Closed xia0pin9 closed 6 years ago
This is what happens in some kernels when having to do with corner cases (i.e. empty graphs etc).
It is on the hand of the user to deal with the Inf/NaN on the result where the standard approach is to apply the numpy.nan_to_num
function on the resulting kernel matrix.
Got the following RuntimeWarning when using graphlet_sampling kernel:
It is probobally caused by INF/NaN value. Not a major problem, but may correspond to some edge cases.