Add features argument to only include certain features

[mnwright]

Example:

library(glex)
library(xgboost)

x <- as.matrix(mtcars[, -1])
y <- mtcars$mpg

xg <- xgboost(data = x, label = y, params = list(max_depth = 4, eta = .1), nrounds = 100, verbose = 0)

glex(xg, x, features = c("cyl", "disp"))

The SHAP values are not correct anymore if we use this. But we should have the same problem with max_interaction?

[jemus42]

Interesting idea, kind of a natural extension to the max_interaction situation I guess.
We could analogously calculate a remainder term, or add to the existing remainder term.
It becomes a bit wonky at the point where we limit glex() in both max_interaction and included features, at which point the remainder term sucks up both higher order interactions and entire features 🤔

Off the top of my head I'm not sure if I completed the remainder term stuff for both xgboost and rpf or just rpf, but I'd like to note that handling that is unfortunately quite fiddly, especially for the multiclass case 🫠

[jemus42]

Is this also superseded by #17?

[mnwright]

Yes, just a small thing that is in #18 now.

[mnwright]

Wait, what about the tests?

[jemus42]

The tests in "tests/testthat/test-glex-xgboost.R" are in master, seems about right.