基于深度学习和高频 GNSS 的大震震源参数快速反演研究——以 2021 年 Mw 7.4 玛多地震为例

近年来，受益于海量测震数据的积累和算力的飞速提升，深度学习发挥了其在复杂特征提取、识别等方面的技术优势，使得中小地震事件检测、震相识别、震源参数测定等的准确性、可靠性与时效性大幅提升。然而，对于大震（M≥7），由于其发震频率低，震例相对稀缺，且存在发震时近场区域地震仪限幅、强震仪基线漂移等问题，针对大震震源参数快速反演的深度学习框架目前尚不多见。

高频全球卫星导航系统（Global Navigation Satellite System，GNSS）能直接以cm甚至mm级精度测量任意大小同震位移，弥补了传统测震台站的观测缺陷。本文以2021年M_w 7.4玛多地震为例，建立了一种基于高频GNSS同震观测值的大震震源参数深度学习框架CLEAR（Continental Large Earthquake Agile Response）。为克服大震震例稀缺问题，利用震级、破裂面积、破裂时长三者之间的尺度关系（scaling law）并施加随机扰动，结合发震断层江错断裂带几何信息，本文首先合成了18000例M_w 6.4到M_w 8.3仿真地震及其对应的高频GNSS同震位移波形，并通过加噪和删站进一步提升真实性。基于此数据集，本文开发了地震定位模型（CLEAR-P）和震级-滑动分布模型（CLEAR-MS）。对仿真测试集，震中平均误差为7.9 km，震级估算准确度可达99%，平均误差0.06级，最大滑移平均误差为1.26 m，且滑动分布具有高于0.7的交并比和结构相似性指数。

以玛多地震实测GNSS同震位移波形为输入，该系统估计震源位置为（34.58°N, 98.44°E, 10.5 km），震后32秒以超过60%的概率估算震级为M_w 7.52，比现有基于GNSS峰值地面位移确定震级的方法提前了约20秒，震后64秒更新至超过90%概率M_w 7.43。64秒时在预设断层上得到地震破裂范围，最大滑移估值约6 m，并确定了至少两个主要滑移凹凸体，与前人基于多源观测数据联合反演得到的破裂分布接近。本文工作为基于高频GNSS的大震震后快速响应提供了重要参考。

[1] SMITH K. Environmental hazards: assessing risk and reducing disaster[M]. Routledge, 2013.
[2] MIZUTORI M, GUHA-SAPIR D. Human Cost of Disasters: An Overview of the Last 20 Years[Z]. CRED and UNDRR. 2020.www.undrr.org/sites/default/files/inline-files/Human%20Cost%20of%20Disasters%202000-2019%20FINAL.pdf
[3] 国务院办公厅. 国家防震减灾规划（2006—2020年）[Z]. 2006
[4] XU P, LU X, ZUO K, et al. Post-Wenchuan earthquake reconstruction and development in China[J]. Disaster and development: examining global issues and cases, 2014: 427-45.
[5] DANIELL J E, KHAZAI B, WENZEL F, et al. The CATDAT damaging earthquakes database[J]. Nat Hazard Earth Sys, 2011, 11(8): 2235-51.
[6] 陈运泰. 地震预测:回顾与展望[J]. 中国科学（地球科学）, 2009, 039(012): 1633-58.
[7] 温联星, 陈颙, 于晟. 我国地震减灾中地震学面临的巨大挑战[M]. 我国地震减灾中地震学面临的巨大挑战, 2011.
[8] ALLEN R M, GASPARINI P, KAMIGAICHI O, et al. The Status of Earthquake Early Warning around the World: An Introductory Overview[J]. Seismol Res Lett, 2009, 80(5): 682-+.
[9] ZOLLO A, FESTA G, EMOLO A, et al. Source characterization for earthquake early warning[J]. Encyclopedia of Earthquake Engineering, 2014: 1-21.
[10] ZHANG J H, HAO J L, ZHAO X, et al. Restoration of clipped seismic waveforms using projection onto convex sets method[J]. Sci Rep-Uk, 2016, 6.
[11] GELLER R J. Scaling Relations for Earthquake Source Parameters and Magnitudes[J]. B Seismol Soc Am, 1976, 66(5): 1501-23.
[12] GOLDBERG D E, MELGAR D, BOCK Y. Seismogeodetic P-wave Amplitude: No Evidence for Strong Determinism[J]. Geophys Res Lett, 2019, 46(20): 11118-26.
[13] RYDELEK P, HORIUCHI S. Earth science: Is earthquake rupture deterministic?[J]. Nature, 2006, 442(7100): E5-E6.
[14] HOSHIBA M, IWAKIRI K, HAYASHIMOTO N, et al. Outline of the 2011 off the Pacific coast of Tohoku Earthquake (M w 9.0) -Earthquake Early Warning and observed seismic intensity-[J]. Earth Planets Space, 2011, 63(7): 547-51.
[15] BÖSE M, HEATON T, HAUKSSON E. Rapid Estimation of Earthquake Source and Ground-Motion Parameters for Earthquake Early Warning Using Data from a Single Three-Component Broadband or Strong-Motion Sensor[J]. B Seismol Soc Am, 2012, 102(2): 738-50.
[16] WU Y M, KANAMORI H. Development of an earthquake early warning system using real-time strong motion signals[J]. Sensors-Basel, 2008, 8(1): 1-9.
[17] BOORE D M, BOMMER J J. Processing of strong-motion accelerograms: needs, options and consequences[J]. Soil Dyn Earthq Eng, 2005, 25(2): 93-115.
[18] CROWELL B W, MELGAR D, BOCK Y, et al. Earthquake magnitude scaling using seismogeodetic data[J]. Geophys Res Lett, 2013, 40(23): 6089-94.
[19] GRAPENTHIN R, JOHANSON I A, ALLEN R M. Operational real-time GPS-enhanced earthquake early warning[J]. J Geophys Res-Sol Ea, 2014, 119(10): 7944-65.
[20] MURRAY J R, CROWELL B W, GRAPENTHIN R, et al. Development of a Geodetic Component for the US West Coast Earthquake Early Warning System[J]. Seismol Res Lett, 2018, 89(6): 2322-36.
[21] VANCE T C, HUANG T, BUTLER K A. Big data in Earth science: Emerging practice and promise[J]. Science, 2024, 383(6688): eadh9607.
[22] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436-44.
[23] 贾佳, 王夫运, 吴庆举. 机器学习在地震检测与震相识别的应用综述[J]. 地震工程学报, 2019, 41(6): 7.
[24] MOUSAVI S M, BEROZA G C. Deep-learning seismology[J]. Science, 2022, 377(6607): 725-+.
[25] KONG Q K, TRUGMAN D T, ROSS Z E, et al. Machine Learning in Seismology: Turning Data into Insights[J]. Seismol Res Lett, 2019, 90(1): 3-14.
[26] KARASÖZEN E, KARASÖZEN B. Earthquake location methods[J]. Gem Int J Geomathema, 2020, 11(1).
[27] LAHR J. HYPOELLIPSE/Version 2.0: A computer program for determining local earthquake hypocentral parameters, magnitude, and first motion pattern[J]. Open File Report, 1999.
[28] SCHWEITZER J. HYPOSAT - An enhanced routine to locate seismic events[J]. Pure Appl Geophys, 2001, 158(1-2): 277-89.
[29] PAVLIS G L, VERNON F, HARVEY D, et al. The generalized earthquake-location (GENLOC) package: an earthquake-location library[J]. Comput Geosci-Uk, 2004, 30(9-10): 1079-91.
[30] LOMAX A, MICHELINI A, CURTIS A, et al. Earthquake location, direct, global-search methods[J]. Encyclopedia of complexity and systems science, 2009, 5: 2449-73.
[31] MYERS S C, JOHANNESSON G, HANLEY W. A Bayesian hierarchical method for multiple-event seismic location[J]. Geophys J Int, 2007, 171(3): 1049-63.
[32] GRIGOLI F, SCARABELLO L, BÖSE M, et al. Pick- and waveform-based techniques for real-time detection of induced seismicity[J]. Geophys J Int, 2018, 213(2): 868-84.
[33] ZHANG M, ELLSWORTH W L, BEROZA G C. Rapid Earthquake Association and Location[J]. Seismol Res Lett, 2019, 90(6): 2276-84.
[34] SATRIANO C, LOMAX A, ZOLLO A. Real-time evolutionary earthquake location for seismic early warning[J]. B Seismol Soc Am, 2008, 98(3): 1482-94.
[35] PEROL T, GHARBI M, DENOLLE M. Convolutional neural network for earthquake detection and location[J]. Sci Adv, 2018, 4(2).
[36] KRIEGEROWSKI M, PETERSEN G M, VASYURA-BATHKE H, et al. A Deep Convolutional Neural Network for Localization of Clustered Earthquakes Based on Multistation Full Waveforms[J]. Seismol Res Lett, 2019, 90(2): 510-6.
[37] MOUSAVI S M, BEROZA G C. Bayesian-Deep-Learning Estimation of Earthquake Location From Single-Station Observations[J]. Ieee T Geosci Remote, 2020, 58(11): 8211-24.
[38] ZHANG X, ZHANG J, YUAN C C, et al. Locating induced earthquakes with a network of seismic stations in Oklahoma via a deep learning method[J]. Sci Rep-Uk, 2020, 10(1).
[39] ZHANG X, ZHANG M, TIAN X. Real-Time Earthquake Early Warning With Deep Learning: Application to the 2016 M 6.0 Central Apennines, Italy Earthquake[J]. Geophys Res Lett, 2021, 48(5).
[40] MÜNCHMEYER J, BINDI D, LESER U, et al. Earthquake magnitude and location estimation from real time seismic waveforms with a transformer network[J]. Geophys J Int, 2021, 226(2): 1086-104.
[41] ZHANG X T, REICHARD-FLYNN W, ZHANG M, et al. Spatiotemporal Graph Convolutional Networks for Earthquake Source Characterization[J]. J Geophys Res-Sol Ea, 2022, 127(11).
[42] ELSAYED H S, SAAD O M, SOLIMAN M S, et al. EQConvMixer: A Deep Learning Approach for Earthquake Location From Single-Station Waveforms[J]. Ieee Geosci Remote S, 2023, 20.
[43] COSTANTINO G, GIFFARD-ROISIN S, MARSAN D, et al. Seismic Source Characterization From GNSS Data Using Deep Learning[J]. J Geophys Res-Sol Ea, 2023, 128(4).
[44] 金星, 张红才, 李军, 等. 地震预警震级确定方法研究[J]. 地震学报, 2012, 34(5): 18.
[45] 李山有. 走近地震预警[J]. 防灾博览, 2018, (2): 10.
[46] NAKAMURA Y. On the urgent earthquake detection and alarm system (UrEDAS); proceedings of the Proc of the 9th World Conference on Earthquake Engineering, F, 1988[C].
[47] ALLEN R M, KANAMORI H. The potential for earthquake early warning in southern California[J]. Science, 2003, 300(5620): 786-9.
[48] KANAMORI H. Real-time seismology and earthquake damage mitigation[J]. Annu Rev Earth Pl Sc, 2005, 33: 195-214.
[49] ZIV A. New frequency-based real-time magnitude proxy for earthquake early warning[J]. Geophys Res Lett, 2014, 41(20): 7035-40.
[50] ATEFI S, HEIDARI R, MIRZAEI N, et al. A frequency-based parameter for rapid estimation of magnitude[J]. Journal of Theoretical and Applied Physics, 2017, 11(4): 319-24.
[51] WANG Y W, LI X J, LI L, et al. New Magnitude Proxy for Earthquake Early Warning Based on Initial Time Series and Frequency[J]. Seismol Res Lett, 2022, 93(1): 216-25.
[52] WU Y M, ZHAO L. Magnitude estimation using the first three seconds P‐wave amplitude in earthquake early warning[J]. Geophys Res Lett, 2006, 33(16).
[53] CHEN D-Y, LIN T-L, WU Y-M, et al. Testing a P-wave earthquake early warning system by simulating the 1999 Chi-Chi, Taiwan, Mw 7.6 earthquake[J]. Seismol Res Lett, 2012, 83(1): 103-8.
[54] 张红才. 地震预警系统关键技术研究[D], 2013.
[55] NODA S, ELLSWORTH W L. Scaling relation between earthquake magnitude and the departure time from wave similar growth[J]. Geophys Res Lett, 2016, 43(17): 9053-60.
[56] CROWELL B W, SCHMIDT D A, BODIN P, et al. Demonstration of the Cascadia G-FAST Geodetic Earthquake Early Warning System for the Nisqually, Washington, Earthquake[J]. Seismol Res Lett, 2016, 87(4): 930-43.
[57] MELGAR D, CROWELL B W, GENG J H, et al. Earthquake magnitude calculation without saturation from the scaling of peak ground displacement[J]. Geophys Res Lett, 2015, 42(13): 5197-205.
[58] RUHL C J, MELGAR D, GENG J H, et al. A Global Database of Strong-Motion Displacement GNSS Recordings and an Example Application to PGD Scaling[J]. Seismol Res Lett, 2019, 90(1): 271-9.
[59] WILLIAMSON A L, MELGAR D, CROWELL B W, et al. Toward Near-Field Tsunami Forecasting Along the Cascadia Subduction Zone Using Rapid GNSS Source Models[J]. J Geophys Res-Sol Ea, 2020, 125(8).
[60] FANG R X, ZHENG J W, GENG J H, et al. Earthquake Magnitude Scaling Using Peak Ground Velocity Derived from High-Rate GNSS Observations[J]. Seismol Res Lett, 2021, 92(1): 227-37.
[61] LOMAX A, MICHELINI A, JOZINOVIC D. An Investigation of Rapid Earthquake Characterization Using Single-Station Waveforms and a Convolutional Neural Network[J]. Seismol Res Lett, 2019, 90(2): 517-29.
[62] MOUSAVI S M, BEROZA G C. A Machine-Learning Approach for Earthquake Magnitude Estimation[J]. Geophys Res Lett, 2020, 47(1).
[63] VAN DEN ENDE M P A, AMPUERO J P. Automated Seismic Source Characterization Using Deep Graph Neural Networks[J]. Geophys Res Lett, 2020, 47(17).
[64] 林彬华, 金星, 康兰池, 等. 基于卷积神经网络的地震震级测定研究[J]. 地球物理学报, 2021, 64(10): 3600-11.
[65] SONG J D, ZHU J B, LI S Y. MEANet: Magnitude estimation via physics-based features time series, an attention mechanism, and neural networks[J]. Geophysics, 2023, 88(1): V33-V43.
[66] ZHU J B, LI S L, LI S Y, et al. Rapid earthquake magnitude estimation combining a neural network and transfer learning in China: Application to the 2022 Lushan M6.1 earthquake[J]. Front Phys-Lausanne, 2023, 11.
[67] WESNOUSKY S G. The Gutenberg-Richter or Characteristic Earthquake Distribution, Which Is It[J]. B Seismol Soc Am, 1994, 84(6): 1940-59.
[68] 冯锐. 趣味地震学(13):震级,向能量标度靠拢[Z]. 地震科学进展. 2020: 40-8.10.3969/j.issn.2096-7780.2020.01.005
[69] LIN J T, MELGAR D, THOMAS A M, et al. Early Warning for Great Earthquakes From Characterization of Crustal Deformation Patterns With Deep Learning[J]. J Geophys Res-Sol Ea, 2021, 126(10).
[70] ZAHRADNÍK J, GALLOVIC F. Toward understanding slip inversion uncertainty and artifacts[J]. J Geophys Res-Sol Ea, 2010, 115.
[71] TRIFUNAC M D. A three-dimensional dislocation model for the San Fernando, California, earthquake of February 9, 1971[J]. Bullseismsocam, 1974, 64(1): 149-72.
[72] OLSON A H, APSEL R J. Finite Faults and Inverse-Theory with Applications to the 1979 Imperial-Valley Earthquake[J]. B Seismol Soc Am, 1982, 72(6): 1969-2001.
[73] HARTZELL S H, HEATON T H. Inversion of Strong Ground Motion and Teleseismic Waveform Data for the Fault Rupture History of the 1979 Imperial-Valley, California, Earthquake[J]. B Seismol Soc Am, 1983, 73(6): 1553-83.
[74] FREYMUELLER J, KING N, SEGALL P. The co-seismic slip distribution of the Landers earthquake[J]. B Seismol Soc Am, 1994, 84(3): 646-59.
[75] YUE H, ZHANG Y, GE Z X, et al. Resolving rupture processes of great earthquakes: Reviews and perspective from fast response to joint inversion[J]. Sci China Earth Sci, 2020, 63(4): 492-511.
[76] COLOMBELLI S, ALLEN R M, ZOLLO A. Application of real-time GPS to earthquake early warning in subduction and strike-slip environments[J]. J Geophys Res-Sol Ea, 2013, 118(7): 3448-61.
[77] ZHANG Y, WANG R J, ZSCHAU J, et al. Automatic imaging of earthquake rupture processes by iterative deconvolution and stacking of high-rate GPS and strong motion seismograms[J]. J Geophys Res-Sol Ea, 2014, 119(7): 5633-50.
[78] KAWAMOTO S, OHTA Y, HIYAMA Y, et al. REGARD: A new GNSS-based real-time finite fault modeling system for GEONET[J]. J Geophys Res-Sol Ea, 2017, 122(2): 1324-49.
[79] CHEN K J, LIU Z, SONG Y T. Automated GNSS and Teleseismic Earthquake Inversion (AutoQuake Inversion) for Tsunami Early Warning: Retrospective and Real-Time Results[J]. Pure Appl Geophys, 2020, 177(3): 1403-23.
[80] LIN J-T. Machine Learning in Large and Small Earthquakes: from Rapid Large Earthquake Characterization to Slow Fault Zone Processes[D]; University of Oregon, 2022.
[81] HAYES G P, MOORE G L, PORTNER D E, et al. Slab2, a comprehensive subduction zone geometry model[J]. Science, 2018, 362(6410): 58-61.
[82] MELGAR D, LEVEQUE R J, DREGER D S, et al. Kinematic rupture scenarios and synthetic displacement data: An example application to the Cascadia subduction zone[J]. J Geophys Res-Sol Ea, 2016, 121(9): 6658-74.
[83] GRAVES R W, PITARKA A. Broadband Ground-Motion Simulation Using a Hybrid Approach[J]. B Seismol Soc Am, 2010, 100(5a): 2095-123.
[84] MAI P M, BEROZA G C. Source scaling properties from finite-fault-rupture models[J]. B Seismol Soc Am, 2000, 90(3): 604-15.
[85] LEVEQUE R J, WAAGAN K, GONZALEZ F I, et al. Generating Random Earthquake Events for Probabilistic Tsunami Hazard Assessment[J]. Pure Appl Geophys, 2016, 173(12): 3671-92.
[86] GREGOR N J, SILVA W J, WONG I G, et al. Ground-motion attenuation relationships for Cascadia subduction zone megathrust earthquakes based on a stochastic finite-fault model[J]. B Seismol Soc Am, 2002, 92(5): 1923-32.
[87] SOMERVILLE P, IRIKURA K, GRAVES R, et al. Characterizing Crustal Earthquake Slip Models for the Prediction of Strong Ground Motion[J]. Seismol Res Lett, 1999.
[88] MENA B, MAI P M, OLSEN K B, et al. Hybrid Broadband Ground-Motion Simulation Using Scattering Green's Functions: Application to Large-Magnitude Events[J]. B Seismol Soc Am, 2010, 100(5a): 2143-62.
[89] TINTI E, FUKUYAMA E, PIATANESI A, et al. A kinematic source-time function compatible with earthquake dynamics[J]. B Seismol Soc Am, 2005, 95(4): 1211-23.
[90] LIU P C, ARCHULETA R J, HARTZELL S H. Prediction of broadband ground-motion time histories: Hybrid low/high-frequency method with correlated random source parameters[J]. B Seismol Soc Am, 2006, 96(6): 2118-30.
[91] ZHU L P, RIVERA L A. A note on the dynamic and static displacements from a point source in multilayered media[J]. Geophys J Int, 2002, 148(3): 619-27.
[92] MCCULLOCH W S, PITTS W. A Logical Calculus of the Ideas Immanent in Nervous Activity[J]. biol math biophys, 1943.
[93] ROSENBLATT F. The Perceptron - a Probabilistic Model for Information-Storage and Organization in the Brain[J]. Psychol Rev, 1958, 65(6): 386-408.
[94] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning Representations by Back-Propagating Errors[J]. Nature, 1986, 323(6088): 533-6.
[95] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. P Ieee, 1998, 86(11): 2278-324.
[96] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Comput, 1997, 9(8): 1735-80.
[97] VASWANI A, SHAZEER N, PARMAR N, et al. Attention Is All You Need[J]. Advances in Neural Information Processing Systems 30 (Nips 2017), 2017, 30.
[98] 潘家伟, 白明坤, 李超, 等. 2021年5月22日青海玛多M_S7.4地震地表破裂带及发震构造[J]. 地质学报, 2021, 95(06): 1655-70.
[99] 邓起东, 张培震, 冉勇康, 等. 中国活动构造与地震活动[J]. 地学前缘, 2003, (S1): 66-73.
[100] BLASER L, KRÜGER F, OHRNBERGER M, et al. Scaling Relations of Earthquake Source Parameter Estimates with Special Focus on Subduction Environment[J]. B Seismol Soc Am, 2010, 100(6): 2914-26.
[101] HANKS T C, KANAMORI H. A moment magnitude scale[J]. Journal of Geophysical Research: Solid Earth, 1979, 84(B5): 2348-50.
[102] CHENG J, RONG Y F, MAGISTRALE H, et al. Earthquake Rupture Scaling Relations for Mainland China[J]. Seismol Res Lett, 2020, 91(1): 248-61.
[103] ALLEN T I, HAYES G P. Alternative Rupture-Scaling Relationships for Subduction Interface and Other Offshore Environments[J]. B Seismol Soc Am, 2017, 107(3): 1240-53.
[104] GENG J H, PAN Y X, LI X T, et al. Noise Characteristics of High-Rate Multi-GNSS for Subdaily Crustal Deformation Monitoring[J]. J Geophys Res-Sol Ea, 2018, 123(2): 1987-2002.
[105] MELGAR D, CROWELL B W, MELBOURNE T I, et al. Noise Characteristics of Operational Real-Time High-Rate GNSS Positions in a Large Aperture Network[J]. J Geophys Res-Sol Ea, 2020, 125(7).
[106] LASKE G, MASTERS G, MA Z, et al. Update on CRUST1. 0—A 1-degree global model of Earth’s crust; proceedings of the Geophysical research abstracts, F, 2013[C].
[107] GIRSHICK R. Fast R-CNN[J]. Ieee I Conf Comp Vis, 2015: 1440-8.
[108] RONNEBERGER O, FISCHER P, BROX T. U-Net: Convolutional Networks for Biomedical Image Segmentation[J]. Lect Notes Comput Sc, 2015, 9351: 234-41.
[109] LECUN Y, BOSER B, DENKER J S, et al. Backpropagation Applied to Handwritten Zip Code Recognition[J]. Neural Comput, 1989, 1(4): 541-51.
[110] IOFFE S, SZEGEDY C. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift[J]. Pr Mach Learn Res, 2015, 37: 448-56.
[111] ZHOU, CHELLAPPA. Computation of optical flow using a neural network; proceedings of the IEEE 1988 international conference on neural networks, F, 1988[C]. IEEE.
[112] ZEILER M D, FERGUS R. Visualizing and Understanding Convolutional Networks[J]. Computer Vision - Eccv 2014, Pt I, 2014, 8689: 818-33.
[113] SRIVASTAVA N, HINTON G, KRIZHEVSKY A, et al. Dropout: A Simple Way to Prevent Neural Networks from Overfitting[J]. Journal of Machine Learning Research, 2014, 15: 1929-58.
[114] MAAS A L, HANNUN A Y, NG A Y. Rectifier nonlinearities improve neural network acoustic models; proceedings of the Proc icml, F, 2013[C]. Atlanta, GA.
[115] HE K M, ZHANG X Y, REN S Q, et al. Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification[J]. Ieee I Conf Comp Vis, 2015: 1026-34.
[116] GLOROT X, BORDES A, BENGIO Y. Deep Sparse Rectifier Neural Networks[J]. Journal of Machine Learning Research, 2011, 15: 315-23.
[117] ABADI M, BARHAM P, CHEN J M, et al. TensorFlow: A system for large-scale machine learning[J]. Proceedings of Osdi'16: 12th Usenix Symposium on Operating Systems Design and Implementation, 2016: 265-83.
[118] KINGMA D, BA J. Adam: A Method for Stochastic Optimization[J]. Computer Science, 2014.
[119] KENDALL A, GAL Y, CIPOLLA R. Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics[J]. 2018 Ieee/Cvf Conference on Computer Vision and Pattern Recognition (Cvpr), 2018: 7482-91.
[120] WANG Z, BOVIK A C, SHEIKH H R, et al. Image quality assessment: From error visibility to structural similarity[J]. Ieee T Image Process, 2004, 13(4): 600-12.
[121] CHEN K J, AVOUAC J P, GENG J H, et al. The 2021 M-w 7.4 Madoi Earthquake: An Archetype Bilateral Slip-Pulse Rupture Arrested at a Splay Fault[J]. Geophys Res Lett, 2022, 49(2).
[122] LYU M, CHEN K J, XUE C H, et al. Overall subshear but locally supershear rupture of the 2021 M-w 7.4 Maduo earthquake from high-rate GNSS waveforms and three-dimensional InSAR deformation[J]. Tectonophysics, 2022, 839.
[123] WANG W L, FANG L H, WU J P, et al. Aftershock sequence relocation of the 2021 M(S)7.4 Maduo Earthquake, Qinghai, China[J]. Sci China Earth Sci, 2021, 64(8): 1371-80.
[124] ZHANG X, FENG W P, DU H L, et al. Supershear Rupture During the 2021 M-w 7.4 Maduo, China, Earthquake[J]. Geophys Res Lett, 2022, 49(6).
[125] GUO R M, YANG H F, LI Y, et al. Complex Slip Distribution of the 2021 Mw 7.4 Maduo, China, Earthquake: An Event Occurring on the Slowly Slipping Fault[J]. Seismol Res Lett, 2022, 93(2a): 653-65.
[126] YUE H, SHEN Z-K, ZHAO Z, et al. Rupture process of the 2021 M7.4 Maduo earthquake and implication for deformation mode of the Songpan-Ganzi terrane in Tibetan Plateau[J]. Proceedings of the National Academy of Sciences, 2022, 119(23): e2116445119.
[127] BILAL M A, JI Y J, WANG Y Z, et al. An Early Warning System for Earthquake Prediction from Seismic Data Using Batch Normalized Graph Convolutional Neural Network with Attention Mechanism (BNGCNNATT)[J]. Sensors-Basel, 2022, 22(17).
[128] ZHANG Z G, ZHANG W, CHEN X F. Three-dimensional curved grid finite-difference modelling for non-planar rupture dynamics[J]. Geophys J Int, 2014, 199(2): 860-79.
[129] CASTELVECCHI D. Can we open the black box of AI?[J]. Nature News, 2016, 538(7623): 20.
[130] RAISSI M, PERDIKARIS P, KARNIADAKIS G E. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations[J]. J Comput Phys, 2019, 378: 686-707.
[131] BAHDANAU D, CHO K, BENGIO Y. Neural Machine Translation by Jointly Learning to Align and Translate[J]. Computer Science, 2014.