Graph variational auto-encoder for deriving EEG-based graph emb e dding

Graph embedding is an effective method for deriving low-dimensional representations of graph data. The power of graph deep learning methods to characterize electroencephalogram (EEG) graph embedding is still in question. We designed a novel graph variational auto-encoder (GVAE) method to extract nodal features of brain functional connections. A new decoder model for the GVAEs network is proposed, which considers the node neighborhood of the reconstructed adjacency matrix. The GVAE is applied and tested on 3 biometric databases which contain 64 to 9 channels' EEG recordings. For all datasets, promising results with more than 95% accuracy and considerably low computational cost are achieved compared to state-of-the-art user identification methods. The proposed GVAE is robust to a limited number of nodes and stable to users' task performance. Moreover, we developed a traditional variational auto-encoder to demonstrate that more accurate features can be obtained when observing EEG-based brain connectivity from a graph perspective. (c) 2021 Elsevier Ltd. All rights reserved.

Automated summary

Graph embedding is a powerful method for deriving low-dimensional representations of graph data. The novel graph variational auto-encoder (GVAE) method was designed to extract nodal features of brain functional connections, achieving promising results with over 95% accuracy on 3 biometric databases.