This is the 5th of 52 posts where I discuss how to bring academic research into DS industry. The primary source is this 2020 paper from and other resources are linked in the post.

As the authors mentioned, the most useful next step would be to develop this method out for non-linear classifiers and continuous variables. However, because it's such a new technique, building out simulations and a codebase would also be quite useful.

If you have any corrections/questions/ideas, leave a comment or message me.