SAGPool implementation and Documentation difference

[arake001]

Hello!

I have a question regarding the implementation of sagpool and the documentation provided for the sagpool.

`attn = x if attn is None else attn
attn = attn.unsqueeze(-1) if attn.dim() == 1 else attn
score = self.gnn(attn, edge_index).view(-1)

if self.min_score is None:
score = self.nonlinearity(score)
else:
score = softmax(score, batch)

perm = topk(score, self.ratio, batch, self.min_score)`

Here, the score values given to the topk pooling are after applying the non-linearity function. But, the documentation suggests the score before applying the non-linearity function.

Documentation:

y = GNN(X, A)
I = topk(y)
X' = (X dot product tanh(y))

Can anyone explain this?

Thank you in advance for your help.

[wsad1]

topk(y) and topk(tanh(y)) will return the same set of nodes, in other words applying tanh will not change the ranking. Because tanh is a increasing function i.e is if if y1>y2 then tanh(y1)>tanh(y2).