Most models that convert a graph into a 2D map of points use force based systems to find a layout. This works wonderfully however there are some odd situations created by local minima: planar graphs don’t become planar due to the original
Existing solutions
- NetworkX provides a comprehensive list of Graph Layout mechanisms
- Historically I found the best results using the Kamada Kawai algorithm.
next-steps Is it possible to bypass local minima in force based layout algorithms by using the force based approach on more than 2 dimensions, and iterate the algorithm along a slow collapse of the extra dimensions? Worth studying :)