MRG FIX for iid weighting in grid-search #1731
Conversation
| @@ -146,7 +146,7 @@ estimator during the construction and exposes an estimator API:: | |||
| >>> clf.fit(X_digits[:1000], y_digits[:1000]) # doctest: +ELLIPSIS | |||
| GridSearchCV(cv=None,... | |||
| >>> clf.best_score_ | |||
| 0.988991985997974 | |||
| 0.98899999999999999 | |||
There was a problem hiding this comment.
We will get rounding errors: computers have finite precisions. We cannot expect all platform to round stuff exactly the same way at the 17th place, can we? I think we should expect:
0.9889999999999... # doctest: +ELLIPSIS
There was a problem hiding this comment.
Ok, then I'll change all the tests in the file...
There was a problem hiding this comment.
The other numbers in the file seemed not to be periodic, so I kept them.
|
That looks good. Maybe @agramfort can you check this fix? I think you were the original author of the iid flag. |
|
rebased to fix merging trouble |
| X, y = make_blobs(centers=[[0, 0], [1, 0], [0, 1], [1, 1]], random_state=0, | ||
| cluster_std=0.1, shuffle=False, n_samples=80) | ||
| # split dataset into two folds that are not iid | ||
| # first on contains data of all 4 blobs, second only from two. |
|
This all looks well to me and the test is good too. it's ready for merge, right? :) |
|
Yes, should be good. I agree with @ogrisel that it would be nice to have @agramfort's 👍 but I'm not sure he is available / has time. |
|
merged by rebase after fixing the typo. |
|
Thanks :) |
This is a fix for the iid weighting in GridSearchCV. It could be that I screwed this up before, but didn't notice as there was no test.