/Pretraining

Primary LanguageJupyter Notebook

Pretraining

Data Location:

Sleep EDF Expanded Dataset (We used Sleep Cassette Data): https://www.physionet.org/content/sleep-edfx/1.0.0/

MDD Dataset: https://figshare.com/articles/dataset/EEG_Data_New/4244171

Python Environment Setup:

Required Python packages are found in Packages folder.

Data Preprocessing Code:

sleep_preprocessing.m - MATLAB script used to preprocess sleep data mdd_preprocessing.py - Python used to preprocessing MDD data

Model Training Code:

sleep_train.py - Python script for training the Model S sleep models that were later used in mdd_pretrained.py

mdd_pretrain.py - Python script for training MDD Models A through D

Performance Analysis Code:

mdd_pretrain_performance_analysis.ipynb - Jupyter Notebook for performing statistical analysis comparing MDD models performance

Explainability Code:

mdd_pretrain_explainability.py - Python script used to generate spatial and spectral explanations for MDD Models A through D

mdd_pretrain_visualize_explainability.ipynb - Jupyter Notebook used to visualize spatial and spectral explainability results

Pretrained Sleep Models:

Pretrained sleep models are found in Sleep_Models folder. They can be loaded using the code in mdd_pretrain.py