David Rudel

I've never done one before, but sure I'd be willing to give it a try.

I have mitigated this by the following patch: ``` X, y, sample_weight, output_weight, missing = self._scrub( X, y, None, None, missing) if y.shape[1] == 1: y=y.ravel() y_hat = self.predict(X, missing=missing)...

Looks like this is more compatible with the rest of the code: ``` X, y, sample_weight, output_weight, missing = self._scrub( X, y, None, None, missing) y_hat = self.predict(X, missing=missing) if...

@jcrudy Thanks for the response. A couple of notes: Reshaping the output really is not a tenable solution because the mathematical operations end up providing a result that is mostly...

@jcrudy , regrettably in its current state reshaping the input also doesn't help. Regardless of the shape of y sent to score_samples, the scrub function will return a 2d-vector so...

@jcrudy I have put a pull request in. You were interested in use cases for score_samples. I'm happy to provide two. (I think it is a pity that scikit-learn does...