PF-TKRR

Intro

This is the public repository for the paper: Efficient Patient-Finetuned Seizure Detection with Tensor Kernel Ridge Regression.

Required packages

Standard packages

The required python packages can be found in requirements.txt. It is recommended to first create a conda environment to install the packages and required python version. This can be done with the environment.yml file, which also installs the conda packages. Note that the remaining packages still need to be installed with pip.

Additional packages

For pre-processing the seizure_data_processing package is needed: https://github.com/sderooij/seizure_data_processing/releases/tag/v0.0.1

In addition to that also the (self-created) TensorLibrary package is required, as it contains the CP-KRR model: https://github.com/sderooij/tensorlibrary/releases/tag/v0.0.1.

Usage

Before running the scripts and notebooks please modify the src/config.py file. The order in which the scripts and notebooks need to be run is as follows:

1. find_tle_patients.ipynb
2. feature_exstraction.py
3. create_set.ipynb
4. preprocess.py
5. classification.py (or run classify.sh)
6. fine_tune.py
7. post_process.py
8. visualize_results.py
9. convergence.py
10. convergence_auc_plot.ipynb

Explanation on their usage is (in principle) provided in the respective files. Note that there may be issues with import due to different (working) paths

Questions

For any questions you can contact: [email protected]