Question about LIME results
ms1948 opened this issue · comments
Hi,
I am trying to use LIME on R, and I need some explanations about the results produced by the package.
I have 144 records with 25 parameters each. Records are labeled “Cat” or “C3” depending on the parameter's value.
If I run a LIME analysis with the command: model <- train(b_train, b_lab, method = 'rf') two times with the same test record I receive two different results (see annex).
Please can you explain to me this behaviour ?
I guess that's because of the instability of the LIME framework.
https://towardsdatascience.com/instability-of-lime-explanations-3e0efc00a7de
Personally, I think if your predictions have no noise then they should be stable. Not sure about this though.