I'm at the beginning of a project and I really wish to try and on the way learn Graph-Neural-Networks.
The goal is to find the underlying graph structure of data points, some could be more clustered together and some are far. I have labeled data so its a supervised training.
In the example, the pink ones are my signal, the ones I need for my calculation, each node has its own external features and I'm trying to think how to approach this clustering problem?
Should I build an initial graph using some prior knowledge on my data?
I would appreciate if you can also point me to similar work to draw inspiration. Thank you