AutoResearch / EEG-GAN

Other
19 stars 1 forks source link

Build Autoencoder Visualizer #37

Closed chadcwilliams closed 1 year ago

chadcwilliams commented 1 year ago

Need a method to visualize the AE training losses and reconstruction accuracy.

whyhardt commented 1 year ago

Visualizer was updated; "multichannel and condition"-enabled; restructured; autoencoder-enabled columns are channels; rows are samples; averaging now over conditions -> each row represents one condition; pca and tsne is working now for autoencoder as well. During ae-training samples are saved in state_dict. Shape of samples is (sample, (real, reconstructed), timeseries, channels) -> 2nd dimension added