ngroebner/Autoencoders

Autoencoders for unsupervised learning. Written in Pytorch.

Various autoencoders for seismic analysis

Written in Pytorch.

Library for building autoencoders. Various encoder and decoder layers, latent layers, clustering layers.

To install, clone the repository and type

pip install .

So far have 2D convolutional regular and variational autoencoders done.

To do:

1. Causal convolutional encoders and decoders for spectrograms (2D) and seismograms (1D)
2. Convolutional temporal encoder/decoder.
3. Deep convolutional embedded clustering.
4. Train encoders together, decoders separately based on basin, source, etc to learn generalized encodings of seismic sources. Like universal music translation network / DeepFakes.