Deep reinforcement learning of model error corrections ...
<!--!introduction!--> Deep reinforcement learning has empowered recent advances in games like chess or in language modelling with chatGPT, but can also control nuclear fusion in a Tokamak reactor. In the often-used actor-critic framework, a neural network is trained to control while another ne...
Main Authors: | , , , |
---|---|
Format: | Conference Object |
Language: | English |
Published: |
GFZ German Research Centre for Geosciences
2023
|
Subjects: | |
Online Access: | https://dx.doi.org/10.57757/iugg23-3336 https://gfzpublic.gfz-potsdam.de/pubman/item/item_5019665 |
id |
ftdatacite:10.57757/iugg23-3336 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.57757/iugg23-3336 2023-07-23T04:21:42+02:00 Deep reinforcement learning of model error corrections ... Finn, Tobias Durand, Charlotte Farchi, Alban Bocquet, Marc 2023 https://dx.doi.org/10.57757/iugg23-3336 https://gfzpublic.gfz-potsdam.de/pubman/item/item_5019665 en eng GFZ German Research Centre for Geosciences Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode cc-by-4.0 Article ConferencePaper Oral 2023 ftdatacite https://doi.org/10.57757/iugg23-3336 2023-07-03T21:27:30Z <!--!introduction!--> Deep reinforcement learning has empowered recent advances in games like chess or in language modelling with chatGPT, but can also control nuclear fusion in a Tokamak reactor. In the often-used actor-critic framework, a neural network is trained to control while another neural network evaluates the actions of the first network. In this talk, we cast model error correction into a remarkably similar framework to learn from temporally sparse observations. A first neural network corrects model errors, while a second, simultaneously trained, estimates the future costs if the model error correction were applied. This allows us to circumvent the need for the model adjoint or any linear approximation for learning in a gradient-based optimization framework. We test this novel framework on low-order Lorenz and sea-ice models. Trained on already existing trajectories, the actor-critic framework can not only correct persisting model errors, but significantly surpasses linear and ensemble ... : The 28th IUGG General Assembly (IUGG2023) (Berlin 2023) ... Conference Object Sea ice DataCite Metadata Store (German National Library of Science and Technology) |
institution |
Open Polar |
collection |
DataCite Metadata Store (German National Library of Science and Technology) |
op_collection_id |
ftdatacite |
language |
English |
description |
<!--!introduction!--> Deep reinforcement learning has empowered recent advances in games like chess or in language modelling with chatGPT, but can also control nuclear fusion in a Tokamak reactor. In the often-used actor-critic framework, a neural network is trained to control while another neural network evaluates the actions of the first network. In this talk, we cast model error correction into a remarkably similar framework to learn from temporally sparse observations. A first neural network corrects model errors, while a second, simultaneously trained, estimates the future costs if the model error correction were applied. This allows us to circumvent the need for the model adjoint or any linear approximation for learning in a gradient-based optimization framework. We test this novel framework on low-order Lorenz and sea-ice models. Trained on already existing trajectories, the actor-critic framework can not only correct persisting model errors, but significantly surpasses linear and ensemble ... : The 28th IUGG General Assembly (IUGG2023) (Berlin 2023) ... |
format |
Conference Object |
author |
Finn, Tobias Durand, Charlotte Farchi, Alban Bocquet, Marc |
spellingShingle |
Finn, Tobias Durand, Charlotte Farchi, Alban Bocquet, Marc Deep reinforcement learning of model error corrections ... |
author_facet |
Finn, Tobias Durand, Charlotte Farchi, Alban Bocquet, Marc |
author_sort |
Finn, Tobias |
title |
Deep reinforcement learning of model error corrections ... |
title_short |
Deep reinforcement learning of model error corrections ... |
title_full |
Deep reinforcement learning of model error corrections ... |
title_fullStr |
Deep reinforcement learning of model error corrections ... |
title_full_unstemmed |
Deep reinforcement learning of model error corrections ... |
title_sort |
deep reinforcement learning of model error corrections ... |
publisher |
GFZ German Research Centre for Geosciences |
publishDate |
2023 |
url |
https://dx.doi.org/10.57757/iugg23-3336 https://gfzpublic.gfz-potsdam.de/pubman/item/item_5019665 |
genre |
Sea ice |
genre_facet |
Sea ice |
op_rights |
Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode cc-by-4.0 |
op_doi |
https://doi.org/10.57757/iugg23-3336 |
_version_ |
1772187758631059456 |