[MRG+2] fix warning and behavior in randomized_svd wrt power iterations #7311
Conversation
|
ping @giorgiop was that the intended behavior? I changed the default value of Also, now, if a user gives a particular integer, we will actually do what they said! |
|
LGTM. The code is indeed much nicer than with the n_iter_specified. Do we have an idea, a track record of any particular reason that we ended up with this variable (n_iter_specified)? Cc @giorgiop |
|
@amueller Before merge, you have a docstring that you need to fix: it gives a failing doctest. |
dd1e77d to
970ace9
Compare
| is set to 7, unless the user specifies a higher number. This improves | ||
| precision with few components. | ||
| problems. When 'auto', it is set to 4, unless `n_components` is small | ||
| (< .1 * min(X.shape)) `n_iter` in which case is set to 7. |
There was a problem hiding this comment.
maybe i'm being slow, but i'm not sure what this sentence means.
There was a problem hiding this comment.
Maybe the following would be clearer:
When `n_iter='auto'`, it is effectively set to 4 to when `n_components` is larger than
`0.1 * min(X.shape)` and set to 7 otherwise.
|
LGTM as well besides failing doctest / the sentence I can't comprehend. |
|
Besides the failing doctest, LGTM as well. |
|
@amueller The failings come from the |
|
I rebased, fixed the doctests and merged. |
|
thanks @ogrisel |
Trying to fix the warning mess left by #5299 and #6749 (issue #6746).