I'm wondering if the output of G = graphtools.Graph(X,decay=None) is working as intended.
My impression was that with decay=None, a kNN graph would be created.
However, inspection of G.K reveals that the kernel matrix is still weighted - some edge weights are 1, other 0.5.
Is this the intended behavior? It probably arises from the symmetrization of neighbors that are not mutual.
I'm wondering if the output of
G = graphtools.Graph(X,decay=None)is working as intended.My impression was that with
decay=None, a kNN graph would be created.However, inspection of
G.Kreveals that the kernel matrix is still weighted - some edge weights are 1, other 0.5.Is this the intended behavior? It probably arises from the symmetrization of neighbors that are not mutual.