Closed SebastianSemper closed 5 years ago
I have some questions and comments about the changes implemented that I would require to be discussed or resolved before proceeding:
largestSingularVal
and not go the full way to consistent naming largestSingularValue
. I find the latter much more intuitive for a user as it is more consistent with the usual naming throughout the package and what the user would expect.largestSingularVecs
(should be largestSingularVectors
), largestEigenVal
(should be largestEigenValue
) and largestEigenVec
(should be largestEigenVector
)largestSingularVecs
of type tuple? I would expect it to be a 2d ndarray, which would also allow slicing and would be consistent with the rest of the package (uniform interfaces, largestEigenVec
is a ndarray)largestSV
and largestEV
hinting the user to the API modifications, not hard API breaks.I added an explanation of the largestSingularVector property to its doc string. right now it is the closest thing to numpy and scipy, since they both return a tuple with the same ordering of (u,v).
We use the linear operator for all sorts of things now.
Especially to calculate largest eigen and singular values as well as the corresponding vector.
This pull request fixes a long and lasting series of complications with the power iteration.