Arrival Time Uncertainties Created Using SWAG

From the paper A Deep-learning Phase Picker with Calibrated Bayesian-derived Uncertainties for Earthquakes in the Yellowstone Volcanic Region by Alysha Armstrong, Zachary Claerhout, Ben Baker, and Keith D. Koper

Overview

Train and evaluate deep-learining P- and S-phase arrival time pickers using Stochastic Weight Averaging - Gaussian (SWAG; Maddox et al., 2019) and Multiple SWAG (MultiSWAG; Wilson and Izmailov, 2020). Using SWAG, these models produce Bayesian-derived uncertainity estimates for the phase arrival times. The uncertainties are calibrated using the method from Kuleshov et al. (2018). The deep-learning model architecture is from Ross et al. (2018).

Examples of MultiSWAG and calibration applied to two waveforms from the University of Utah Seismograph Stations (UUSS) catalog. The vertical-component waveform (Z) is shown with a histogram of predicted picks from three separate SWAG models (m1, m2, and m3), the UUSS analyst pick (y_act), the predicted pick from MultiSWAG (y_pred), the standard deviation of the MultiSWAG predictions (1 st. dev.) and the calibrated 90% credible interval (C.I.). The x-axis is relative to the analyst pick (0.0 s). (a) A P arrival labeled as high quality (0) by a seismic analyst. (b) A P arrival labeled as low quality (2) by a seismic analyst. The signal-to-noise (SNR) value for the arrival is in the top right corner.

Directories

• swa_gaussian-master: SWAG code, downloaded on May 3, 2022
• swag_modified: Modified SWAG code used in Armstrong et al. (2023) to produce uncertainty estimates on phase arrival time estimates
• uuss: Hyperparameter tuning, training, evaluation, and uncertainty calibration of models presented in Armstrong et al. (2023)

References

• Armstrong, A. D., Z. Claerhout, B. Baker, and K. D. Koper (2023). A deep-learning phase picker with calibrated bayesian-derived uncertainties for earthquakes in the Yellowstone Volcanic Region, Bull. Seismol. Soc. Am. XX, 1–22
• Kuleshov, V., N. Fenner, and S. Ermon (2018). Accurate uncertainties for deep learning using calibrated regression, in Proceedings of the 35th International Conference on Machine Learning Stockholm, Sweden, PMRL.
• Maddox, W., T. Garipov, P. Izmailov, D. Vetrov, and A. G. Wilson (2019). A simple baseline for Bayesian uncertainty in deep learning, Advances in Neural Information Processing Systems 32, 13153–13164.
• Ross, Z. E., M. Meier, and E. Hauksson (2018). P wave arrival picking and first‐motion polarity determination with deep learning, J. Geophys. Res. Solid Earth 123, 5120–5129.
• Wilson, A. G., and P. Izmailov (2020). Bayesian deep learning and a probabilistic perspective of generalization, in Advances in Neural Information Processing Systems, 4697–4708.