Hands-on tutorial on deep learning for EEG classification.
In this hands-on tutorial, you will train a convolutional neural network to identify sleep stages from raw EEG signals, and try to improve the classification performance of an existing model. The tutorial will guide you through the different steps of loading publicly available sleep recordings, preparing the data for training, creating a neural network, training it, and analyzing its behaviour. The tutorial is based on the open source packages MNE-Python and pytorch and can be run using Google Colab.
Inspiration for this tutorial was taken from the MNE-Python and braindecode sleep staging examples, as well as the mne-torch repository.
Open this tutorial in Google Colab
To run this tutorial locally, make sure the repo has been cloned on your machine, and that the required packages are installed in a Python environment (3.6 or above). The packages can be installed by running:
pip install -r requirements.txt
The entire tutorial is contained in sleep_staging_physionet.ipynb and can be run with jupyter notebook.
Tested on Ubuntu 20.04.