컨텐츠 시작
학술대회/행사
초록검색
| 제출번호(No.) | 0269 |
|---|---|
| 분류(Section) | Special Session |
| 분과(Session) | (SS-17) Scientific Computing and Machine Learning (SS-17) |
| 발표시간(Time) | 19th-B-13:30 -- 14:00 |
| 영문제목 (Title(Eng.)) |
Incorporating power spectral density into seismic wave generative model |
| 저자(Author(s)) |
Keunsuk Cho1, Jeongun Ha1, Jihun Lim1, Jongwon Han1, Seongryong Kim1, Donghun Lee1 Korea University1 |
| 초록본문(Abstract) | Recent advances in deep generative models resulted in general purpose tools for laypeople such as ChatGPT and Midjourney. In this talk, we present a waveform generative model specialized for professional-level tasks in earthquake sensing: a data-driven method to generate site-specific seismic wave measurement signals when no earthquake events are recorded. To achieve seismological realism in generated samples, we deploy unconditional WGAN-GP (Wasserstein GAN with Gradient Penalty) training framework augmented with PPSD (Probabilistic Power Spectral Density) loss function. * This work is supported by the Korea Meteorological Administration Research and Development Program under Grant KM12021-01112. |
| 분류기호 (MSC number(s)) |
68T05 |
| 키워드(Keyword(s)) | Generative adversarial network, probabilistic power spectral density, seismic waveform |
| 강연 형태 (Language of Session (Talk)) |
English |