quinteros@fias.uni-frankfurt.de
2022-12-21 11:36:11
Dear Colleagues,
Abstract submission for the 2023 SSA Annual Meeting (https://meetings.seismosoc.org/) is open until 17:00 PM, January 11, 2023.
We would like to draw your attention and invite you to submit your abstracts to the session:
Opportunities and Challenges for Machine Learning Applications in Seismology
Owing to the increase in the availability of large amounts of high-quality open source data, in recent years, we observed a successful surge in Machine Learning (ML) applications in Seismology. For instance, ML has largely been adopted in earthquake detection & seismic phase picking, in generating high-resolution earthquake catalogs, in discrimination and classification of seismic events, in earthquake early warning, in seismicity forecasting, in ground motion modeling and simulation, as well as in seismic inversion. Today, traditional ML techniques, such as Convolutional Neural and Long-Short Term Memory networks trained over very large datasets, are successfully employed in operational conditions. Nonetheless, efficient training with small and imbalanced datasets, as well as extrapolation to new data are among the challenges that are still unresolved. On one hand, advanced ML techniques such as attention layers, autoencoders, and transformers provide accurate and faster alternatives. On the other hand, physics-informed learning attempts to solve mathematical problems using neural networks or kernel based approaches, nourished by real world data. Moreover, ML techniques are adopted to improve existing predictive tools, in a non-intrusive way. However, a thorough investigation of those data-driven techniques is demanded, in both existing and new research branches of seismology, before their deployment as operational models.
In this session, we invite contributions that explore the potential of ML for seismology. In particular, we are interested in studies focusing on developing innovative state-of-the-art ML models for seismology and earthquake engineering, ML investigations of new research areas, and works highlighting issues related to methodologies in ML, data quantity, and quality. Furthermore, we welcome contributions on research topics including null hypothesis testing, open databases for collaborative research, architecture and framework, software packages, and development of research capabilities.
Best regards,
Conveners:
Nishtha Srivastava, Frankfurt Institute for Advanced Studies.
Claudia Quinteros Cartaya, Frankfurt Institute for Advanced Studies.
Quentin Brissaud, NORSAR Norwegian Seismic Array.
Filippo Gatti, CentraleSupélec, Université Paris-Saclay.
Florent Aden, GNS Science.
Kiran Thingbaijam, GNS Science.
Abstract submission for the 2023 SSA Annual Meeting (https://meetings.seismosoc.org/) is open until 17:00 PM, January 11, 2023.
We would like to draw your attention and invite you to submit your abstracts to the session:
Opportunities and Challenges for Machine Learning Applications in Seismology
Owing to the increase in the availability of large amounts of high-quality open source data, in recent years, we observed a successful surge in Machine Learning (ML) applications in Seismology. For instance, ML has largely been adopted in earthquake detection & seismic phase picking, in generating high-resolution earthquake catalogs, in discrimination and classification of seismic events, in earthquake early warning, in seismicity forecasting, in ground motion modeling and simulation, as well as in seismic inversion. Today, traditional ML techniques, such as Convolutional Neural and Long-Short Term Memory networks trained over very large datasets, are successfully employed in operational conditions. Nonetheless, efficient training with small and imbalanced datasets, as well as extrapolation to new data are among the challenges that are still unresolved. On one hand, advanced ML techniques such as attention layers, autoencoders, and transformers provide accurate and faster alternatives. On the other hand, physics-informed learning attempts to solve mathematical problems using neural networks or kernel based approaches, nourished by real world data. Moreover, ML techniques are adopted to improve existing predictive tools, in a non-intrusive way. However, a thorough investigation of those data-driven techniques is demanded, in both existing and new research branches of seismology, before their deployment as operational models.
In this session, we invite contributions that explore the potential of ML for seismology. In particular, we are interested in studies focusing on developing innovative state-of-the-art ML models for seismology and earthquake engineering, ML investigations of new research areas, and works highlighting issues related to methodologies in ML, data quantity, and quality. Furthermore, we welcome contributions on research topics including null hypothesis testing, open databases for collaborative research, architecture and framework, software packages, and development of research capabilities.
Best regards,
Conveners:
Nishtha Srivastava, Frankfurt Institute for Advanced Studies.
Claudia Quinteros Cartaya, Frankfurt Institute for Advanced Studies.
Quentin Brissaud, NORSAR Norwegian Seismic Array.
Filippo Gatti, CentraleSupélec, Université Paris-Saclay.
Florent Aden, GNS Science.
Kiran Thingbaijam, GNS Science.
