Should ModelSystem allow for passing more args to score function?

[SRFU-NN]

I would like the ability to control whether my score function prints where it has been evaluated and what the result was - It is useful when making examples.

I think the best way to do it is just like we handle acquisition function args - Allow for ModelSystem to have a score_kwargs property that is a dict (or None, which is translated to {}), and ModelSystem.get_score() then calls self.score(X,**self.score_kwargs).

Is this too complicated functionality to solve my problem?

[dk-teknologisk-mon]

I'm not sure I understand why you want a function that is meant to return "data" when queried to keep a memory of the results it has returned? It adds a lot of code complexity to the ModelSystem class, when you could just as easily keep a "data points" array that you continually append or grab them from your Optimizer, which stores both the positions and the scores anyway?

[SRFU-NN]

I don't want to store the data. I want to print it to console when it calculates it, so my jupyter notebooks can report on their progress.

But I don't want to do it every time - I want to be able to plot the entire map without printing 10 000 results.

But perhaps it is best done by having two functions? I will consider this option more.

[dk-teknologisk-mon]

Well, depending on the exact context there are many ways around this problem that seem simpler to me than expanding the ModelSystem class. If you want the iterative optimization step to report on its progress every N iterations, you could just keep a running counter of how many points you have sampled and have a command ala print("Best result after {} iterations is {} at X: {}".format(some-stuff-here)) that gets executed if len(scores) %N == 0?

[SRFU-NN]

Conclusion: no, it should not.