Large-Scale Graph Neural Networks

With the recent rise in machine learning, neural network is also applied to graph data, which has a more versatile data structure, has gained significant attention. This project is working on developing technologies to process graph deep learning with higher accuracy and speed, as well as creating benchmarks to evaluate the performance of various graph neural networks.
The project focuses primarily on three key pillars.
The first pillar involves developing integrated graph deep learning techniques that complement the advantages and address the limitations of multiple graph deep learning methods. This includes working on techniques for weighting multiple Graph Neural Networks (GNNs) with consideration for robustness.
The second pillar is dedicated to researching graph generators [1] that control the characteristics of real-world graphs and creating benchmarks using artificial graph generators to elucidate the strengths and weaknesses of various graph neural networks [2].
This includes developing techniques to extract graph properties (such as homophily/heterophily and core/border) and class distributions from input graphs, and generating various graphs based on these distributions.
The final pillar focuses on enhancing the accuracy and speed of graph neural networks [2, 4, 5] for a variety of graph data.

Back to the research page