Francesco Bruzzesi

@MBrouns before closing the issue, could it be worth adding an example in the docs?

Having a closer look at this: actually the two approaches are a bit different. The implementation of ``` for each parameters in grid: fit model with parameters for each threshold...

Thanks for raising the issue. In principle I don't see any blocker in allowing a mapping `{class_label: n_components}` as input in place of a single fixed value.

@koaning do you mean to make explicit in the docs that `GMMClassifier` _does not_ support a variable number of components?

I started to take a look into this. Maybe I am missing something, but by changing ```diff if self.use_global_model: - self.fallback_ = clone(self.transformer).fit(X_value) + self.fallback_ = clone(self.transformer).fit(X_value, y) ``` and...

If there is still interest in this feature, I would be happy to give it a try, this looks like a nice feature to have. However I have a couple...

Hi @samuelcolvin, if there is still interest in this issue I can take it! Any drawback in using `math.isnan` and `math.isinf` for the checks?

I would consider a more general `FloatInf` as well (that's the use case it brought me here in the first place). Just to make sure on naming convention, they should...

Fixed in #626

`Bernoulli` probs are [checked to be in the range `[eps, 1-eps]`](https://github.com/jmschrei/pomegranate/blob/master/pomegranate/distributions/bernoulli.py#L60), where `eps = torch.finfo(torch.float32).eps`. Hence `Bernoulli([0.0])` and `Bernoulli([1.0])` will always fail to initialize. I have no idea if there...