Neuroevolutionary Transfer Learning for Time Series Forecasting

Aymeric Vellinger; J. F. Torres; Federico Divina; Wim Vanhoof

doi:10.1007/978-3-031-42529-5_21

Neuroevolutionary Transfer Learning for Time Series Forecasting

Aymeric Vellinger, J. F. Torres, Federico Divina, Wim Vanhoof

Résultats de recherche: Contribution à un événement scientifique (non publié) › Papier › Revue par des pairs

Résumé

In this paper, we propose a neuroevolution technique specifically designed for evolving LSTM networks. The proposed technique uses a grammar-based approach to evolve LSTM neural networks for time series prediction tasks, and is based on a previous technique which was designed in order to evolve CNN networks.

We use transfer learning in order to reduce the computational time of our approach. We have compared results obtained with other state of the art time series forecasting techniques on twenty time series, which contains data generated by sensors placed on a number of Iberian pigs. Results obtained confirm the effectiveness of the strategy proposed in this work.

Overall, we showcase the potential of our proposal in producing precise and efficient deep learning models for time series prediction, as well as the adaptability of transfer learning to new datasets.

langue originale	Anglais
Pages	219-228
Nombre de pages	10
Les DOIs	https://doi.org/10.1007/978-3-031-42529-5_21
Etat de la publication	Publié - 31 août 2023

Financement

Acknowledgements. Iztok Fister Jr. is grateful the Slovenian Research Agency for the financial support under Research Core Funding No. P2-0057. Iztok Fister thanks the Slovenian Research Agency for the financial support under Research Core Funding No. P2-0042 - Digital twin. This research is also supported by the PDE-GIR project from the European Union Horizon 2020 research and innovation programme under the Marie Skodowska-Curie grant agreement No 778035, and the project PID2021-127073OB-I00 of the MCIN/AEI/10.13039/501100011033/FEDER, EU. This research work has been supported by the Spanish national project PID2021-127073OB-I00 of the MCIN/AEI/10.13039/501100011033/FEDER, EU, and the European project PDE-GIR with grant number H2020-MSCA-RISE-2017-778035 of the European Union’s Horizon 2020 research and innovation programme, Marie Sklodowska-Curie Actions (MSCA) programme. Acknowledgement. This work was partially supported by the European Commission, under European Project 5G-Induce, grant number 101016941. This work is partially supported by Universidad de León, under the “Programa Propio de Investigación de la Universidad de León 2021” grant. Acknowledgements. This research is supported by the project Future Artificial Intelligence Research (FAIR) - PNRR MUR Cod. PE0000013 - CUP: E63C22001940006. Acknowledgements. The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the projects PID2020-117954RB-C22 and TED2021-131311B and Junta de Andalućıa for the project PYC20 RE 078 USE. Acknowledgments. The authors would like to acknowledge the financial support of the Ministerio de Ciencias, Tecnología e InnovaciÃşn (Minciencias) through Scholarship Program No. 860. This work has also benefited from a State grant managed by the National Research Agency under the “Investissements d’Avenir” program with the reference ANR-18-EURE-0021. Work also partially supported by the Spanish project TED2021-132470B-I00, funded by MCIN-AEI-10.13039-501100011033, and the GOMINOLA project (PID2020-118112RB-C21, funded by MCIN-AEI-10.13039-501100011033). CITIC, as a Research Center of the University System of Galicia, is funded by Consellería de Educación, Universidade e Formación Profesional of the Xunta de Galicia through the European Regional Development Fund (ERDF) and the Secretaría Xeral de Universidades (Ref. ED431G 2019/01). Acknowledgements. The authors would like to thank the Spanish Ministry of Science and Innovation and the Junta de Andalucía for their support within the projects PID2020-117954RB-C21 and TED2021-131311B-C22, PY20-00870, UPO-138516, respectively. The authors would also like to thank the Fundación Tatiana Pérez de Guzmán el Buenofor the support offered through the Beca Predoctoral en Medioambiente de 2018. Acknowledgements. Míriam Timiraos’s research was supported by the “Xunta de Galicia” (Regional Government of Galicia) through grants to industrial PhD (http:// gain.xunta.gal/), under the “Doutoramento Industrial 2022” grant with reference: 04 IN606D 2022 2692965. Acknowledgments. Álvaro Michelena’s research was supported by the Spanish Ministry of Universities (https://www.universidades.gob.es/), under the ”Formación de Profesorado Universitario” grant with reference FPU21/00932. Particular thanks go as well to the conference’s main sponsors, Startup Olé, the CYL-HUB project financed with NEXT-GENERATION funds from the European Union and channeled by Junta de Castilla y León through the Regional Ministry of Industry, Trade and Employment, BISITE research group at the University of Salamanca, CTC research group at the University of A Coruña, and the University of Salamanca. They jointly contributed in an active and constructive manner to the success of this initiative. Acknowledgements. Authors acknowledge funding under grant AI4TES, by the Spanish Min. of Economic Affairs and Digital Transformation and by EU Next Generation EU/PRTR. Carlos J. Gallego acknowledges funding for his scholarship from UPM RP180022025. Álvaro Michelena’s research was supported by the Spanish Ministry of Universities (https://www.universidades.gob.es/), under the “Formación de Profesorado Universi-tario” grant with reference: FPU21/00932. Acknowledgments. The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the project PID2020-117954RB-C21 and the European Regional Development Fund and Junta de Andalućıa for projects PY20-00870 and UPO-138516. Míriam Timiraos’s research was supported by the Xunta de Galicia (Regional Government of Galicia) through grants to industrial Ph.D. (http://gain.xunta.gal), under the Doutoramento Industrial 2022 grant with reference: 04 IN606D 2022 2692965.

Bailleurs de fonds	Numéro du bailleur de fonds
ANR-18-EURE-0021	TED2021-132470B-I00, MCIN-AEI-10.13039-501100011033, PID2020-118112RB-C21
Consellería de Educación, Universidade e Formación
FAIR	E63C22001940006, PE0000013
Formación de Profesorado Universi-tario
Marie Skodowska-Curie	778035, PID2021-127073OB-I00
Regional Ministry of Industry, Trade and Employment
Secretaría Xeral de Universidades	ED431G 2019/01
Spanish Ministry of Universities	FPU21/00932
Universidad de León 2021
Universidad de Salamanca
Xunta de Galicia (Regional Government of Galicia
Horizon 2020 Framework Programme
HORIZON EUROPE Marie Sklodowska-Curie Actions
European commission	H2020-MSCA-RISE-2017-778035, 101016941
Agence Nationale de la Recherche
Ministerio de Ciencia, Tecnología e Innovación	860
Javna Agencija za Raziskovalno Dejavnost RS	P2-0057, P2-0042
Universiti Putra Malaysia
Ministerio de Ciencia e Innovación	TED2021-131311B, PID2020-117954RB-C22
Horizon 2020
Consejería de Educación, Junta de Castilla y León
European Regional Development Fund
Xunta de Galicia	04 IN606D 2022 2692965
Junta de Andalucía	PY20-00870, PID2020-117954RB-C21, TED2021-131311B-C22, PYC20 RE 078 USE, UPO-138516
Universidad de León
project Future Artificial Intelligence Research

Accès au document

10.1007/978-3-031-42529-5_21

https://dblp.org/rec/conf/softcomp/VellingerTDV23

Autres fichiers et liens

Lien vers la publication sur Scopus

Contient cette citation

@conference{d95b0f2ed57f4c77a8f5280546a13470,

title = "Neuroevolutionary Transfer Learning for Time Series Forecasting",

abstract = "In this paper, we propose a neuroevolution technique specifically designed for evolving LSTM networks. The proposed technique uses a grammar-based approach to evolve LSTM neural networks for time series prediction tasks, and is based on a previous technique which was designed in order to evolve CNN networks.We use transfer learning in order to reduce the computational time of our approach. We have compared results obtained with other state of the art time series forecasting techniques on twenty time series, which contains data generated by sensors placed on a number of Iberian pigs. Results obtained confirm the effectiveness of the strategy proposed in this work.Overall, we showcase the potential of our proposal in producing precise and efficient deep learning models for time series prediction, as well as the adaptability of transfer learning to new datasets.",

keywords = "Deep Learning, Neuroevolution, Time series forecasting",

author = "Aymeric Vellinger and Torres, {J. F.} and Federico Divina and Wim Vanhoof",

note = "Funding Information: Acknowledgements. Iztok Fister Jr. is grateful the Slovenian Research Agency for the financial support under Research Core Funding No. P2-0057. Iztok Fister thanks the Slovenian Research Agency for the financial support under Research Core Funding No. P2-0042 - Digital twin. This research is also supported by the PDE-GIR project from the European Union Horizon 2020 research and innovation programme under the Marie Skodowska-Curie grant agreement No 778035, and the project PID2021-127073OB-I00 of the MCIN/AEI/10.13039/501100011033/FEDER, EU. Funding Information: This research work has been supported by the Spanish national project PID2021-127073OB-I00 of the MCIN/AEI/10.13039/501100011033/FEDER, EU, and the European project PDE-GIR with grant number H2020-MSCA-RISE-2017-778035 of the European Union{\textquoteright}s Horizon 2020 research and innovation programme, Marie Sklodowska-Curie Actions (MSCA) programme. Funding Information: Acknowledgement. This work was partially supported by the European Commission, under European Project 5G-Induce, grant number 101016941. Funding Information: This work is partially supported by Universidad de Le{\'o}n, under the “Programa Propio de Investigaci{\'o}n de la Universidad de Le{\'o}n 2021” grant. Funding Information: Acknowledgements. This research is supported by the project Future Artificial Intelligence Research (FAIR) - PNRR MUR Cod. PE0000013 - CUP: E63C22001940006. Funding Information: Acknowledgements. The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the projects PID2020-117954RB-C22 and TED2021-131311B and Junta de Andalu{\'c}ıa for the project PYC20 RE 078 USE. Funding Information: Acknowledgments. The authors would like to acknowledge the financial support of the Ministerio de Ciencias, Tecnolog{\'i}a e Innovaci{\~A}{\c s}n (Minciencias) through Scholarship Program No. 860. This work has also benefited from a State grant managed by the National Research Agency under the “Investissements d{\textquoteright}Avenir” program with the reference ANR-18-EURE-0021. Work also partially supported by the Spanish project TED2021-132470B-I00, funded by MCIN-AEI-10.13039-501100011033, and the GOMINOLA project (PID2020-118112RB-C21, funded by MCIN-AEI-10.13039-501100011033). Funding Information: CITIC, as a Research Center of the University System of Galicia, is funded by Conseller{\'i}a de Educaci{\'o}n, Universidade e Formaci{\'o}n Profesional of the Xunta de Galicia through the European Regional Development Fund (ERDF) and the Secretar{\'i}a Xeral de Universidades (Ref. ED431G 2019/01). Funding Information: Acknowledgements. The authors would like to thank the Spanish Ministry of Science and Innovation and the Junta de Andaluc{\'i}a for their support within the projects PID2020-117954RB-C21 and TED2021-131311B-C22, PY20-00870, UPO-138516, respectively. The authors would also like to thank the Fundaci{\'o}n Tatiana P{\'e}rez de Guzm{\'a}n el Buenofor the support offered through the Beca Predoctoral en Medioambiente de 2018. Funding Information: Acknowledgements. M{\'i}riam Timiraos{\textquoteright}s research was supported by the “Xunta de Galicia” (Regional Government of Galicia) through grants to industrial PhD (http:// gain.xunta.gal/), under the “Doutoramento Industrial 2022” grant with reference: 04 IN606D 2022 2692965. Funding Information: Acknowledgments. {\'A}lvaro Michelena{\textquoteright}s research was supported by the Spanish Ministry of Universities (https://www.universidades.gob.es/), under the ”Formaci{\'o}n de Profesorado Universitario” grant with reference FPU21/00932. Funding Information: Particular thanks go as well to the conference{\textquoteright}s main sponsors, Startup Ol{\'e}, the CYL-HUB project financed with NEXT-GENERATION funds from the European Union and channeled by Junta de Castilla y Le{\'o}n through the Regional Ministry of Industry, Trade and Employment, BISITE research group at the University of Salamanca, CTC research group at the University of A Coru{\~n}a, and the University of Salamanca. They jointly contributed in an active and constructive manner to the success of this initiative. Funding Information: Acknowledgements. Authors acknowledge funding under grant AI4TES, by the Spanish Min. of Economic Affairs and Digital Transformation and by EU Next Generation EU/PRTR. Carlos J. Gallego acknowledges funding for his scholarship from UPM RP180022025. Funding Information: {\'A}lvaro Michelena{\textquoteright}s research was supported by the Spanish Ministry of Universities (https://www.universidades.gob.es/), under the “Formaci{\'o}n de Profesorado Universi-tario” grant with reference: FPU21/00932. Funding Information: Acknowledgments. The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the project PID2020-117954RB-C21 and the European Regional Development Fund and Junta de Andalu{\'c}ıa for projects PY20-00870 and UPO-138516. Funding Information: M{\'i}riam Timiraos{\textquoteright}s research was supported by the Xunta de Galicia (Regional Government of Galicia) through grants to industrial Ph.D. (http://gain.xunta.gal), under the Doutoramento Industrial 2022 grant with reference: 04 IN606D 2022 2692965. Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.",

year = "2023",

month = aug,

day = "31",

doi = "10.1007/978-3-031-42529-5_21",

language = "English",

pages = "219--228",

}

TY - CONF

T1 - Neuroevolutionary Transfer Learning for Time Series Forecasting

AU - Vellinger, Aymeric

AU - Torres, J. F.

AU - Divina, Federico

AU - Vanhoof, Wim

N1 - Funding Information: Acknowledgements. Iztok Fister Jr. is grateful the Slovenian Research Agency for the financial support under Research Core Funding No. P2-0057. Iztok Fister thanks the Slovenian Research Agency for the financial support under Research Core Funding No. P2-0042 - Digital twin. This research is also supported by the PDE-GIR project from the European Union Horizon 2020 research and innovation programme under the Marie Skodowska-Curie grant agreement No 778035, and the project PID2021-127073OB-I00 of the MCIN/AEI/10.13039/501100011033/FEDER, EU. Funding Information: This research work has been supported by the Spanish national project PID2021-127073OB-I00 of the MCIN/AEI/10.13039/501100011033/FEDER, EU, and the European project PDE-GIR with grant number H2020-MSCA-RISE-2017-778035 of the European Union’s Horizon 2020 research and innovation programme, Marie Sklodowska-Curie Actions (MSCA) programme. Funding Information: Acknowledgement. This work was partially supported by the European Commission, under European Project 5G-Induce, grant number 101016941. Funding Information: This work is partially supported by Universidad de León, under the “Programa Propio de Investigación de la Universidad de León 2021” grant. Funding Information: Acknowledgements. This research is supported by the project Future Artificial Intelligence Research (FAIR) - PNRR MUR Cod. PE0000013 - CUP: E63C22001940006. Funding Information: Acknowledgements. The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the projects PID2020-117954RB-C22 and TED2021-131311B and Junta de Andalućıa for the project PYC20 RE 078 USE. Funding Information: Acknowledgments. The authors would like to acknowledge the financial support of the Ministerio de Ciencias, Tecnología e InnovaciÃşn (Minciencias) through Scholarship Program No. 860. This work has also benefited from a State grant managed by the National Research Agency under the “Investissements d’Avenir” program with the reference ANR-18-EURE-0021. Work also partially supported by the Spanish project TED2021-132470B-I00, funded by MCIN-AEI-10.13039-501100011033, and the GOMINOLA project (PID2020-118112RB-C21, funded by MCIN-AEI-10.13039-501100011033). Funding Information: CITIC, as a Research Center of the University System of Galicia, is funded by Consellería de Educación, Universidade e Formación Profesional of the Xunta de Galicia through the European Regional Development Fund (ERDF) and the Secretaría Xeral de Universidades (Ref. ED431G 2019/01). Funding Information: Acknowledgements. The authors would like to thank the Spanish Ministry of Science and Innovation and the Junta de Andalucía for their support within the projects PID2020-117954RB-C21 and TED2021-131311B-C22, PY20-00870, UPO-138516, respectively. The authors would also like to thank the Fundación Tatiana Pérez de Guzmán el Buenofor the support offered through the Beca Predoctoral en Medioambiente de 2018. Funding Information: Acknowledgements. Míriam Timiraos’s research was supported by the “Xunta de Galicia” (Regional Government of Galicia) through grants to industrial PhD (http:// gain.xunta.gal/), under the “Doutoramento Industrial 2022” grant with reference: 04 IN606D 2022 2692965. Funding Information: Acknowledgments. Álvaro Michelena’s research was supported by the Spanish Ministry of Universities (https://www.universidades.gob.es/), under the ”Formación de Profesorado Universitario” grant with reference FPU21/00932. Funding Information: Particular thanks go as well to the conference’s main sponsors, Startup Olé, the CYL-HUB project financed with NEXT-GENERATION funds from the European Union and channeled by Junta de Castilla y León through the Regional Ministry of Industry, Trade and Employment, BISITE research group at the University of Salamanca, CTC research group at the University of A Coruña, and the University of Salamanca. They jointly contributed in an active and constructive manner to the success of this initiative. Funding Information: Acknowledgements. Authors acknowledge funding under grant AI4TES, by the Spanish Min. of Economic Affairs and Digital Transformation and by EU Next Generation EU/PRTR. Carlos J. Gallego acknowledges funding for his scholarship from UPM RP180022025. Funding Information: Álvaro Michelena’s research was supported by the Spanish Ministry of Universities (https://www.universidades.gob.es/), under the “Formación de Profesorado Universi-tario” grant with reference: FPU21/00932. Funding Information: Acknowledgments. The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the project PID2020-117954RB-C21 and the European Regional Development Fund and Junta de Andalućıa for projects PY20-00870 and UPO-138516. Funding Information: Míriam Timiraos’s research was supported by the Xunta de Galicia (Regional Government of Galicia) through grants to industrial Ph.D. (http://gain.xunta.gal), under the Doutoramento Industrial 2022 grant with reference: 04 IN606D 2022 2692965. Publisher Copyright: © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2023/8/31

Y1 - 2023/8/31

N2 - In this paper, we propose a neuroevolution technique specifically designed for evolving LSTM networks. The proposed technique uses a grammar-based approach to evolve LSTM neural networks for time series prediction tasks, and is based on a previous technique which was designed in order to evolve CNN networks.We use transfer learning in order to reduce the computational time of our approach. We have compared results obtained with other state of the art time series forecasting techniques on twenty time series, which contains data generated by sensors placed on a number of Iberian pigs. Results obtained confirm the effectiveness of the strategy proposed in this work.Overall, we showcase the potential of our proposal in producing precise and efficient deep learning models for time series prediction, as well as the adaptability of transfer learning to new datasets.

AB - In this paper, we propose a neuroevolution technique specifically designed for evolving LSTM networks. The proposed technique uses a grammar-based approach to evolve LSTM neural networks for time series prediction tasks, and is based on a previous technique which was designed in order to evolve CNN networks.We use transfer learning in order to reduce the computational time of our approach. We have compared results obtained with other state of the art time series forecasting techniques on twenty time series, which contains data generated by sensors placed on a number of Iberian pigs. Results obtained confirm the effectiveness of the strategy proposed in this work.Overall, we showcase the potential of our proposal in producing precise and efficient deep learning models for time series prediction, as well as the adaptability of transfer learning to new datasets.

KW - Deep Learning

KW - Neuroevolution

KW - Time series forecasting

UR - http://www.scopus.com/inward/record.url?scp=85172148660&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-42529-5_21

DO - 10.1007/978-3-031-42529-5_21

M3 - Paper

SP - 219

EP - 228

ER -

Neuroevolutionary Transfer Learning for Time Series Forecasting

Résumé

Financement

Accès au document

Autres fichiers et liens

Empreinte digitale

Contient cette citation