search

christophM / interpretable-ml-book

Book about interpretable machine learning
https://christophm.github.io/interpretable-ml-book/
Other
4.73k stars 1.05k forks source link

fixes #245 #259

Closed christophM closed 3 years ago

martinju commented 3 years ago

Thanks, for your willingness to improve this minor element in your book. Although I would have done some tiny adjustment to the text, I think your suggestions are more or less OK. (I don't think I can commit to this PR?)

In any case, as mentioned over in #245, I think the real issue here is in the Shapley section (5.9), not the SHAP section (5.10). I would be happy to contribute a PR with a small paragraph on dependence vs independence for Shapley values at least referencing the "true to the model vs true to the data" paper. This could be a new subsection 5.9.3.4 to replace the current text under disadvantages. The new text in this PR should then reference that. What do you think @christophM?

christophM commented 3 years ago

Sounds good. Just fork the repo and open a new PR from you fork.

Thanks a lot!