| 1. Identity statement | ||||
| Reference Type | Conference Paper (Conference Proceedings) | |||
| Site | mtc-m21c.sid.inpe.br | |||
| Holder Code | isadg {BR SPINPE} ibi 8JMKD3MGPCW/3DT298S | |||
| Identifier | 8JMKD3MGP3W34R/43L8TKH | |||
| Repository | sid.inpe.br/mtc-m21c/2020/11.23.10.37 | |||
| Metadata Repository | sid.inpe.br/mtc-m21c/2020/11.23.10.37.22 | |||
| Metadata Last Update | 2022:01.04.01.35.38 (UTC) administrator | |||
| Secondary Key | INPE--PRE/ | |||
| Citation Key | RosaACPBSSC:2020:SuNoPh | |||
| Title | Deep neural networks for learning spatiotemporal pattern formation: a survey in nonlinear physics  | |||
| Year | 2020 | |||
| Access Date | 2024, May 16 | |||
| Secondary Type | PRE CN | |||
| 2. Context | ||||
| Author | 1 Rosa, Reinaldo Roberto 2 An, Wu Chun 3 Caproni, Anderson 4 Pontes, José 5 Barchi, Paulo Henrique 6 Stalder, Diego H. 7 Sautter, Rubens Andreas 8 Carvalho, Reinaldo Ramos de | |||
| Resume Identifier | 1 8JMKD3MGP5W/3C9JJ5D 2 3 4 5 6 7 8 8JMKD3MGP5W/3C9JJ5B | |||
| Group | 1 LABAC-COCTE-INPE-MCTIC-GOV-BR 2 CAP-COMP-SESPG-INPE-MCTIC-GOV-BR 3 4 5 CAP-COMP-SESPG-INPE-MCTIC-GOV-BR 6 7 CAP-COMP-SESPG-INPE-MCTIC-GOV-BR | |||
| Affiliation | 1 Instituto Nacional de Pesquisas Espaciais (INPE) 2 Instituto Nacional de Pesquisas Espaciais (INPE) 3 UNICSUL 4 Universidade do Estado do Rio de Janeiro (UERJ) 5 Instituto Nacional de Pesquisas Espaciais (INPE) 6 7 Instituto Nacional de Pesquisas Espaciais (INPE) | |||
| Author e-Mail Address | 1 reinaldo.rosa@inpe.br 2 3 4 5 paulo.barchi@inpe.br 6 7 rubens.sautter@inpe.br | |||
| Conference Name | Encontro de Outono Sociedade Brasileira de Física | |||
| Conference Location | Online | |||
| Date | 23 a 26 nov. | |||
| History (UTC) | 2020-11-23 10:38:20 :: simone -> administrator :: 2020 2022-01-04 01:35:38 :: administrator -> simone :: 2020 | |||
| 3. Content and structure | ||||
| Is the master or a copy? | is the master | |||
| Content Stage | completed | |||
| Transferable | 1 | |||
| Content Type | External Contribution | |||
| Abstract | Mining valuable knowledge from spatiotemporal data in nonlinear physics is critically important to many real world applications including reaction-diffusion, chaos and turbulence. As the complexity (volume, variety and resolution) of spatiotemporal data sets increases dramatically, traditional methods of data mining, especially methods based on supervised statistics, are becoming insufficient. With the recent advances in deep learning techniques (DLT), such as the recurrent neural network (RNN) and the convolutional neural network (CNN), considerable successes have been achieved in invariant machine learning tasks due to their powerful ability to learn hierarchical characteristics in spatial and temporal domains, and have been widely applied in various spatiotemporal data modeling tasks, such as pattern classification, predictive learning, representation learning and spatiotemporal anomaly detection. In this study, we provide a comprehensive survey on recent progress in applying deep learning techniques for spatiotemporal data mining (recognition, classification and prediction) from canonical nonlinear regimes in physics as reaction-diffusion from Ginzburg-Landau equation, spatiotemporal chaos from coupled map lattices and weak and fully developed turbulence from MHD. To measure the input features for the traditional machine learning methodology, we have developed a system called CyMorph, with a novel non-parametric approach to spatiotemporal pattern classification. We first categorize the types of spatiotemporal data combining accurate machine learning classifications from the CyMorph analysis with deep learning methodologies. Then a framework is introduced to show a general pipeline of the utilization of deep learning models. Next we investigated the power of generalization of DLT by operating small variations in the control parameters that are responsible for subtle changes in each group of simulated nonlinear processes including transitions from regular to irregular patterns and the appearance of remarkable structural aspects. Finally, we conclude the limitations of current research and point out future research directions. | |||
| Area | COMP | |||
| Arrangement 1 | urlib.net > BDMCI > Fonds > Produção anterior à 2021 > LABAC > Deep neural networks... | |||
| Arrangement 2 | urlib.net > BDMCI > Fonds > Produção pgr ATUAIS > CAP > Deep neural networks... | |||
| doc Directory Content | there are no files | |||
| source Directory Content | there are no files | |||
| agreement Directory Content |
| |||
| 4. Conditions of access and use | ||||
| User Group | simone | |||
| Reader Group | administrator simone | |||
| Visibility | shown | |||
| Update Permission | not transferred | |||
| 5. Allied materials | ||||
| Next Higher Units | 8JMKD3MGPCW/3ESGTTP 8JMKD3MGPCW/3F2PHGS | |||
| Host Collection | urlib.net/www/2017/11.22.19.04 | |||
| 6. Notes | ||||
| Empty Fields | archivingpolicy archivist booktitle callnumber copyholder copyright creatorhistory descriptionlevel dissemination doi e-mailaddress edition editor format isbn issn keywords label language lineage mark mirrorrepository nextedition notes numberoffiles numberofvolumes orcid organization pages parameterlist parentrepositories previousedition previouslowerunit progress project publisher publisheraddress readpermission rightsholder schedulinginformation secondarydate secondarymark serieseditor session shorttitle size sponsor subject targetfile tertiarymark tertiarytype type url versiontype volume | |||
| 7. Description control | ||||
| e-Mail (login) | simone | |||
| update | ||||