1:55 PM - 2:15 PM
[DX-05] Automatic Detection of First Arrival Time of Seismic Waves with the Fine-tuning: Applying to Observed Data in Hachijojima
It was reported that volcanic activity has been getting active near Hachijojima recently (Kimura et al.,2002). In order to evaluate a risk of eruption, we aim to find underground structure by technique using natural earthquake. Hence, we need many and accurate first arrivals of seismic waves generated by natural earthquakes to improve the accuracy of underground structure analysis. However, the number of earthquakes near Hachijojima is small. Moreover, we identify first arrivals of seismic waves manually. It takes much time to do it. Therefore, we try to increase the number of earthquakes to improve the accuracy of underground structure analysis, to automate the identification of first arrivals of seismic waves and to improve the accuracy of the identification by applying deep learning.
The recent development of the deep learning is leading innovations in various fields. But it is necessary for deep learning to prepare a large-scale dataset. It takes vast time to prepare the dataset. In order to solve the problem, we use the method called fine-tuning. The fine-tuning is a method to retrain a part of a trained model’s parameters with a small amount of data and enables to extract the features of data efficiently.
We applied Generalized Seismic Phase Detection with Deep learning (GPD) (Ross et al., 2018) to waveform data observed in Hachijojima. But there were many false detections. Therefore, we applied fine-tuning to solve the problem.
We evaluated the effects of fine-tuning by comparing the result of fine-tuning with manual detection in terms of the number of detection and the accuracy of detection time. As a result, we could reduce false detections and succeed in highly accurate detection in case of applying to a data includes earthquakes. But we couldn’t reduce false detection in case of applying to a data doesn’t include earthquakes.
The recent development of the deep learning is leading innovations in various fields. But it is necessary for deep learning to prepare a large-scale dataset. It takes vast time to prepare the dataset. In order to solve the problem, we use the method called fine-tuning. The fine-tuning is a method to retrain a part of a trained model’s parameters with a small amount of data and enables to extract the features of data efficiently.
We applied Generalized Seismic Phase Detection with Deep learning (GPD) (Ross et al., 2018) to waveform data observed in Hachijojima. But there were many false detections. Therefore, we applied fine-tuning to solve the problem.
We evaluated the effects of fine-tuning by comparing the result of fine-tuning with manual detection in terms of the number of detection and the accuracy of detection time. As a result, we could reduce false detections and succeed in highly accurate detection in case of applying to a data includes earthquakes. But we couldn’t reduce false detection in case of applying to a data doesn’t include earthquakes.
Abstract password authentication.
The password was sent to the registered participants by e-mail in advance.