Symmetric Variational Autoencoders (SymVAE) disentangle coherent source information from path effects in grouped waveforms. The model enables extraction of high‑resolution source time functions (STFs) and reveals directivity patterns that envelope averaging obscures. It examines the spatial patterns in earthquake data and effectively separates slow-scale source effects from fast-scale path effects. This separation helps us to determine earthquake source time functions more accurately, without being influenced by path effects.

On extracting coherent seismic wavefield using variational symmetric autoencoders
https://arxiv.org/abs/2411.15613
​​​​​​​Learning earthquake sources using symmetric autoencoders, BSSA, 2025
https://doi.org/10.1785/0120250071