Utilize este identificador para referenciar este registo:
https://hdl.handle.net/1822/87937
Registo completo
Campo DC | Valor | Idioma |
---|---|---|
dc.contributor.author | Coelho, C. | por |
dc.contributor.author | Costa, M. Fernanda P. | por |
dc.contributor.author | Ferrás, Luís Jorge Lima | por |
dc.date.accessioned | 2024-01-08T09:38:07Z | - |
dc.date.available | 2024-01-08T09:38:07Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | https://hdl.handle.net/1822/87937 | - |
dc.description.abstract | Neural Ordinary Differential Equations (ODEs) have been used extensively to model physical systems because they represent a continuous-time function that can make predictions over the entire time domain. However, most of the time, the parameters of these physical systems are subject to strict laws/constraints. But there is no guarantee that the Neural ODE model satisfies these constraints. Therefore, we propose a two-stage training for Neural ODE. The first stage aims at find ing feasible parameters by minimizing a loss function defined by the constraints violation. The second stage aims at improving the feasible solution by minimiz ing the distance between the predicted and ground-truth values. By training the Neural ODE in two stages, we ensure that the governing laws of the system are satisfied and the model fits the data. | por |
dc.language.iso | eng | por |
dc.relation | info:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDP%2F00013%2F2020/PT | por |
dc.relation | info:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDB%2F00013%2F2020/PT | por |
dc.rights | openAccess | por |
dc.title | Prior knowledge meets neural odes: a two stage training method for improved explainability | por |
dc.type | conferencePaper | por |
dc.peerreviewed | yes | por |
sdum.event.title | ICLR 2023 | por |
sdum.event.type | conference | por |
oaire.citationStartPage | 1 | por |
oaire.citationEndPage | 6 | por |
dc.subject.fos | Ciências Naturais::Matemáticas | por |
sdum.conferencePublication | 11th International Conference on Learning Representations - ICLR 2023 | por |
Aparece nas coleções: |
Ficheiros deste registo:
Ficheiro | Descrição | Tamanho | Formato | |
---|---|---|---|---|
paper_revised.pdf | 498,03 kB | Adobe PDF | Ver/Abrir |