I am trying to create a Graph NN that will be able to predict the forces in truss elements of a space frame.

The input for the NN will be a graph, where the nodes represent the nodes of the spaceframe. And the output should be the forces in the edges of the supplied graph/frame.

The problem that I am facing is that for the NN to be beneficial I need to encode a lot of data per node:

  • 3 floats for the position of the node,
  • 3 floats for the vector of the force applied to the node,
  • a boolean/int to determine whether a node is a support.

I am not sure how to design my Graph NN to allow for so many parameters per node.

Maybe I should try a different approach?

Any help is greatly appreciated!

  • $\begingroup$ If I remember correctly, you can have a feature vector (or arbitrary size) for each node. Why do you think that 7 features per node are too many? $\endgroup$
    – nbro
    Commented Jun 25, 2020 at 16:30


You must log in to answer this question.