The article introduces a new framework called H2G2-Net, which uses multi-modal physiological signals to predict cognitive states. The authors argue that existing graph neural networks (GNNs) struggle with this type of data due to its hierarchical structure and the need for predefined graph structures. The proposed H2G2-Net overcomes these limitations by automatically learning a graph structure without domain knowledge. The researchers validate their method using the CogPilot dataset, demonstrating improved prediction accuracy over existing GNNs.
Publication date: 8 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.02905
