Graph drawing aims to produce readable visualizations that reveal a graph’s structure through the placement of its nodes and edges.

This topic investigates how graph neural networks or graph transformers can learn to generate layouts, either from existing drawings or by directly optimizing drawing criteria. Possible objectives include reducing edge crossings and node overlap, preserving graph distances, or balancing several competing criteria. The learned methods would be compared with established layout algorithms in terms of drawing quality, runtime, and generalization to unfamiliar graph structures and sizes.

Starting literature:

• DeepGD (2021)
• Graph Neural Networks for Graph Drawing