簡易檢索 / 詳目顯示

回結果列表
研究生： 張睿哲
Chang, Ruei-Jhe
論文名稱： 使用卷積神經網路偵測緻密雙星合併的重力波
Detecting Gravitational Waves from Compact Binary Coalescences using Convolutional Neural Network
指導教授： 李君樂
Li, Kwan-Lok
共同指導教授: 林峻哲
Lin, Chun-Che
學位類別： 碩士
Master
系所名稱： 理學院 - 物理學系
Department of Physics
論文出版年： 2024
畢業學年度： 112
語文別： 英文
論文頁數： 54
中文關鍵詞： 重力波 、緻密雙星合併 、深度學習 、卷積神經網路
外文關鍵詞： gravitational waves, compact binary coalescence, deep learning, convolutional neural network
相關次數： 點閱：114  下載：11 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 自2015年LIGO首次探測到重力波以來，研究重力波已開啟了解宇宙的新途徑。然而，目前用於偵測緻密雙星合併的重力波的管道大多依賴匹配濾波，這是一種需要大量計算資源實現低延遲偵測的方法。基於深度學習的替代方法為這一挑戰提供了解決方案。到目前為止，已經有許多基於卷積神經網路（CNN）的深度學習算法被應用於重力波偵測，尤其是雙黑洞（BBH）合併訊號的偵測。然而，很少有研究探討額外將雙中子星 (BNS) 合併的訊號納入基於 CNN 的重力波偵測模型中。 在用於偵測 BBH 訊號的 CNN 模型的建構方法中，我們的研究目的在探討這些方法是否適合建構 CNN 模型以從雜訊中偵測 BBH 和 BNS 訊號。 我們將這些建構 CNN 模型的方法分為三類：訓練資料類型、訓練策略和 CNN 架構。 我們的研究表明，使用模擬探測器雜訊產生的資料訓練CNN 會導致靈敏度更高，但誤報也會更高，而使用真實探測器雜訊產生的資料進行訓練會導致靈敏度降低，但誤報率也會降低。在我們測試的訓練策略中，使用低信噪比數據訓練CNN是一種有效的重力波偵測策略。此外，在 CNN 架構中加入批量歸一化層可以提高偵測訊號的靈敏度，但會增加偵測的時間成本。 有趣的是，在使用 3 到 6 個卷積層的測試中我們觀察到，一旦模型的超參數被適當選擇，使用更深的 CNN 似乎有利於 GW 偵測。

    Since the first detection of gravitational waves (GW) by LIGO in 2015, studying GW has opened new avenues for understanding the universe. However, current GW detection pipelines for the GW from compact binary coalescences mostly rely on matched filtering, a computationally intensive technique which requires substantial computing resources for low-latency detections. An alternative approach based on deep learning offers a solution to this challenge. So far, numerous deep learning algorithms that based on convolutional neural network (CNN) have been applied to GW detection, especially for detecting signals of binary black hole (BBH) mergers. This leads to various methods for building CNN models. However, few studies have explored the additional inclusion of signals of binary neutron star (BNS) mergers in CNN-based GW detection models. Among the CNN construction methods applied for the BBH signal detection, our study aims to explore whether these methods are suitable for constructing CNNs to detect both BBH and BNS signals from noises. We analyzed these methods of constructing CNN models based on three aspects: training data type, training strategy, and CNN architecture. Our studies show that training a CNN with the data generated from simulated detector noise results in higher sensitivity but also higher false positives, while training with the data generated from real detector noise leads to lower sensitivity but also lower false positives. Among the training strategies we examined, training the CNN on low signal-to-noise ratio data is an effective strategy for GW detection. Additionally, adding batch normalization layers to the CNN architecture enhances the sensitivity but increases time costs of their detections. Interestingly, in our tests involving 3 to 6 convolutional layers, we observed that using deeper CNNs appears to be beneficial for GW detection once the hyperparameters of the model are appropriately selected.

    中文摘要 I Abstract II Acknowledgements III Contents IV List of Tables VI List of Figures VIII 1 Introduction 1 1.1 Background 1 1.2 Gravitational Wave detection using Matched Filtering 2 1.3 Gravitational Wave Detection Using Convolutional Neural Network 3 1.4 Motivation and Research Methodology 3 2 Deep Neural Network 5 2.1 Basic Concept of Neural Network 5 2.2 Basic concept of Convolutional Neural Network 6 2.2.1 Convolutional layer 7 2.2.2 Pooling layer 8 2.2.3 Dense layer 9 2.3 Training a Neural Network 9 2.3.1 Data 9 2.3.2 Cost function 10 2.3.3 Optimizer 11 3 Training Data Type 12 3.1 Data Preparation 12 3.1.1 Simulated detector noise 12 3.1.2 Real detector noise 12 3.1.3 Signal data 13 3.2 General setup 13 3.2.1 CNN architecture and training configuration 13 3.2.2 Data Set construction 15 3.3 Result and Discussion 16 4 Training Strategy 19 4.1 Method 19 4.1.1 Training Strategy 19 4.2 General setup 20 4.2.1 CNN architecture and training configuration 20 4.2.2 Data Set construction 21 4.3 Result and Discussion 21 5 CNN Architecture 24 5.1 Method 24 5.1.1 Batch Normalization in CNN 24 5.1.2 Different number of convolutional layers in CNN 26 5.1.3 Hyperparameter tuning 26 5.2 General setup 27 5.2.1 CNN architecture and training configuration 27 5.2.2 Data Set construction 28 5.3 Result and Discussion 28 5.3.1 Batch Normalization in CNN 28 5.3.2 Different number of convolutional layers in CNN 31 6 Conclusion 36 References 37 Appendix 42 A.1 Development Environment 42 A.1.1 Hardware environment 42 A.1.2 Software Environment 42

    [1] LIGO Scientific Collaboration, J. Aasi, B. P. Abbott, and et al. Advanced LIGO. Classical and Quantum Gravity, 32(7):074001, April 2015.
    [2] B. P. Abbott, R. Abbott, T. D. Abbott, and et al. Observation of Gravitational Waves from a Binary Black Hole Merger. , 116(6):061102, February 2016.
    [3] F. Acernese, M. Agathos, K. Agatsuma, D. Aisa, and et al. Advanced Virgo: a second- generation interferometric gravitational wave detector. Classical and Quantum Gravity, 32(2):024001, January 2015.
    [4] Kagra Collaboration, T. Akutsu, M. Ando, and et al. KAGRA: 2.5 generation interfer- ometric gravitational wave detector. Nature Astronomy, 3:35–40, January 2019.
    [5] The LIGO Scientific Collaboration, the Virgo Collaboration, the KAGRA Collabo- ration, and et al. GWTC-3: Compact Binary Coalescences Observed by LIGO and Virgo During the Second Part of the Third Observing Run. arXiv e-prints, page arXiv:2111.03606, November 2021.
    [6] B. P. Abbott, R. Abbott, T. D. Abbott, and et al. Prospects for observing and localizing gravitational-wave transients with Advanced LIGO, Advanced Virgo and KAGRA. Living Reviews in Relativity, 23(1):3, September 2020.
    [7] Craig Cahillane and Georgia Mansell. Review of the Advanced LIGO Gravitational Wave Observatories Leading to Observing Run Four. Galaxies, 10(1):36, February 2022.
    [8] B. P. Abbott, R. Abbott, T. D. Abbott, and et al. GW170817: Observation of Gravita- tional Waves from a Binary Neutron Star Inspiral. , 119(16):161101, October 2017.
    [9] B. S. Sathyaprakash and Bernard F. Schutz. Physics, Astrophysics and Cosmology with Gravitational Waves. Living Reviews in Relativity, 12(1):2, December 2009.
    [10] Samantha A. Usman, Alexander H. Nitz, Ian W. Harry, and et al. The PyCBC search for gravitational waves from compact binary coalescence. Classical and Quantum Gravity, 33(21):215004, November 2016.
    [11] Bruce Allen, Warren G. Anderson, Patrick R. Brady, Duncan A. Brown, and Jolien D. E. Creighton. FINDCHIRP: An algorithm for detection of gravitational waves from inspiraling compact binaries. , 85(12):122006, June 2012.
    [12] Benjamin J. Owen and B. S. Sathyaprakash. Matched filtering of gravitational waves from inspiraling compact binaries: Computational cost and template placement. , 60(2):022002, July 1999.
    [13] Kipp Cannon, Sarah Caudill, Chiwai Chan, and et al. GstLAL: A software framework for gravitational wave discovery. SoftwareX, 14:100680, June 2021.
    [14] Tito Dal Canton, Alexander H. Nitz, Bhooshan Gadre, and et al. Real-time Search for Compact Binary Mergers in Advanced LIGO and Virgo’s Third Observing Run Using PyCBC Live. , 923(2):254, December 2021.
    [15] F. Aubin, F. Brighenti, R. Chierici, and et al. The MBTA pipeline for detecting compact binary coalescences in the third LIGO-Virgo observing run. Classical and Quantum Gravity, 38(9):095004, May 2021.
    [16] Qi Chu, Manoj Kovalam, Linqing Wen, and et al. Spiir online coherent pipeline to search for gravitational waves from compact binary coalescences. Phys. Rev. D, 105:024023, Jan 2022.
    [17] Koustav Chandra, Juan Calderón Bustillo, Archana Pai, and I. W. Harry. First gravitational-wave search for intermediate-mass black hole mergers with higher-order harmonics. , 106(12):123003, December 2022.
    [18] Rahul Dhurkunde and Alexander H. Nitz. Sensitivity of spin-aligned searches for neutron star-black hole systems using future detectors. , 106(10):103035, November 2022.
    [19] Daniel George and E. A. Huerta. Deep neural networks to enable real-time multimes- senger astrophysics. , 97(4):044039, February 2018.
    [20] Daniel George and E. A. Huerta. Deep Learning for real-time gravitational wave detection and parameter estimation: Results with Advanced LIGO data. Physics Letters B, 778:64–70, March 2018.
    [21] Heming Xia, Lijing Shao, Junjie Zhao, and Zhoujian Cao. Improved deep learning techniques in gravitational-wave data analysis. , 103(2):024040, January 2021.
    [22] Plamen G. Krastev, Kiranjyot Gill, V. Ashley Villar, and Edo Berger. Detection and parameter estimation of gravitational waves from binary neutron-star mergers in real LIGO data using deep learning. Physics Letters B, 815:136161, April 2021.
    [23] Richard Qiu, Plamen G. Krastev, Kiranjyot Gill, and Edo Berger. Deep learning detection and classification of gravitational waves from neutron star-black hole mergers. Physics Letters B, 840:137850, May 2023.
    [24] Hunter Gabbard, Michael Williams, Fergus Hayes, and Chris Messenger. Match- ing Matched Filtering with Deep Networks for Gravitational-Wave Astronomy. , 120(14):141103, April 2018.
    [25] Plamen G. Krastev. Real-time detection of gravitational waves from binary neutron stars using artificial neural networks. Physics Letters B, 803:135330, April 2020.
    [26] Warren Mcculloch and Walter Pitts. A logical calculus of ideas immanent in nervous activity. Bulletin of Mathematical Biophysics, 5:127–147, 1943.
    [27] F. Rosenblatt. The perceptron: A probabilistic model for information storage and organization in the brain. Psychological Review, 65(6):386–408, 1958.
    [28] Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learn- ing applied to document recognition. In Proceedings of the IEEE, volume 86, pages 2278–2324, 1998.
    [29] Ronan Collobert and Jason Weston. A unified architecture for natural language pro- cessing: Deep neural networks with multitask learning. In Proceedings of the 25th International Conference on Machine Learning, ICML ’08, pages 160–167, New York, NY, USA, 2008. ACM.
    [30] Yann Lecun. Generalization and network design strategies. Elsevier, 1989.
    [31] Sebastian Ruder. An overview of gradient descent optimization algorithms. arXiv e-prints, page arXiv:1609.04747, September 2016.
    [32] Diederik P. Kingma and Jimmy Ba. Adam: A Method for Stochastic Optimization. arXiv e-prints, page arXiv:1412.6980, December 2014.
    [33] Ian Goodfellow, Yoshua Bengio, and Aaron Courville. Deep Learning. MIT Press, 2016. http://www.deeplearningbook.org.
    [34] Marlin B. Schäfer, Ondřej Zelenka, Alexander H. Nitz, Frank Ohme, and Bernd Brügmann. Training strategies for deep learning gravitational-wave searches. , 105(4):043002, February 2022.
    [35] LIGO Scientific Collaboration. LALSuite: LIGO Scientific Collaboration Algorithm Library Suite. Astrophysics Source Code Library, record ascl:2012.021, December 2020.
    [36] R. B. Blackman and J. W. Tukey. The measurement of power spectra from the point of view of communications engineering — part i. The Bell System Technical Journal, 37(1):185–282, 1958.
    [37] Barak Zackay, Tejaswi Venumadhav, Javier Roulet, Liang Dai, and Matias Zaldarriaga. Detecting gravitational waves in data with non-stationary and non-Gaussian noise. , 104(6):063034, September 2021.
    [38] Timothy D. Gebhard, NikiKilbertus, IanHarry, and Bernhard Schölkopf. Convolutional neural networks: A magic bullet for gravitational-wave detection? , 100(6):063015, September 2019.
    [39] R. Abbott, T. D. Abbott, S. Abraham, and et al. Open data from the first and second observing runs of Advanced LIGO and Advanced Virgo. SoftwareX, 13:100658, January 2021.
    [40] Caleb Devine, Zachariah B. Etienne, and Sean T. McWilliams. Optimizing spinning time-domain gravitational waveforms for advanced LIGO data analysis. Classical and Quantum Gravity, 33(12):125025, June 2016.
    [41] Francesco Messina, Reetika Dudi, Alessandro Nagar, and Sebastiano Bernuzzi. Quasi- 5.5PN TaylorF2 approximant for compact binaries: Point-mass phasing and impact on the tidal polarizability inference. , 99(12):124051, June 2019.
    [42] Alexander H. Nitz, Sumit Kumar, Yi-Fan Wang, Shilpa Kastha, Shichao Wu, Marlin Schäfer, Rahul Dhurkunde, and Collin D. Capano. 4-OGC: Catalog of Gravitational Waves from Compact Binary Mergers. , 946(2):59, April 2023.
    [43] Genevieve B. Orr and Klaus-Robert Müller, editors. Neural Networks: Tricks of the Trade, volume 1524 of Lecture Notes in Computer Science. Springer, 1998.
    [44] Gonzalo Morrás, Jose Francisco Nuño Siles, Juan García-Bellido, and Ester Ruiz Morales. False alarms induced by Gaussian noise in gravitational wave detectors. , 107(2):023027, January 2023.
    [45] Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th Annual International Conference on Machine Learning, ICML ’09, page 41–48, New York, NY, USA, 2009. Association for Com- puting Machinery.
    [46] Marlin B. Schäfer, Frank Ohme, and Alexander H. Nitz. Detection of gravitational- wave signals from binary neutron star mergers using machine learning. , 102(6):063015, September 2020.
    [47] Sergey Ioffe and Christian Szegedy. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. arXiv e-prints, page arXiv:1502.03167, February 2015.
    [48] Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. Imagenet large scale visual recognition challenge. Int. J. Comput. Vision, 115(3):211–252, dec 2015.
    [49] Grégory Baltus, Justin Janquart, Melissa Lopez, Amit Reza, Sarah Caudill, and Jean- René Cudell. Convolutional neural networks for the detection of the early inspiral of a gravitational-wave signal. , 103(10):102003, May 2021.

    下載圖示 校內：立即公開
    校外：立即公開
    QR CODE