I have just finished SHAP chapter, and decided to dig deeper in the method. In the original paper and library, I found some model specific method such as gradient, DeepLIFT and its combination with SHAP called DeepSHAP. They yield good results, and reasonable explanation. I suggest that you can add them to Neural network interpretation chapter.
I have just finished SHAP chapter, and decided to dig deeper in the method. In the original paper and library, I found some model specific method such as gradient, DeepLIFT and its combination with SHAP called DeepSHAP. They yield good results, and reasonable explanation. I suggest that you can add them to Neural network interpretation chapter.