RGNN - Regularized Graph Neural Network

Bartolomeo Caruso, Gabriele Costanzo, Giuseppe Fallica

---

This repository is an extention of the Python implementation of RGNN.

It's the result of a final assignment for the Cognitive Computing 2020/21 course at UniCT - University of Catania.

We refactored some code from other contributors, such as PieraRiccio's implementation of the EmotionDL regularizer, and added all the missing pieces necessary for a minimal execution of the model on the SEED-IV EEG dataset.

We provide the code both as a Python module and as a Colab/Jupyter notebook, and we provide additional files to be integrated with the SEED-IV dataset.

Note:

In order to execute the code correctly, it may be necessary installing torch_geometric and all of its dependencies. Here is the code we used to import it in Google Colab:

TORCH_version = torch.__version__
TORCH = TORCH_version.split('+')[0]

CUDA_version = torch.version.cuda
CUDA =  'cu' + CUDA_version.replace('.', '')

!pip install torch-scatter     -f https://pytorch-geometric.com/whl/torch-{TORCH}+{CUDA}.html
!pip install torch-sparse      -f https://pytorch-geometric.com/whl/torch-{TORCH}+{CUDA}.html
!pip install torch-cluster     -f https://pytorch-geometric.com/whl/torch-{TORCH}+{CUDA}.html
!pip install torch-spline-conv -f https://pytorch-geometric.com/whl/torch-{TORCH}+{CUDA}.html
!pip install torch-geometric 

Following, is the original README from zhongpeixiang's repo:

---

This repo illustrates the RGNN model implementation in the paper EEG-Based Emotion Recognition Using Regularized Graph Neural Networks. The model is based on torch geometric v1.2.1

The EmotionDL regularizer is easy to implement and thus not included in the repo. More details can be found in the paper.

If you find the paper or this repo useful, please cite

@article{zhong2020eeg,
  title={EEG-Based Emotion Recognition Using Regularized Graph Neural Networks},
  author={Zhong, Peixiang and Wang, Di and Miao, Chunyan},
  journal={IEEE Transactions on Affective Computing},
  year={2020},
  publisher={IEEE}
}