Copyright © 2024 R. Ortega, C. Meneses-Ponce & J.D. Castro Morales
EQNeMix is a mixed architecture that combines two widely-used neural networks in seismology: ConvNetQuake (Perol et al., 2018) and EQTransformer (Mousavi et al., 2020). Our algorithm employs a Gaussian Mixture Model to perform a Bayesian Inference using the outputs generated by both neural networks. The ultimate outcome is a probabilistic location pinpointed using just a single seismic station.
-
An integral facet of its versatile design is the algorithm's adaptability, as it is not confined to a single travel-time algorithm. It accommodates a spectrum of options ranging from simpler to more intricate travel-time methods. Furthermore, various sampling techniques such as variational inference, Hamiltonian sampling, among others, can be seamlessly integrated.
-
This algorithm is applicable not only to individual seismic stations but can also be extended to entire seismic networks.