I find this DOC confusing ij the library

In this documentation. I find this GNN equation wrong. Why x_i is included in aggregation process? It should be there in final update stage only right? Please correct me with resource. Also why is it saying EXOR symbol as differentiable function , it should represent only permutation invariant no? Differentiable is activation function which is there.

Hi @cody

First of all, the documentation is for PyG and here is a forum for DGL.

For your questions:

  1. It depends. You can choose aggregators that include features of node i or not. It’s up to your choice of aggregation implementation.
  2. The differentiability ensures that the whole model is learnable, which means the model can be trained with gradient descent.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.