Dear author:
Thank you for your distribution in tIme series explaining!
I have read your paper “Explaining Time Series Predictions with Dynamic Masks” carefally. To the best of my knowledge, many datasets that contain time series include some CATEGORICAL FILEDS. Some perturbation methods in your paper like Gaussian blur and moving windows may be more useful and reasonable for numerical filed.But when it comes to categorical filed with some int input, it may not work very well.Did you think about how to perturb categorical feature in this framework? Could you share your idea with me?
Looking forward to your reply!
Dear author:
Thank you for your distribution in tIme series explaining!
I have read your paper “Explaining Time Series Predictions with Dynamic Masks” carefally. To the best of my knowledge, many datasets that contain time series include some CATEGORICAL FILEDS. Some perturbation methods in your paper like Gaussian blur and moving windows may be more useful and reasonable for numerical filed.But when it comes to categorical filed with some int input, it may not work very well.Did you think about how to perturb categorical feature in this framework? Could you share your idea with me?
Looking forward to your reply!