HyperGCN: Interpreting Hyperscanning EEG Signal Common Multi-Tasks Classification Based on Graph Convolutional Network

Welcome to HyperGCN – a cutting-edge algorithm that blends the power of Graph Convolutional Networks (GCN) with multi-task learning to unlock the potential of hyperscanning datasets. Designed for the next generation of brain-computer interface (BCI) research, HyperGCN enables simultaneous analysis of multiple brain states, making it possible to extract meaningful relationships across multiple subjects during complex cognitive tasks.

HyperGCN empowers researchers to analyze individual brain regions and explore their interdependencies in a shared space, creating an intricate neural network of multi-subject brain activity. Whether you're working with fMRI, EEG, or other types of neural data, HyperGCN is your key to deeper insights and more powerful results.

🚀 Features

• Multi-Task Classification: Solve multiple prediction tasks at once using a shared graph-based learning model.
• Hyperscanning-Friendly: Designed specifically to process brain data from multiple subjects performing tasks concurrently.
• Graph Convolutional Networks: Harnesses GCN's ability to capture spatial relationships between brain regions and their interactions across subjects.
• Scalable & Modular: Easily customizable to suit different types of datasets and brain research applications.

Table of Contents

• Installation
• Usage
• Model Architecture
• Datasets
• Results

Installation

Create an environment using requirement.txt (All packages mentioned there)

Usage

Once the dependencies are installed then, just run the main.py

Model Architecture

HyperGCN utilizes Graph Convolutional Networks...

Datasets

The datasets for common Multi Tasks and code for Hyper-CSP are available on Github.

Results

Accuracy: 92.86%