Deep learning for the classification of transient noises and signals in the LIGO detectors
Autor(a) principal: | |
---|---|
Data de Publicação: | 2022 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
Texto Completo: | http://hdl.handle.net/10773/36277 |
Resumo: | In this work, data from the aLIGO detectores collected during the first two aLIGO and AdV observing runs (O1 and O2), in the form of spectrograms, were classified using Deep Learning models based on Convolutional Neural Networks. As well as training models from scratch, pre-trained models were also employed, and their performance compared. Initially, a brief theoretical introduction on gravitational wave detection was performed, focusing on the LIGO detectors. In addition, the foundations of Deep Learning and current best practices for the training of image classification models were also presented. The computational experiments showed that encoding information from different time windows in the different colour channels enhanced the performance of the models and that small architectures were capable of separating the 22 classes present in the Gravity Spy dataset. Moreover, transfer learning was able to accelerate the training process and achieve classifiers with competitive performance. The best models obtained a macro-averaged F1 score of 96.84% (fine-tuned model) and 97.18% (baseline trained from scratch), which are in line with the best results in the literature for the same dataset. In addition, these models were evaluated on real gravitational wave signals from Compact Binary Coalescences from the first two aLIGO and AdV observing runs, and they achieved recalls of 75% and 25%, respectively, while only having been trained with a small number of signals from gravitational wave simulations. |
id |
RCAP_335b86fc95c2661d22243264a6eced25 |
---|---|
oai_identifier_str |
oai:ria.ua.pt:10773/36277 |
network_acronym_str |
RCAP |
network_name_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
repository_id_str |
7160 |
spelling |
Deep learning for the classification of transient noises and signals in the LIGO detectorsDeep learningaLIGOTransient noiseGravitational wavesImage classificationIn this work, data from the aLIGO detectores collected during the first two aLIGO and AdV observing runs (O1 and O2), in the form of spectrograms, were classified using Deep Learning models based on Convolutional Neural Networks. As well as training models from scratch, pre-trained models were also employed, and their performance compared. Initially, a brief theoretical introduction on gravitational wave detection was performed, focusing on the LIGO detectors. In addition, the foundations of Deep Learning and current best practices for the training of image classification models were also presented. The computational experiments showed that encoding information from different time windows in the different colour channels enhanced the performance of the models and that small architectures were capable of separating the 22 classes present in the Gravity Spy dataset. Moreover, transfer learning was able to accelerate the training process and achieve classifiers with competitive performance. The best models obtained a macro-averaged F1 score of 96.84% (fine-tuned model) and 97.18% (baseline trained from scratch), which are in line with the best results in the literature for the same dataset. In addition, these models were evaluated on real gravitational wave signals from Compact Binary Coalescences from the first two aLIGO and AdV observing runs, and they achieved recalls of 75% and 25%, respectively, while only having been trained with a small number of signals from gravitational wave simulations.Neste trabalho, dados dos detetores aLIGO recolhidos nos dois primeiros períodos de observação de LIGO e Virgo (O1 e O2), na forma de espectrogramas, foram classificados usando modelos de Deep Learning baseados em redes neuronais convolucionais. Além de serem usados modelos treinados do zero, também se testaram modelos pré-treinados, e os resultados foram comparados. Para isso, começou por se fazer uma breve introdução às ondas gravitacionais e sua deteção nos detetores de LIGO. Foram também introduzidos os fundamentos relacionados com algoritmos de Deep Learning e das boas práticas para o treino de modelos para a classificação de imagens. Verificou-se que usar os diferentes canais de cor das imagens para apresentar informação com diferentes janelas temporais melhora os resultados dos modelos e que, além disso, arquiteturas pequenas são capazes de separar eficazmente as 22 classes presentes no dataset Gravity Spy. Adicionalmente, a técnica de transfer learning permite acelerar a fase de treino e obter classificadores com um desempenho competitivo. Os melhores modelos obtiveram um F1-score médio (macro) de 96.84% para o modelo pré-treinado e de 97.18% para o modelo base treinado do zero. Estes resultados estão em linha com os melhores resultados encontrados na literatura para o mesmo dataset. Adicionalmente, os modelos foram testados em sinais reais de ondas gravitacionais de Coalescências Binárias Compactas detetadas por LIGO, obtendo sensibilidades de, respetivamente, 25% e 75%, apesar de terem sido treinados com um número reduzido de sinais provenientes de simulações de ondas gravitacionais.2023-02-10T09:06:05Z2022-12-15T00:00:00Z2022-12-15info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10773/36277engFernandes, Tiago Saraivainfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2024-02-22T12:10:05Zoai:ria.ua.pt:10773/36277Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-20T03:07:11.254905Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse |
dc.title.none.fl_str_mv |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
title |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
spellingShingle |
Deep learning for the classification of transient noises and signals in the LIGO detectors Fernandes, Tiago Saraiva Deep learning aLIGO Transient noise Gravitational waves Image classification |
title_short |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
title_full |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
title_fullStr |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
title_full_unstemmed |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
title_sort |
Deep learning for the classification of transient noises and signals in the LIGO detectors |
author |
Fernandes, Tiago Saraiva |
author_facet |
Fernandes, Tiago Saraiva |
author_role |
author |
dc.contributor.author.fl_str_mv |
Fernandes, Tiago Saraiva |
dc.subject.por.fl_str_mv |
Deep learning aLIGO Transient noise Gravitational waves Image classification |
topic |
Deep learning aLIGO Transient noise Gravitational waves Image classification |
description |
In this work, data from the aLIGO detectores collected during the first two aLIGO and AdV observing runs (O1 and O2), in the form of spectrograms, were classified using Deep Learning models based on Convolutional Neural Networks. As well as training models from scratch, pre-trained models were also employed, and their performance compared. Initially, a brief theoretical introduction on gravitational wave detection was performed, focusing on the LIGO detectors. In addition, the foundations of Deep Learning and current best practices for the training of image classification models were also presented. The computational experiments showed that encoding information from different time windows in the different colour channels enhanced the performance of the models and that small architectures were capable of separating the 22 classes present in the Gravity Spy dataset. Moreover, transfer learning was able to accelerate the training process and achieve classifiers with competitive performance. The best models obtained a macro-averaged F1 score of 96.84% (fine-tuned model) and 97.18% (baseline trained from scratch), which are in line with the best results in the literature for the same dataset. In addition, these models were evaluated on real gravitational wave signals from Compact Binary Coalescences from the first two aLIGO and AdV observing runs, and they achieved recalls of 75% and 25%, respectively, while only having been trained with a small number of signals from gravitational wave simulations. |
publishDate |
2022 |
dc.date.none.fl_str_mv |
2022-12-15T00:00:00Z 2022-12-15 2023-02-10T09:06:05Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/10773/36277 |
url |
http://hdl.handle.net/10773/36277 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.source.none.fl_str_mv |
reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação instacron:RCAAP |
instname_str |
Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
instacron_str |
RCAAP |
institution |
RCAAP |
reponame_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
collection |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
repository.name.fl_str_mv |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
repository.mail.fl_str_mv |
|
_version_ |
1799137727288967168 |