Primary Convener:
Maeva Pourpoint, Air Force Research Laboratory Albuquerque

Convener:
Ian McBrearty, Stanford University
Jesse Williams, GTC Analytics
Weiqiang Zhu, University of California, Berkeley

Chair:
Ian McBrearty, Stanford University
Jesse Williams, GTC Analytics
Weiqiang Zhu, University of California, Berkeley

Recent advances in Artificial Intelligence/Machine Learning (AI/ML) have transformed the field of seismology and particularly seismic monitoring by decreasing detection thresholds and improving the accuracy and efficiency of source characteristics estimation. However, there remains critical need for systematic uncertainty quantification (UQ) of AI/ML models to assess their reliability and build confidence in their predictions. This session focuses on addressing some of the challenges inherent in estimating uncertainties in AI/ML models and aims to foster conversations within the seismic community. We welcome contributions demonstrating recent progress in UQ approaches and their integration into AI/ML models for seismic monitoring. Topics of interest include the incorporation of UQ techniques such as Monte-Carlo Dropout, Deep Ensembles, Bayesian inferences in seismic signal detection, phase association, and event location and characterization. Efforts that evaluate the strengths/limitations of different UQ techniques and the generalizability of uncertainty estimates in the face of domain shift are of particular interest.

Index Terms
0555 Neural networks, fuzzy logic, machine learning
3275 Uncertainty quantification
7219 Seismic monitoring and test-ban treaty verification
7290 Computational seismology

Suggested Itineraries:
Machine Learning and AI

Neighborhoods:
2. Earth Interior