Showing 1–1 of 1 results for author: Marthindale, N

Toward path-invariant embeddings for local distance source characterization

Authors: Lisa Linville, Chengping Chai, Nathan Marthindale, Jacob Smith, Scott Stewart, Asmeret Naugle

Abstract: This work builds on recent advances in foundation models in the language and image domains to explore similar approaches for seismic source characterization. We rely on an architecture called Barlow Twins, borrowed from an understanding of the human visual cortical system and originally envisioned for the image domain and adapt it for learning path invariance in seismic event time series. Our mode… ▽ More This work builds on recent advances in foundation models in the language and image domains to explore similar approaches for seismic source characterization. We rely on an architecture called Barlow Twins, borrowed from an understanding of the human visual cortical system and originally envisioned for the image domain and adapt it for learning path invariance in seismic event time series. Our model improves the performance on event characterization tasks such as source discrimination across catalogs by 10-12% and provides more reliable predictive uncertainty estimates. We suggest that dataset scale and diversity more than architecture may determine aspects of the current ceiling on performance. We leverage decision trees, linear models, and visualization to understanding the dependencies in learned representations. △ Less

Submitted 23 October, 2024; originally announced October 2024.