making mutual_info_regression more verbose

[arfogg]

Describe the workflow you want to enable

Hi,

Thanks for coding up mutual_info_regression, I use it for novel space science research and it's super helpful for me.

I'm currently using it to analyse quite large amounts of data, and understandably this can be quite slow. I'm wondering if it would be possible to implement an optional flag in mutual_info_regression to make the code verbose while it's running. For example, printing out something to the terminal when it is 10%, 20%, 30% (etc) done.

Just an idea - hopefully I have raised this in the correct way!

Thanks,
Alexandra Fogg

Describe your proposed solution

Implementation of an optional flag in mutual_info_regression to make the function verbose, where it prints out statements at certain stages of the process, e.g. 10% done or saying that a certain stage is completed, etc.

Describe alternatives you've considered, if relevant

No response

Additional context

No response

[betatim]

I don't know if there is precedent for having progress bars in scikit-learn. I don't think so? Unless there is precedent I think we shouldn't add it. I think if people want a progress bar or some other kind of "spinner" it should be implemented "on top of" scikit-learn. The reason for this is that there are plenty of things in scikit-learn that could/should display a progress bar and then you have progress bars inside of progress bars, etc. And you need to somehow choose when to show a progress bar, etc.

All these things are in some way easier to do in a library that uses scikit-learn (showing just one progress bar, knowing which bits are slow, etc).

[glemaitre]

We show probably first get the callback: #22000

Then, we could use the progress bar callback.