How about the XRAI which is based on Integrated saliency
DLyzhou opened this issue · 0 comments
DLyzhou commented
Thanks for your great work!
The XRAI might be a more intuitive interpretability tool for CNN models, which is based on the Integrated saliency.
Paper: https://arxiv.org/abs/1906.02825
TF 1.x version: https://github.com/PAIR-code/saliency