Alljoined-1.6M

Setup

Before setting up the environment, ensure you have mamba installed.

Create the Environment

Run the following command to create the environment using mamba:

mamba env create -f environment.yml

This will create a new environment with all required dependencies as specified in environment.yml.

Activate the Environment

After the installation completes, activate the environment:

mamba activate aj-preprocessing 

Updating the Enviroment

If new dependencies are added to environment.yml, update your environment with:

mamba env update -f environment.yml

Preprocessing the data

We can run the preprocessing script from the root directory as a module:

python -m preprocessing --sub <subject>

By default, it will not show you warnings unless you add the --verbose argument.

The preprocessed data is then saved to preprocessed_eeg_{test,train}_flat.npy.

Epoching

In this stage, we concatenate all the raw edf files for each block, filter out oddball trials, and then throw the data into mne.Epoch to get epoched data. After this stage, we already get a numpy array of desired shape (trials, channels, times).

MVNN

The next stage is whitening the data, we throw the epoched training data, for each image condition, we compute the covariance of the channels for each timestamp, and then take the average. Then we take the average of all the average covariance matrix for each image condition, to obtain this average of average covariance matrix $\Sigma_{cond}$ and then apply $\Sigma_{cond}^{-1/2}$ to every epoched train and test trial.Z