GAN Package Main Functions

GAN Package Details

EEG-GAN is a command line interface (CLI) package that allows users to train a Generative Adversarial Network (GAN) on EEG data. Once installed, you can run eeggan functions <function> in your terminal or command prompt alongside their parameters <params>: eeggan <function> <params>.

eeggan has four functions:

  gan_training - This trains a GAN
  autoencoder_training - This trains an autoencoder
  visualize - This visualizes components of a trained GAN, such as the training losses
  generate_samples - This generates synthetic samples using the trained GAN

Arguments

Each function can be followed by function parameters. Parameters are structured in that:
  Boolean arguments are passed as their argument name (e.g., ddp):
    eeggan gan_training ddp
  While other arguments are passed with an equals sign =:
    eeggan gan_training data=data/eeg_training_data.csv
   Arguments are separated by a space:
     eeggan gan_training ddp data=data/eeg_training_data.csv

Parameters

You can use the help argument to see a list of possible parameters with a brief description:
  eeggan gan_training help
  eeggan autoencoder_training help
  eeggan visualize help
  eeggan generate_samples help

You can also see these parameters the Parameters page.