Jacob

Sounds good to me too, thanks for looking into this @remiadon

My two cents: Looking into what is currently available in `skmultiflow` and what `streamz` brings into the table I see (at a high level): `streamz` provides a step forward to...

> > Stream metadata is rather important/useful on the learning side > > That much ? To me the ideal world would be to have skmultiflow provide models which are...

Considering the discussion so far, I rather not rush this feature into the next release (0.3.0). I propose to target it for 0.4.0 so we have more time to put...

Hi, unfortunately, there is currently no active effort on this. The topic is still relevant, but we have not enough resources at the moment.

Thanks for the complete discussion @smastelini. `RegressorChain` does not perform explicitly data normalization, this is performed by the base estimator **if applicable**. For example, if you use `RegressorChain` with `RegressionHoeffdingTree`...

Thanks for reporting this @ginop The bug comes from a typo in the internal variable name: ```python self.max_wind_size = max_window_size ``` The `reset` method fails to set the `max_window_size` resulting...

Hello @luccaportes >Would something like this be possible in the scikit-multiflow? Yes! And I agree with you on the potential of this functionality. I also like the idea of having...

+1 for this. My comments: * `BaseEnsemble` would inherit from both `MetaEstimatorMixin` and `BaseEstimator`. * For compatibility with `scikit-learn` we should use the same variable name `ensemble_` * For completeness,...

I am then assigning the issue to you @luccaportes. Let us know if we can help.