Event‐Based Training in Label‐Limited Regimes

Abstract

Semi-supervised learning (SSL) offers a way to use the structural characteristics of an entire data set, regardless of label availability, for specific task learning. In this paper, structure refers to the salient waveform characteristics imparted by underlying physical processes that persist through the nonlinear transforms inherent to deep neural networks. Although an ideal feature set would exhibit structure dominated by target characteristics, in practice features can inherit dependencies which are not causally related to the prediction task. For seismic event data sets, where segmented input data spans the duration of the transient energy from a specific and discrete seismogenic process, inherited structure can come from processes including travel path, geologic site conditions, source properties, nonstationary seismic background noise, and various data artifacts. Although persistent seismic monitoring networks generate data-rich event catalogs that enable the development of deep neural network (DNN) models capable of making accurate decisions about source attributes, they do so with a complex set of features only some of which may relate to the seismogenic processes that matter for predictive modeling tasks. Yet despite imperfect feature learning, performance on seismic processing tasks using stable network geometries can be superior to traditional methods (Linville et al., 2019;Ross et al., 2019;Tibi et al., 2019) and provide capabilities that scale as data volumes increase (Nguyen et al., 2019). Therefore, even when models generalize poorly to new regions, they can add consequential value to the specific event processing pipelines they were developed for. While some authors have achieved meaningful generalization on seismic processing tasks beyond the regional networks their models were trained on (e.g., Ross et al., 2018), this work is focused on scenarios in the seismic domain where we can expect that we have collected data that imperfectly represents the set of possible states our system is capable of expressing, and we have, or can obtain at some cost, ground truth for a limited number of the observations that we have collected.

Global Nuclear Explosion Discrimination Using a Convolutional Neural Network

“…(2019) for earthquake phase detection, in which the authors did not have such issues while using a similar approach but with a training set that was an order of magnitude larger than ours. Without substantially more nuclear event waveforms this will remain a limitation of our work, however, event‐based training could prove useful to combat these limitations (L. M. Linville, 2022). The issue is still quite small, affecting only 2% of the signals.…”

Machine learning applied to apatite compositions for determining mineralization potential

“…Previous research has suggested that magmas characterized by high water content, high Sr/Y ratio, and high oxidation state play a vital role in the genesis of porphyry Cu deposits (Lu et al, 2015;Richards, 2011;2015). Recent investigations have indicated that chlorine and sulfur are crucial components of ore-forming fluids due to their ability to form complexes with ore metals, including Cu, Au, Pb, Zn, Fe, and Mo (Hsu et al, 2019;Xu et al, 2021;2022). These geochemical signatures of magma could be inherited by apatite crystallized from such fertile magmas.…”

“…Over the past few years, there has been an explosion of interest in the applications of ML to solid Earth geoscience (Li et al, 2023). ML has been widely applied in earthquake phase detection and seismicity classification (Cianetti et al, 2021;Linville, 2022), geophysical data processing and image interpretation (Xiao et al, 2021), geophysical inversion (Cai et al, 2022), and multi-physical and multidisciplinary information integration. Given the complexity and diversity of geochemistry data, ML-based classification methods have emerged as a promising approach that outperforms conventional methods, especially in large-scale geological processes, such as in predicting mantle metasomatism worldwide (Qin et al, 2022), revealing source compositions of intraplate basaltic rocks (Guo et al, 2021), identifying primary water concentrations in mantle pyroxene (Chen et al, 2021), determining the quartz-forming environments , and classifying the source rocks of detrital zircons (Zhong et al, 2023a(Zhong et al, , 2023b.…”

Machine Learning Developments and Applications in Solid‐Earth Geosciences: Fad or Future?

“…Seismicity classification is a more challenging step in the earthquake monitoring workflow as fewer labeled data are available, and more uncertainty exists in current manually labeled datasets. Therefore, earthquake seismologists appeal to semi-supervised methods (Linville, 2022) and unsupervised ML methods, which do not require manual labels and are often better generalized. Zhu, McBrearty, et al (2022) proposes a new earthquake phase association algorithm based on a Bayesian Gaussian Mixture model to aggregate picked seismic phases into individual seismic events.…”