Hi I'm wondering if there's any plan or roadmap for scikit learn inspection package development. I notice that there has been effort in creating a generalized partial dependence plot from #12599. But as suggested in the book Interpretable Machine Learning, PDP is usually used together with some other interpretation methods, such as Individual Conditional Expectation (ICE), but which are missing in current scikit package.

An existing project Skater aims to create a library for model inspection but doesn't seem to be well-maintained. Another one eli5 focuses more on the model-specific methods for now. I think it'll be helpful to have these inspection methods for black-box learning models as part of sklearn package.

I'd be happy to contribute to this if it's in development.