DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics

Tom Dooney; Stefano Bromuri; Lyana Curier

doi:10.1109/BigData55660.2022.10021080

DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics

Tom Dooney^*, Stefano Bromuri, Lyana Curier

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference Article in proceeding › Academic › peer-review

Abstract

Simulating time-domain observations of gravitational wave (GW) detector environments will allow for a better understanding of GW sources, augment datasets for GW signal detection and help in characterizing the noise of the detectors, leading to better physics. This paper presents a novel approach to simulating fixed-length time-domain signals using a three-player Wasserstein Generative Adversarial Network (WGAN), called DVGAN, that includes an auxiliary discriminator that discriminates on the derivatives of input signals. An ablation study is used to compare the effects of including adversarial feedback from an auxiliary derivative discriminator with a vanilla two-player WGAN. We show that discriminating on derivatives can stabilize the learning of GAN components on 1D continuous signals during their training phase. This results in smoother generated signals that are less distinguishable from real samples and better capture the distributions of the training data. DVGAN is also used to simulate real transient noise events captured in the advanced LIGO GW detector.

Original language	English
Title of host publication	Proceedings
Subtitle of host publication	2022 IEEE International Conference on Big Data, Big Data 2022
Editors	Shusaku Tsumoto, Yukio Ohsawa, Lei Chen, Dirk Van den Poel, Xiaohua Hu, Yoichi Motomura, Takuya Takagi, Lingfei Wu, Ying Xie, Akihiro Abe, Vijay Raghavan
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	5468-5477
Number of pages	10
ISBN (Electronic)	9781665480451
DOIs	https://doi.org/10.1109/BigData55660.2022.10021080
Publication status	Published - 2022
Event	2022 IEEE International Conference on Big Data, Big Data 2022 - Osaka, Japan Duration: 17 Dec 2022 → 20 Dec 2022 https://bigdataieee.org/BigData2022/

Conference

Conference	2022 IEEE International Conference on Big Data, Big Data 2022
Country/Territory	Japan
City	Osaka
Period	17/12/22 → 20/12/22
Internet address	https://bigdataieee.org/BigData2022/

Keywords

Big Data
Data Augmentation
Deep Learning
GAN
Gravitational Waves
Machine Learning
Physics
Time Series

Access to Document

10.1109/BigData55660.2022.10021080

Cite this

Dooney, T., Bromuri, S., & Curier, L. (2022). DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics. In S. Tsumoto, Y. Ohsawa, L. Chen, D. Van den Poel, X. Hu, Y. Motomura, T. Takagi, L. Wu, Y. Xie, A. Abe, & V. Raghavan (Eds.), Proceedings: 2022 IEEE International Conference on Big Data, Big Data 2022 (pp. 5468-5477). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/BigData55660.2022.10021080

Dooney, Tom ; Bromuri, Stefano ; Curier, Lyana. / DVGAN : Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics. Proceedings: 2022 IEEE International Conference on Big Data, Big Data 2022. editor / Shusaku Tsumoto ; Yukio Ohsawa ; Lei Chen ; Dirk Van den Poel ; Xiaohua Hu ; Yoichi Motomura ; Takuya Takagi ; Lingfei Wu ; Ying Xie ; Akihiro Abe ; Vijay Raghavan. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 5468-5477

@inproceedings{1e955224882a40ef9b42642ddf299802,

title = "DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics",

abstract = "Simulating time-domain observations of gravitational wave (GW) detector environments will allow for a better understanding of GW sources, augment datasets for GW signal detection and help in characterizing the noise of the detectors, leading to better physics. This paper presents a novel approach to simulating fixed-length time-domain signals using a three-player Wasserstein Generative Adversarial Network (WGAN), called DVGAN, that includes an auxiliary discriminator that discriminates on the derivatives of input signals. An ablation study is used to compare the effects of including adversarial feedback from an auxiliary derivative discriminator with a vanilla two-player WGAN. We show that discriminating on derivatives can stabilize the learning of GAN components on 1D continuous signals during their training phase. This results in smoother generated signals that are less distinguishable from real samples and better capture the distributions of the training data. DVGAN is also used to simulate real transient noise events captured in the advanced LIGO GW detector.",

keywords = "Big Data, Data Augmentation, Deep Learning, GAN, Gravitational Waves, Machine Learning, Physics, Time Series",

author = "Tom Dooney and Stefano Bromuri and Lyana Curier",

note = "Funding Information: 1Python code can be found at https://git.ligo.org/tom.dooney/dvgan. Computational resources were provided by the LIGO Laboratory and supported by the National Science Foundation Grants No. PHY-0757058 and No. PHY-0823459. Publisher Copyright: {\textcopyright} 2022 IEEE.; 2022 IEEE International Conference on Big Data, Big Data 2022 ; Conference date: 17-12-2022 Through 20-12-2022",

year = "2022",

doi = "10.1109/BigData55660.2022.10021080",

language = "English",

pages = "5468--5477",

editor = "Shusaku Tsumoto and Yukio Ohsawa and Lei Chen and {Van den Poel}, Dirk and Xiaohua Hu and Yoichi Motomura and Takuya Takagi and Lingfei Wu and Ying Xie and Akihiro Abe and Vijay Raghavan",

booktitle = "Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

url = "https://bigdataieee.org/BigData2022/",

}

Dooney, T , Bromuri, S & Curier, L 2022, DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics. in S Tsumoto, Y Ohsawa, L Chen, D Van den Poel, X Hu, Y Motomura, T Takagi, L Wu, Y Xie, A Abe & V Raghavan (eds), Proceedings: 2022 IEEE International Conference on Big Data, Big Data 2022. Institute of Electrical and Electronics Engineers Inc., pp. 5468-5477, 2022 IEEE International Conference on Big Data, Big Data 2022, Osaka, Japan, 17/12/22. https://doi.org/10.1109/BigData55660.2022.10021080

DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics. / Dooney, Tom ; Bromuri, Stefano ; Curier, Lyana.
Proceedings: 2022 IEEE International Conference on Big Data, Big Data 2022. ed. / Shusaku Tsumoto; Yukio Ohsawa; Lei Chen; Dirk Van den Poel; Xiaohua Hu; Yoichi Motomura; Takuya Takagi; Lingfei Wu; Ying Xie; Akihiro Abe; Vijay Raghavan. Institute of Electrical and Electronics Engineers Inc., 2022. p. 5468-5477.

Research output: Chapter in Book/Report/Conference proceeding › Conference Article in proceeding › Academic › peer-review

TY - GEN

T1 - DVGAN

T2 - 2022 IEEE International Conference on Big Data, Big Data 2022

AU - Dooney, Tom

AU - Bromuri, Stefano

AU - Curier, Lyana

N1 - Funding Information: 1Python code can be found at https://git.ligo.org/tom.dooney/dvgan. Computational resources were provided by the LIGO Laboratory and supported by the National Science Foundation Grants No. PHY-0757058 and No. PHY-0823459. Publisher Copyright: © 2022 IEEE.

PY - 2022

Y1 - 2022

N2 - Simulating time-domain observations of gravitational wave (GW) detector environments will allow for a better understanding of GW sources, augment datasets for GW signal detection and help in characterizing the noise of the detectors, leading to better physics. This paper presents a novel approach to simulating fixed-length time-domain signals using a three-player Wasserstein Generative Adversarial Network (WGAN), called DVGAN, that includes an auxiliary discriminator that discriminates on the derivatives of input signals. An ablation study is used to compare the effects of including adversarial feedback from an auxiliary derivative discriminator with a vanilla two-player WGAN. We show that discriminating on derivatives can stabilize the learning of GAN components on 1D continuous signals during their training phase. This results in smoother generated signals that are less distinguishable from real samples and better capture the distributions of the training data. DVGAN is also used to simulate real transient noise events captured in the advanced LIGO GW detector.

AB - Simulating time-domain observations of gravitational wave (GW) detector environments will allow for a better understanding of GW sources, augment datasets for GW signal detection and help in characterizing the noise of the detectors, leading to better physics. This paper presents a novel approach to simulating fixed-length time-domain signals using a three-player Wasserstein Generative Adversarial Network (WGAN), called DVGAN, that includes an auxiliary discriminator that discriminates on the derivatives of input signals. An ablation study is used to compare the effects of including adversarial feedback from an auxiliary derivative discriminator with a vanilla two-player WGAN. We show that discriminating on derivatives can stabilize the learning of GAN components on 1D continuous signals during their training phase. This results in smoother generated signals that are less distinguishable from real samples and better capture the distributions of the training data. DVGAN is also used to simulate real transient noise events captured in the advanced LIGO GW detector.

KW - Big Data

KW - Data Augmentation

KW - Deep Learning

KW - GAN

KW - Gravitational Waves

KW - Machine Learning

KW - Physics

KW - Time Series

U2 - 10.1109/BigData55660.2022.10021080

DO - 10.1109/BigData55660.2022.10021080

M3 - Conference Article in proceeding

AN - SCOPUS:85147931986

SP - 5468

EP - 5477

BT - Proceedings

A2 - Tsumoto, Shusaku

A2 - Ohsawa, Yukio

A2 - Chen, Lei

A2 - Van den Poel, Dirk

A2 - Hu, Xiaohua

A2 - Motomura, Yoichi

A2 - Takagi, Takuya

A2 - Wu, Lingfei

A2 - Xie, Ying

A2 - Abe, Akihiro

A2 - Raghavan, Vijay

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 17 December 2022 through 20 December 2022

ER -

Dooney T , Bromuri S , Curier L. DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics. In Tsumoto S, Ohsawa Y, Chen L, Van den Poel D, Hu X, Motomura Y, Takagi T, Wu L, Xie Y, Abe A, Raghavan V, editors, Proceedings: 2022 IEEE International Conference on Big Data, Big Data 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 5468-5477 doi: 10.1109/BigData55660.2022.10021080

DVGAN: Stabilize Wasserstein GAN training for time-domain Gravitational Wave physics

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this