full set of equations for any component, such as the atmosphere or ocean, is likely to be a very difficult task. Faster progress may be made by keeping the existing components that have been derived using theory and experience with Earth simulations and using machine learning to compose an additive term that corrects their errors. This approach also has several other advantages. Here, studies that have used this approach are reviewed, and additional results are presented to display how well it performs at improving simulation of the Lorenz '96 dynamical equations. The outlook for applying this approach to Earth system modeling and simulating climate change is discussed.","language":"en","urldate":"2021-08-28","booktitle":"Deep learning for the Earth Sciences","publisher":"John Wiley & Sons, Ltd","author":[{"propositions":[],"lastnames":["Watson"],"firstnames":["Peter","A.","G."],"suffixes":[]}],"year":"2021","doi":"10.1002/9781119646181.ch22","note":"Section: 22 _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781119646181.ch22","keywords":"artificial neural networks, deep learning algorithms, machine learning algorithms, root mean square error, theoretically-derived models","pages":"315–327","bibtex":"@incollection{watson_using_2021,\n\ttitle = {Using {Deep} {Learning} to {Correct} {Theoretically}-derived {Models}},\n\tisbn = {978-1-119-64618-1},\n\turl = {https://onlinelibrary.wiley.com/doi/abs/10.1002/9781119646181.ch22},\n\tabstract = {Machine learning holds great promise for identifying better equations of motion to use in Earth system simulators. However, learning a full set of equations for any component, such as the atmosphere or ocean, is likely to be a very difficult task. Faster progress may be made by keeping the existing components that have been derived using theory and experience with Earth simulations and using machine learning to compose an additive term that corrects their errors. This approach also has several other advantages. Here, studies that have used this approach are reviewed, and additional results are presented to display how well it performs at improving simulation of the Lorenz '96 dynamical equations. The outlook for applying this approach to Earth system modeling and simulating climate change is discussed.},\n\tlanguage = {en},\n\turldate = {2021-08-28},\n\tbooktitle = {Deep learning for the {Earth} {Sciences}},\n\tpublisher = {John Wiley \\& Sons, Ltd},\n\tauthor = {Watson, Peter A. G.},\n\tyear = {2021},\n\tdoi = {10.1002/9781119646181.ch22},\n\tnote = {Section: 22\n\\_eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781119646181.ch22},\n\tkeywords = {artificial neural networks, deep learning algorithms, machine learning algorithms, root mean square error, theoretically-derived models},\n\tpages = {315--327},\n}\n\n","author_short":["Watson, P. A. G."],"key":"watson_using_2021","id":"watson_using_2021","bibbaseid":"watson-usingdeeplearningtocorrecttheoreticallyderivedmodels-2021","role":"author","urls":{"Paper":"https://onlinelibrary.wiley.com/doi/abs/10.1002/9781119646181.ch22"},"keyword":["artificial neural networks","deep learning algorithms","machine learning algorithms","root mean square error","theoretically-derived models"],"metadata":{"authorlinks":{}}},"bibtype":"incollection","biburl":"https://bibbase.org/zotero/manabsaharia","dataSources":["XpugPdSrCaPJgR6v7"],"keywords":["artificial neural networks","deep learning algorithms","machine learning algorithms","root mean square error","theoretically-derived models"],"search_terms":["using","deep","learning","correct","theoretically","derived","models","watson"],"title":"Using Deep Learning to Correct Theoretically-derived Models","year":2021}