[MRG + 2] Sample-weight support for GaussianNB #4346
Conversation
sklearn/tests/test_naive_bayes.py
Outdated
There was a problem hiding this comment.
I think I'd prefer a test where you explicitly construct a dataset with duplicate points, so that not all sample weights are the same.
There was a problem hiding this comment.
I've extended the test by comparing a dataset with duplicate entries to a dataset without duplicates but correspondingly increased sample weight
sklearn/naive_bayes.py
Outdated
There was a problem hiding this comment.
Shouldn't it be " None for uniform weights"? Not sure what the docstring usually says ^^
There was a problem hiding this comment.
This is copy-paste from BaseDiscreteNB to keep docstrings consistent in the module
|
I've refactored _update_mean_variance() to reduce repetitions and added comments |
|
Great, thanks :) LGTM |
sklearn/naive_bayes.py
Outdated
There was a problem hiding this comment.
shape (n_samples,)
with comma. It's a tuple with 1 element
|
besides +1 for merge thx @jmetzen ! |
FIX Converting sample_weight.sum() to float REFACTOR Reorganizing and commenting _update_mean_variance() in GaussianNB
TST Test GaussianNB sample-weight by somparing with duplicate entries
|
Docstring inconsistencies have been addressed |
|
Is the backticks commit from this PR? Otherwise could you rebase on master? |
|
Also, please add an entry to whatsnew under a new 0.17 heading. |
|
added to whatsnew and pushed. |
|
Thanks for merging! I was too slow ;-) |
|
Or I was too fast ;) Thanks for the contribution! |
This PR adds support for sample weights to GaussianNB.