Extrapolation: confine 2-dimensional plots to convex hull of inputs?

[brandongreenwell-8451]

Would it be feasible to confine two-dimensional plots (i.e., for pairwise interactions) from show(ebm.explain_global()) to the boundary of the convex hull between the data points of the two variables in question? I've found this to make it much easier to avoid interpreting such graphs in areas where there is little to no data. Example plot below (see https://doi.org/10.32614/RJ-2017-016 for details and more examples).

[paulbkoch]

Hi @brandongreenwell-8451 -- This would be a very nice feature. I'm mostly focused on lower-level improvements to EBMs at the moment, but a PR that adds this would be welcome if you or someone else is interested in contributing it.

[bgreenwell]

Thanks @paulbkoch, I’ll see if I can pull something together.