ENH Add Array API compatibility to KernelCenterer
#27556
Conversation
|
|
||
| K_pred_cols = (np.sum(K, axis=1) / self.K_fit_rows_.shape[0])[:, np.newaxis] | ||
| K_pred_cols = (xp.sum(K, axis=1) / self.K_fit_rows_.shape[0])[:, None] |
There was a problem hiding this comment.
can be replaced with xp.newaxis for readability
There was a problem hiding this comment.
I actually tried with xp.newaxis initially, but I got an AttributeError both with torch and numpy
AttributeError: module 'array_api_compat.torch' has no attribute 'newaxis'
AttributeError: module 'numpy.array_api' has no attribute 'newaxis'
@betatim do you know why newaxis may not be included?
There was a problem hiding this comment.
Unfortunately I don't know why newaxis doesn't exist for them. I would have guessed "bug in array_api_compat" if that was the only place where its missing, but I think ofnumpy.array_api as the "very much exactly what the standard says" implementation. One option is that it is something that only got added in a newer version of the standard
There was a problem hiding this comment.
I opened data-apis/array-api-compat#64 upstream.
Ok for using None in scikit-learn in the mean time.
Reference Issues/PRs
Towards #26024
What does this implement/fix? Explain your changes.
It makes the
KernelCentererimplementation compatible and tested with the Array API.