Structure-Preserving Graph Kernel for Brain Network Classification

Brain network analysis is of great importance in clinical diagnosis and treatments. In this paper, we present a novel graph-based kernel learning approach for brain network classification. Specifically, we demonstrate how to exploit the natural graph structure of brain networks to encode prior knowledge in the kernel using the tensor product operator. For each brain network, we first proposed to apply sparse matrix factorization with a symmetric constraint to extract tensor product based approximation. We then used them to derive a structure-persevering symmetric graph kernel to be fed into the support vector machine (SVM). Quantitative evaluations on challenging EEG-based emotion recognition tasks with respect to different frequency bands demonstrate the superior performance of our proposed method, compared with the state-of-the-art traditional and deep learning methods. Together, results show that relevant EEG signals are primarily encoded in the alpha and theta bands during the emotion regulation task, which is consistent with previous findings.