Can an NN model plainly learn planar layouts?

TL;DR

This paper investigates whether neural networks can learn to generate planar graph layouts and generalize beyond planarity, showing promising results for some classes but also highlighting limitations in robustness.

Contribution

It demonstrates the potential of neural networks to learn planar graph layouts and compares their performance to traditional methods, revealing strengths and weaknesses.

Findings

Neural networks can outperform traditional techniques for certain planar graph classes.

The model's performance varies with data randomness, indicating sensitivity.

The approach shows promise but has limitations in robustness and generalization.

Abstract

Planar graph drawings tend to be aesthetically pleasing. In this poster we explore a Neural Network's capability of learning various planar graph classes. Additionally, we also investigate the effectiveness of the model in generalizing beyond planarity. We find that the model can outperform conventional techniques for certain graph classes. The model, however, appears to be more susceptible to randomness in the data, and seems to be less robust than expected.