So Graph Neural Networks is about representation learning where initially representation of graph is learned in the form of node embeddings. My question is: Are the output values back propagated and influence learned node embeddings? The paper I read seems to not account for output being backpropogated so that embeddings get learned. So does that mean learning node embeddings is agnostic of output values during training?
$\begingroup$
$\endgroup$
Add a comment
|