AlCheNet

Pathology-informed Graph · Channel-wise Encoder · ChebNet

A graph-based model for EEG analysis combining pathology-informed graph construction, channel-wise encoding, and ChebNet.

Model Overview · Setup · Dataset · Train & Test · Results · Project Retrospective

---

🧠 Model Overview

AlCheNet brings together three components:

Component	Role
Pathology-informed Graph	Incorporates pathology information into graph construction.
Channel-wise Encoder	Encodes EEG channel features.
ChebNet	Learns graph representations for classification.

🛠️ Setup

Ubuntu setup

1. Create a container

Run these commands on the host. Adjust the container name and shared memory size to suit your environment.

docker pull ubuntu:22.04
docker run -itd --gpus=all --shm-size=8g --name <container_name> ubuntu:22.04
docker exec -it <container_name> bash

2. Install system packages and clone the repository

Run the following commands inside the container.

apt-get update
apt-get install -y sudo
sudo apt-get install -y git curl unzip python3 python3-venv

git clone https://github.com/drawcodeboy/AlCheNet.git
cd AlCheNet

3. Create a Python environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

📁 Dataset

Download the Open Nuro Dataset and extract it under data/ from the repository root.

mkdir -p data
curl -L -o ./data/open-nuro-dataset.zip https://www.kaggle.com/api/v1/datasets/download/yosftag/open-nuro-dataset
unzip data/open-nuro-dataset.zip -d data/open-nuro-dataset

The preprocessing script expects the dataset at data/open-nuro-dataset/dataset.

Data processing

Prepare the DTW adjacency matrices and node features (EEG channel features) before training.

python datasets/data2np.py

🚀 Train & Test

Use a configuration name without the .yaml extension. Training configurations are in configs/train/, and test configurations are in configs/test/.

# Train
python train.py --config=<config_name>

# Test
python test.py --config=<config_name>

For AlCheNet:

python train.py --config=alchenet
python test.py --config=alchenet

Before running, review the configuration files:

Setting	Where to configure
GPU device	Set device in both configurations to match your environment; the AlCheNet defaults use cuda:1.
Dataset path	Check the root values in the training and test configurations.
Test checkpoint	Set save_dir and weights_file_name in the test configuration to the checkpoint you want to evaluate.

📊 Results