Bruno Kacper Mlodozeniec

No worries, let me know if there is anything I can do to explain and document it better in the code.

To answer the two points: > 1. This PR introduces a new method that isn't declared in any model interfaces. That's backwards from Emukit's point of view. Are you planning...

@apaleyes sorry, resolving the interfaces has been on the stack of my todos for the past couple of weeks, I will try to get around to it soon. Yes, this...

Thanks for the follow-up @apaleyes! I'm still happy to finish off this PR (if I don't do it before the end-of-year, feel free to close it). I've been a bit...

I added an interface that specifies the covariance between points derivative method. Let me know your thoughts! As for the new acquisition function, I think I'll be able to add...

I'd be curious to hear how others approached this topic, but I can chime in with the approach I've been using. I've defined a new _batch_ candidate calculator, and a...

One small possible change would be to cache hyperparameter samples from `model.generate_hyperparameter_samples()`. This way each new instantiation of `IngregratedHyperparameterAcquisition` wouldn't needlessly regenerate the samples unless the model's been updated.

I guess the main point was that combining `LocalPenlizationPointCalculator` and `IntegratedHyperParameterAcquisition` can lead to unexpected (although logical) and undesirable behaviour, since `LocalPenlizationPointCalculator` won't marginalise out the samples when computing the...

If the idea is that it's on the user to construct a batch point calculator that marginalises out all the HMC samples when computing the penalized acquisition, I was wondering...

As an aside, this means that the package doesn't work with Python 3.12 and above, since `pip