Roman Yurchak

> Need high-performance matrix library that abstracts away sparse and dense (SIMD) matrices. For sparse maybe https://github.com/vbarrielle/sprs ? They would be happy to accept SIMD improvements I think.

In [latest output of Travis CI](https://travis-ci.org/RJT1990/pyflux/jobs/189963864) it looks like there is another dependency missing for coverage with py.test, which can be fixed with ```bash pip install pytest-cov ``` or by...

To clarify, given an arbitrary classification or regression pipeline, currently to fit it with sample weights one has to do something like, ```py kwargs = {pipe.steps[-1][0] + '__sample_weight': sample_weight} pipe.fit(X,...

> For the one expecting, sample_weight to be dispatch everywhere, it will make them angry. At least now, we just annoy everyone. Right, I understand SLEP006 is much more general...

> What I would personally expect is neither 1. or 2., but rather 3. that pipeline passes sample_weight to all steps and raises an error if any of the estimator...

> It's a (debatable) choice Or it might not be when going back to the definition of sample weights. If setting some of them to zero should be equivalent to...

> if sample_weight is passed to fit in pipeline then all steps need to support it. > > to me this ensures backwards compatibility. It cannot silently behave differently. I...

> Just a note that the if all steps support sample_weight is not trivial since a meta estimator may accept **kwargs and we don't know if sample_weight is really supported...

Interesting. > if tags are to be used in the code, we need deprecation cycle / warnings to indicate developers that they need to set the tags on their own...

Indeed, we should add, ``` X = _enforce_estimator_tags_x(X, estimator) ``` to `check_regressors_train` together with running that check on your example estimator to `scikit-learn/sklearn/utils/tests/test_estimator_checks.py`, since apparently scikit-learn has no regressors which...