Remo
`tkinter` should already be a part of your python distribution Are you running Python 2.7? In python 2.7 `tkinter` is called as `Tkinter`. Try rerunning your code in Python 3....
I've gotten this error before when an explainer could not take a sparse vector representation (not deep explainer). Are you using a dense array or a sparse array? If the...
A numpy array should be dense already, if you're not sure that it is, you can call `your_data.toarray()`
The masker error should be related to the background distribution you pass. I think it might be solved in shap == 0.41 There are a few things I would try:...
One last thing, regarding this warning: `WARNING:tensorflow:Layers in a Sequential model should only have a single input tensor, but we receive a ` Are you passing the input as a...
Yes, if you are using a 3D array, then this is likely to be the issue. You can copy what the authors do in the documentation and define a wrapper...
whether a feature's value is -.23 of .23, its absolute contribution in terms of percentage will be the same, you can just transform your features to absolute values and calculate...
Hm, I'm struggling to see how you would calculate percentile contribution any other way in this case. Percentages might just not be a good measure for this scenario. Does it...
So, I'm not sure why @slundberg 's says that you cannot simply use absolute values here, but that's neither here nor there. Regarding your proposed solution: I don't see how...
Have you tried `shap.plots.bar(shap_values[:, :, 'mix'])`?