🔥 [ICML 2024] Learning High-Order Relationships of Brain Regions

This is the official implementation of ICML 2024 paper Learning High-Order Relationships of Brain Regions.

Installation

Make sure you have git-lfs installed in order to clone the preprocessed dataset. Altenatively, you can download from Google Drive. After cloning the repo, please check the provided environment.yml to install the conda environment.

conda env create -f environment.yml

Usage

If you want to integrate HyBRiD into your own project

Copy the folder src/hybrid to your local storage, and use it by from hybrid import HyBRiD. Check the docstring and type hints in the file for more details.

Note: The repo is implemented in python3.10 and I use the new typing convention (e.g. list[int] instead of List[int]) so it is not backward compatible. However, adapting it to a lower version is always straightforward.

If you want to run our experiment

Make sure you follow the guidance in the Installation section and run the following command

python main.py -c config/hybrid-fiq.yaml

This will train the model and report the metrics on the ABIDE FIQ task.

TODO

• Data preprocessing scripts
• Baselines comparison (including Brain Network Transformer)

Note: Please urge me if you need any of the above items.