Can a project be in two categories?
paulbkoch opened this issue · 1 comments
I'm one of the developers of InterpretML. InterpretML is currently listed in the interpretability category, which makes sense and is also our primary focus. Recently though, we have added the ability to create differentially private EBM models (DP-EBMs) which we and at least one other outside paper have found perform significantly better than other known differentially private models. InterpretML would therefore seem to fit into the privacy category as well.
Papers:
http://proceedings.mlr.press/v139/nori21a/nori21a.pdf
I encountered similar "classification issues" in my own best-of-atomistic-machine-learning list. After some thought, I'd argue against multiple categories. People may request a package to belong to not only two categories, but three or more. This clutters the list. Instead, extend tagging and view categories as merely emphasized tagging. In the case of InterpretML, this would mean that the package stays in the category interpretability, but gains the tags interpretability, privacy.