interpretable-ml-book
interpretable-ml-book copied to clipboard
Unclear notations in chapter 9.6 SHAP
Hi, I found it very confusing when read about the notations in SHAP chapter. Here are the details:
- In chapter 9.6.1, the SHAP explanation is definded with a function includes $z'_j$, where $j$ here denotes the index of the feature. However, in chapter 9.6.2 Kernel SHAP, there is a $k$ in the sample coalitions $z'_k$, where $k$ means the index of the coalition. For consistancy, maybe the k could be a upper notation like $z'^{(k)}$ to denote this is a vector of $k$ th coalition, not a singular in the vector.
- No clear definition of $X_C$ and $X_S$ in chapter 9.6.2. I suppose $X_C$ is the current instance of interest, and $X_S$ is the randomly sampled instance. Not sure about that, please correct me if it's wrong.