Graph Convolutional Neural Network Learning （2017 – 2019）
- Proposed a novel graph convolutional neural network architecture based on depth-based representation of graph structure deriving from quantum walks.
- Proposed a novel graph convolutional auto-encoder architecture that integrates the global topological structure and local connectivity information within a graph.
Motif-based Graph Representation Learning （2019 – 2021）
- Proposed a theoretical basis for motif-based graph representation learning, which derived novel measurement quantities by mapping the graph motifs into clusters in the thermodynamic system.
- Extended motif-based representations into directed graph field, and validated in anomaly detection for the stock market.
Graph Representation Learning in Computer Vision （2021 – 2023）
- For 2D images: Proposed a position-aware embedding network based on subgraphs for graph matching, which combined relative position information at the node level and high-order structural arrangement information at the subgraph level.
Graph Learning in Brain Network Analysis （2021 – 2024）
- Dynamic Representation Learning
- Proposed a dynamic graph modeling method for brain functional and effective connectivity.
- Proposed a spatio-temporal graph neural network with brain functional and effective connectivity fusion for MCI diagnosis.
- Directed Representation Learning
- Proposed a learnable subdivision method to encode brain networks into multiple latent feature subspaces, which extracted representations of brain networks in discrete subspaces.
- Proposed a heterogeneous brain network modeling and representation method for brain cognitive disorder diagnosis.
- Graph Structure Learning
- Proposed a graph structure learner that adaptively characterizes general brain connectivity networks for various brain disorders.
- Proposed a fundamental brain network analysis framework based on graph contrastive learning for multi-dataset brain disorder diagnosis.
