Neural networks are effective for picking seismic phase arrival times, but previous models often used arbitrary hyperparameters, limiting performance. This study tests a hyperparameter optimization scheme across multiple regions and shows that PhaseNet can be simplified and improved with better hyperparameter choices. Optimized models perform consistently well, even for small events, benefiting applications from microseismicity studies to explosion monitoring.
-
The Value of Hyperparameter Optimization in Phase-Picking Neural Networks
Yongsoo Park and David R. Shelly
Penerbit :
The Seismic Record
Tahun :
2024
Jurnal
Earthquake Geofisika
-
No Scan-
-
No Klasifikasi-
-
ISBN-
-
ISSN-
-
No Registrasi-
-
Lokasi Terbit-
-
Jumlah Hal9
-
Label-
-
Versi DigitalYA
-
Versi FisikTIDAK
-
Lokasi Rak Buku Fisik//
-
Jumlah Exemplar Fisik Tersedia-
