Graph Neural Networks icon Graph Neural Networks#

Graph Neural Network

This category contains graph-based models that treat EEG electrodes as graph nodes and learn inter-channel relationships dynamically.

Available Models

  • DGCNN — Dynamic Graph Convolutional Neural Network. Treats electrodes as graph nodes and learns the adjacency matrix jointly with all other parameters via back-propagation. Uses Chebyshev spectral graph convolution to extract spatial features from the learned graph Laplacian. Based on Song et al. (2018).