This repository contains all of the source code used to train, test, evaluate and compare all of the deep learning models that make up the work from Thinker Invariance: Enabling Deep Neural Networks for BCI Across More People written by Kostas and Rudzicz (https://doi.org/10.1088/1741-2552/abb7a7).
To reproduce results from each of the datasets, the ./experiments directory contains BASH scripts that should be sufficient. Similarly, the ./analysis directory should provide all the tools needed to produce plots and figures.
If you would like to extend this work as is, or run the tests with different hyperparameters or options, run:
python3 main.py --help
This should provide many options for trying something new and or different.
If you would simply like to run the exact same tests with a new dataset, adding a new file to the directory ./datasets and using one of the existing files as a template should be enough to get up and running quickly.
This was originally written, and heavily relies on using:
- Specifically
- python >= 3.5
- mne ~= 20.0
- pytorch >= 1.0
- More generally
- tqdm
- numpy
- matplotlib
- pandas