2

I am new to graph deep learning and I am trying to generate a graph from a matrix of node features by computing graph edges for all nodes within a given distance. I have found a concise way to do this with torch_geometric.nn.radius_graph() but I would like a similarly concise method of accomplishing the same task with tensorflow tensors.

My code is kind of a mess right now and I would like to change it significantly based on whatever method that is convenient for processing a graph dataset in this form. I am aware that I can write a class of connectivity utility functions but I don't think that is ideal.

playerJX1
  • 209
  • 1
  • 11
  • Hi @playerJX1, Could you please take a look at this similar api available in TF Graphics as [tfg.geometry.convolution.graph_convolution.edge_convolution_template](https://www.tensorflow.org/graphics/api_docs/python/tfg/geometry/convolution/graph_convolution/edge_convolution_template). Thank you. –  Jan 13 '23 at 09:18
  • I am not too sure about this being similar. I don't think I want to do graph convolution for this project. – playerJX1 Jan 14 '23 at 22:22
  • Okay, Here is one more API related to [TF_geometric](https://tf-geometric.readthedocs.io/en/latest/modules/layers.html). –  Jan 15 '23 at 16:27

0 Answers0