search

shap / shap

A game theoretic approach to explain the output of any machine learning model.
https://shap.readthedocs.io
MIT License
22.76k stars 3.28k forks source link

Possible to include Anchor? #32

Closed keithcallenberg closed 6 years ago

keithcallenberg commented 6 years ago

A new method was released by the authors of LIME. Would there be any value in adding this somehow to the SHAP package?

Paper title: "Anchors: High-Precision Model-Agnostic Explanations" Paper: https://homes.cs.washington.edu/~marcotcr/aaai18.pdf Source: https://github.com/marcotcr/anchor

slundberg commented 6 years ago

Anchor is a great method, but it does not easily fall in the class of additive feature attribution methods unified by SHAP, and I would like to keep the shap package focused on doing one thing well at the moment.

That being said...there might be a way to interpret SHAP values in order to make a "anchoring statement". That could be useful for a variety of reasons, including that (last I talked with Marco about it) Anchor currently just focuses on classification problems and not regression.

keithcallenberg commented 6 years ago

Makes sense! Thanks.