MNT fix failing slope test

[jolars]

Context of the PR

I think the failing test for SLOPE observed in #282 was due to the Slope estimator
normalizing data by default.

Contributions of the PR

I've turned off normalization in the test now.
(I realize that it might not be completely kosher to normalize in these sklearn
estimators so I'll change the default for the sortedl1 package as well, but
we might as well make it explicit here).

The test is working at least locally for me now.

Checks before merging PR

• added documentation for any new feature
• added unit tests
• edited the what's new (if applicable)

[mathurinm]

flake is going to complain about extra spaces around equal here and line below

[mathurinm]

@jolars There are 6 failing not due to Slope, but the last one is because the shape mismatch, you need to squeeze the coefs of libslope:

(shapes (10,), (10, 1) mismatch)
 ACTUAL: array([ 1.267978,  0.254253, -0.641967,  0.406315, -0.147703,  0.451324,
       -0.025444, -0.292776,  0.158982, -0.172514])
 DESIRED: array([[ 1.267984],
       [ 0.254267],
       [-0.642008],...

also probably the numerical tolerances are too low

[jolars]

@jolars There are 6 failing not due to Slope, but the last one is because the shape mismatch, you need to squeeze the coefs of libslope:

(shapes (10,), (10, 1) mismatch)
 ACTUAL: array([ 1.267978,  0.254253, -0.641967,  0.406315, -0.147703,  0.451324,
       -0.025444, -0.292776,  0.158982, -0.172514])
 DESIRED: array([[ 1.267984],
       [ 0.254267],
       [-0.642008],...

also probably the numerical tolerances are too low

Sorry! I manage to miss this due to not having the package installed 😬

I realized that I am also using arguments that are just in the development version. I suppose the reason it's not failing here is that you're installing straight from the github repo and not pypi. Let me push a new version to pypi and then I'll tidy this up.

[jolars]

I believe the SLOPE tests are fixed now. Let me know if this is fine and I'll squash the fixup commits.

[mathurinm]

@jolars thanks for the PR !

No need to squash on your side, we perform "squash and merge" directly on the UI